3 Student Conference on Operational Research

3rd Student Conference on
Operational Research
Nottingham, UK
20-22 April 2012
Programme and Abstracts
Directions
A very warm Welcome to SCOR 2012
We are delighted that you are joining us at the University of Nottingham for the 3rd Student Conference on Operational Research
(SCOR 2012). We have delegates who have travelled from across Europe and from further afield and we are glad to see you have
all arrived safely.
SCOR 2012 is a conference aimed at students who are studying Operational Research, Management Science or a related field.
In planning this conference, our objective has been to provide a friendly environment in which students can share their work,
practice their presentation skills, receive constructive feedback, and meet others who are researching similar topics.
In addition to the high calibre of student speakers, we are very pleased to have five excellent invited speakers: Gavin Blackett
(The OR Society), David Buxton (DSE Consulting Ltd.), Tony O’Connor (GORS), Vincent Knight (Cardiff University) and
Louise Orpin (The OR Society).
We have a full conference programme over the next three days, so please take a few minutes to read through the important
information provided in this pack. The conference is fully catered, with the conference dinner being held on Friday evening.
Please ensure that you read the information on page 6 regarding the location and provision of other meals.
We hope you have a great time at the conference here in Nottingham. If we can do anything to help you over the weekend then
please let us know. Thank you for joining us at SCOR 2012.
The SCOR Committee 2012
Directions
The conference will take place at the University Jubilee Campus. This is NOT the main campus of the university, but a little
closer to the city centre of Nottingham.
The University of Nottingham
Jubilee Campus
Wollaton Road
Nottingham
NG8 1BB, UK
Phone: (+44) 115 846 6543
Fax: (+44) 115 846 7877
Contact numbers during the conference:
Stefan Ravizza: (+44) 796 490 8133
Penny Holborn: (+44) 782 572 9540
Urszula Neuman: (+44) 784 561 4083
Detailed maps and directions can be found at:
http://www.nottingham.ac.uk/about/visitorinformation/mapsanddirections/jubileecampus.aspx
Travelling from East Midlands Airport:
From East Midlands Airport you can take the Nottingham City Transport’s Skylink service. Buses leave from outside the Airport
Arrivals hall. The journey takes 60 minutes and buses are every 30 minutes until 11.30pm, then hourly until 5am when the half
hourly service resumes. You can also walk to the taxi rank on the terminal forecourt and take a direct taxi to the university. The
cost of a single/one way journey is approximately £20. Taxis are normally available 24 hours.
Travelling by Train:
From London: St. Pancras. Tickets between London and Nottingham are available from the national rail website. There are also
regular services to Nottingham from Birmingham, Derby, Leicester, Crewe, Sheffield, and Leeds (http://www.nationalrail.co.uk/).
Turn right out of the main station exit (not the smaller, Station Street exit) for an easy 5 minute walk to the city centre. The
Broadmarsh bus station is on the right, after approx 250m (Carrington Street).
From M1 Motorway:
Leave the M1 motorway at Junction 25 to join the A52 to Nottingham. After five miles turn left onto the A6514, Middleton
Boulevard. Turn right at the next roundabout onto the A609 Wollaton Road. The main entrance to Jubilee Campus is clearly
signposted on the right.
SCOR 2012
3
Conference Registration and Accommodation
Travelling from Nottingham:
There are a number of bus services running from Nottingham to Jubilee Campus (“the Two", 28 and 30). More information can
be found on the Trent Barton Website: http://www.trentbuses.co.uk/university/index.html.
There are taxi ranks throughout the city and immediately adjacent to the main railway and bus stations. The journey to the
campus takes approximately 15 minutes.
Travelling by Taxi to Jubilee Campus:
There are taxi ranks throughout the City Centre and immediately adjacent to the main railway and bus stations. The journey to
the campus takes about 15 minutes.
Taxi Companies:
DG Cars
Trent Cars
Nottingham Cars
Fast-Lane Cabs
(+44) 115 960 7607
(+44) 115 950 5050
(+44) 115 970 0700
(+44) 115 950 1501
Parking on the Campus
Parking in the car park in front of the halls of residence is available for SCOR students for the duration of the event (nonNottingham students only). When you arrive at the campus please indicate to the security gate that you are here for the SCOR
conference and they will provide you with a security pass for your car as well as directions to the halls. Please note that the
accommodation will not be available on the Friday until after 3pm therefore please proceed to the lecture theatre (for more
information see the accommodation section).
Conference Registration and Accommodation
Delegates are asked to go straight to “The Exchange" (see map on page 8, building no 2) for the welcome buffet and registration,
as access to the halls is only available from 3pm onwards. We have arranged for the bags to be left in a secure room during the
first talks on Friday afternoon. Another room will be used to leave bags on the Sunday during the programme as guests are asked
to check out of the halls before 10am.
The conference will be opened by the Conference Chair, Stefan Ravizza, in “The Exchange" room LT2 at 1.15pm. The first
session of the conference will begin at 2.45pm. For the full conference programme, see page 17.
Southwell Hall (En-suite)
The hall is a 2 minute walk across the campus from “The Exchange" where the initial welcome buffet and registration will be
held. Access can be granted to bedrooms from 3pm onwards. The hall manager or assistant hall manager will check in guests
and give you a room key from the hall reception. Keys are to be handed in to the main reception upon departure and any missing
keys will be charged for. Guests are asked to retain their keys at all times.
Internet Access
These instructions clarify the procedures for visitors accessing the wireless network when the helpline is closed. It also highlights
alternative methods to access the internet while at the University of Nottingham, when not a member of the university. A visitor
is classed as anyone who is not a member of staff or a student of the University of Nottingham, including conference guests,
ad-hoc visitors, etc.
There are three ways to access the internet:
• Wired access in the halls of residence (SNS)
• Public access terminals in the halls of residence and walk-in PC’s in the libraries
• Wireless access in the public areas (UoN-Guest)
The following information is for visitors to the university. Nottingham students should continue to use the information on the IS
website at: http://www.nottingham.ac.uk/is/
4
SCOR 2012
Conference Registration and Accommodation
Wired Access in Halls of Residence (SNS)
Each bedroom has access to the SNS network requiring a data cable which plugs into the data socket. In order to connect, a
visitor should plug a data cable into their laptop and connect it to the data point.
On opening your internet browser, you will be presented with the “Welcome to the Student Network Service" screen. Visitors
should click Visitor to continue, before being asked to confirm you agree to abide by the terms and conditions. The next stage
will be to register your computer. All users will need to provide some details so that the university can contact them if necessary
(for example in an emergency or abuse of the service). By clicking Continue, visitors are assumed to have agreed to these terms
and conditions. After completing this final stage, you will be registered and can use the SNS for up to seven days.
Public Access Terminals in the Halls of Residence and Walk-in PC’s in the Libraries
Each university hall has been supplied with two PCs which connect to the university network, using a pre-supplied username
and password (if these have been lost, the Staff helpline will be able to supply them). These PCs are configured exactly the same
as the computer room PCs on campus.
Wireless Access in the Public Areas (UoN-Guest)
In public areas such as coffee bars or Junior Common Rooms, there is a wireless access point. Guests and visitors can use this to
connect to the internet through the UoN-guest network. When in study bedrooms, the fixed line SNS point should be used (see
section above).
To access the guest wireless service, visitors simply need to select UoN-guest from the list of available wireless networks and
connect. No password or additional configuration is necessary. Once connected, open up a web browser and you will be presented with a page to register onto the network. You must enter a valid email address and agree to the terms and conditions
before being granted access onto the internet.
Notes:
When using the university wireless service, some Android devices can only connect to internal websites and not the internet due
to an issue with proxy settings on some Android devices. To protect the university only certain activity will be permitted on the
wireless network. This includes web browsing and reading emails, but some applications will be blocked. The use of Virtual
Private Networks (VPNs) is not allowed. The network connection is unencrypted. More information can be found on the IS
website at http://www.nottingham.ac.uk/is/wireless.
Hall Facilities
Free use of the fitness centre is available to all residential guests. Room keys are asked to be shown at the reception of each
facility for access. For further information go to http://www.nottingham.ac.uk/sport/unipark.php. Breakfast will be served from
7.30am until 9.00am on Saturday and from 8.00am until 9.30am on Sunday. The menu (unrestricted amount) is as follows:
Grilled Bacon
Fried Egg
Veggie Grill
Preserves
Tea
Water
SCOR 2012
Grilled Pork Sausage
Scrambled Egg
Grilled Tomato
Butter/Flora
Coffee
Baked Beans
Quorn Sausage
Toast
Cereals - minimum of four choices
Fresh Fruit Juice
5
Helpful Telephone Numbers and Emergencies
Meals and Entertainment
Lunch will be served on all three days of the conference on Friday in “The Exchange" and during the weekend in the “Business
School South" (the same buildings where the conference is taking place).
Friday Night
The dinner will be held at 7.30pm in “Fat Cat" (Chapel Quarter, Nottingham, NG1 6JR, see on map). In order to get there you
will need to take a bus to the city centre (detailed information on the Trent Barton Website: www.trentbarton.co.uk). The Chapel
Quarter is at the far end of the Market Square and at the top of Maid Marion Way. Buses stop at the Market Square. After the
dinner social events are planned and this will be a good opportunity for networking.
Menu
Starters
Honey roast butternut squash soup with warm crusty bread
Chicken and bacon Caesar salad with herb croutons
Creamy garlic mushrooms on sea salt and rosemary toasted foccacia
Main Courses
Slow cooked short rib of beef, dauphinoise potatoes and buttered greens
Baked puff pastry tart filled with melting goats cheese, chargrilled vegetables, sweet balsamic dressed leaves
Grilled fillet of lemon pepper hake on a cassoulet of tomatoes, garlic, peppers, beans and spinach
Desserts
Black Forest Torte with mulled black cherries
Chocolate Brownie Sundae
Sticky Toffee Pudding with Custard
Saturday Night
The dinner will be held at 7.30pm in “Peachy Keens" (114 Upper Parliament Street, Nottingham, NG1 6LF, see on map). In order
to get there you will need to take a bus to the city centre (one possible option is to take “the Two" from the “Wollaton, Jubilee
Campus" bus stop and leave at the “Upper Parliament Street" bus stop). “Peachy Keens" is an all you can eat multi-cuisine
buffet.
Helpful Telephone Numbers and Emergencies
Local Taxi Numbers:
DG Cars
Trent Cars
Nottingham Cars
Fast-Lane Cabs
(+44) 115 960 7607
(+44) 115 950 5050
(+44) 115 970 0700
(+44) 115 950 1501
SCOR Organiser:
Stefan (Chairman)
Penny (Vice-Chair)
Urszula (Local organiser)
(+44) 796 490 8133
(+44) 782 572 9540
(+44) 784 561 4083
Anyone requiring an ambulance should dial 8888 on the internal phone rather than a mobile. The call will be channelled through
security who can meet the ambulance and quickly direct them to the scene. If the fire alarm sounds at any time during your stay
with us then you must evacuate the building as quickly and safely as possible by the nearest emergency exit. The duty porter for
that area will arrive with the Fire Brigade as quickly as possible and will give you further instructions.
6
SCOR 2012
Maps and Floor Plans
Maps and Floor Plans
SCOR 2012
7
Maps and Floor Plans
On Friday we will be based in “The Exchange" (building no 2) while during the weekend we will be in the “Business School
South" (building no 7+8). Breakfast is served in “The Atrium" (building no 5).
Woll
aton
Rd.
Western
Boulevard
A6514
Crown
Island
Jubilee Campus
MAIN
ENTRANCE
& University of Nottingham
Innovation Park (UNIP)
24hr Security Gatehouse
Melton
Hall PD
G
Wolla
to
SC
n Ro
ad A
609
Ilke
ston
To city
centre
Roa
d
A60
9
uleva
ton Bo
Middle
1
The Exchange
2
514
rd A6
The Sir Harry & 3
Lady Djanogly
Learning Resource
Centre
To
University Park
Campus
(0.5 miles)
Southwell
Hall
4
SC
The Atrium
5
B
The Dearing
Building
6
Business
School
South
7
Auditorium
8
2
2/5/7/11
11
11
3/7
9
11
11
12
Sports
Centre
Newark
Hall SC
ay
Railw
Computer Science
Academic schools and departments
Centre for English Language Education 10
Contemporary Chinese Studies
10
Computer Science
4
Civil Engineering
13
Education
6/15
International Office
10
Nottingham University
1/7/10/14
Business School
Other services
Banks/Retail
Cafes
Faith/Prayer rooms
Graduate Centre
Libraries
Sports
Student Services Centre
Students’ Union
UNIP reception
SC
Triumph Road
Business
School
North
9
International
House
10
11
13
Amenities
Building
B
Nottingham
Geospatial
Building
IPD
12
B
National
College
entrance
National College
for School Leadership
(NCSL)
Academic buildings
Halls of residence
Sir Colin
Campbell
Building
Other services
Footpath
PD
University of Nottingham
Innovation Park (UNIP)
Sports
ground
15
9 Triumph
Road
PD Pay & Display visitor parking
IPD UNIP pay & display visitor parking
6 Triumph
Road
14
N
Blue-badge parking
oad
ph R
Trium
Tennis
courts
G Gatehouse
B Barrier-access control
SC
Secure cycle parking
Hopper bus stop
0
metres
100
Public bus stop
Public/Hopper bus stop
Pedestrian/cycle route
to University Park
Campus
(0.3 miles)
Building public entrances
Aspire Sculpture
24-hour ambulance/fire/police
(0115) 951 8888
24-hour security contact
(0115) 951 3013
8
To
University Park
Campus
(0.5 miles)
08/2011 © Crown Copyright Licence no. 100030223
De
To city
centre
00
A62
oad
R
r by
UNITED KINGDOM
CHINA
MALAYSIA
SCOR 2012
Plenary Talks
Plenary Talks
Gavin Blackett
The OR Society
About the speaker
Following a maths degree from the University of Bath, Gavin started his working career in the Operational Research Executive
of British Coal in 1988. In 1992, with a part-time MBA under his belt, Gavin (and the rest of his O.R. colleagues) joined what
would become Capgemini, and during his 14 years with the company he worked on O.R. projects in just about every business
sector. In 2006 he gave up the consultancy lifestyle for a fixed location when he became Secretary & General Manager at the
OR Society.
Abstract
Gavin’s talk will focus on the benefits of OR Society membership. The OR community needs a strong professional body, and it’s
not always about what it can do for me! Being an active member of the OR Society is all about promoting a healthy subject area
and ensuring that it remains vibrant. Critical mass is important for many of the Society’s activities - regional societies, special
interest groups and especially accreditation.
David Buxton
Entrepreneurship in Operational Research - Essential bedfellows?
DSE Consulting Ltd.
David Buxton has had a varied career history. Starting out as a Geographer before moving into Analytical roles and then into
senior positions in general management. Recently, David move into academia, before taking the risky step to strike out and
founded dseConsulting. Enjoying commercial success and a reputation as a simulation expert, last year he joined forces to also
launch decisionLab, styled as the modern approach to OR consultancy.
These business ventures and the previous experiences have each given different insights into the role of OR in business, and had
led David to believe that an entrepreneurial spirit in Operational Research is an essential component.
Looking at the definition of an entrepreneur: “someone who identifies an opportunity and organises, operates and assumes the
risk for a business venture to exploit that opportunity“, it is easy to see that this should be the very essence of OR. However, in
practice, we frequently see OR as a ’behind the scenes’ or supporting function? As evidence of this - given the difference our
work can make, and the decisions we work on - wouldn’t we expect to see some of the high flyers of the business and political
worlds coming from an OR background?
So why is it that we don’t? Perhaps we should focus more on risk taking?
And what about those characteristics associated with the most successful entrepreneurs: interpersonal skills, the ability to persuade, the ability to lead and motivate, charisma? Are those skills traditionally valued and developed in OR?
In the end, does it really matter? Perhaps other disciplines are better suited to the cut and thrust of a commercially-driven world
- we can let them take the plaudits whilst we continue with the decision support. But does this threaten to marginalise OR? And
in a competitive job-market what can an OR graduate do stand out?
Using examples from his own experience, David will highlight the importance of the inner entrepreneur. It hasn’t all been plainsailing and in this talk David will share his mistakes as well as successes to provide you with insight to help you prepare for your
own career and recognise the skills you’ll need to move seamlessly from academic to commercial to consultancy environments.
Tony O’Connor
Chair of the Government Operational Research Service (GORS)
Senior Analytical Strategist: Department of Health
Former Chief Analyst for the Prime Minister’s Delivery Unit
Tony has over 25 years experience of OR in Government across Education, Cabinet Office HMT Treasury and since 2009 in
the Department of Health’s Strategy Group. Worked 15 years in the Dept for Education (in its various guises) manly as an OR
analyst, but also in policy, working across a wide range of education policies.
Key central Government role was as the Chief Analyst of the Prime Minister’s Delivery Unit (2001-2008) where he established
a small multi-disciplinary analytical team to improve the use of evidence at the heart of Government, working in particular on
performance measurement across all the key indicators underpinning the then Government’s Public Service Agreements for the
Prime Minister.
In addition since April 2004 he has been the Chair of GORS representing the interests of all the government OR analysts. He has
also been an active member across the wider OR community - regularly attending the OR Soc Conferences, delivering plenary
talks in 2008 and 2010. Also Chair of the NATCOR Advisory Board, and member of Heads of OR Forum (HORF) and of the
Lancaster Management Science Advisory Board.
SCOR 2012
9
Instructions
In 2007 he was awarded the CBE in the Queen’s Birthday Honours for his work promoting Operational Research across Government and his analytical contribution to the Prime Minister’s Delivery Unit. In 2010 he was made an OR Companion of Honour
by the OR Society
Vincent Knight & Louise Orpin
OR in Schools
Vincent Knight, Cardiff University
Vincent Knight is a LANCS lecturer in Operational Research. He is currently chair of the OR in schools taskforce and carries
out a wide variety of outreach activities. His research interests are in game theory and queueing theory.
Louise Orpin, The OR Society
Louise Orpin is the Education Officer at the Operational Research Society. Her role is to raise awareness of OR among young
people and their teachers. OR can help encourage young people to continue studying maths by showing them some applications
of the maths they are learning and also highlights a potential career for someone who enjoys maths.
Abstract
Participants in Science, Technology, Engineering and Mathematics (STEM) subjects need to actively encourage and enthuse
young school children to explore the various possibilities available to them. Operational Research is no different. In this talk we
will present what the OR society already does in means of outreach activities as well as playing a game (audience participation
is required) used to introduce game theory to school children. It is hoped that this talk will encourage and enthuse young
Operational Researchers to participate in such events and further the great work done by the Society to disseminate the subject.
Instructions
For Speakers
• We ask all speakers to be familiar with the time and the location of their stream and talk, as specified in the conference
booklet.
• Speakers should arrive at the location of their stream and talk 10 minutes prior to the scheduled start time of the session.
• Upon arrival you will be met by the chair of the session. Please introduce yourself and, if applicable, provide the chair
with a copy of your presentation to upload onto the seminar room computer.
• Each seminar room will contain a computer equipped for Powerpoint and PDF presentations. Please ensure that you are
familiar with the equipment before the start of your talk.
• Talks are strictly 20 minutes long plus 5 minutes for questions and answers. Anyone going over this time will be asked to
stop by the chair.
• To aid you with the timing of your presentation, the chair will show the ‘time remaining’ cards when you have 5 minutes
and then 1 minute remaining for your presentation.
For Chairs
• Please arrive at the appropriate seminar room 10 minutes before the start of the stream you will be chairing. You should
familiarise yourself with the equipment and ensure there are no obvious problems which would prevent the stream from
running to schedule.
• In the event of a problem you should immediately seek the help of a local conference organiser.
• Delegates presenting in the stream should also be present in the seminar room 10 minutes before the start of the stream.
You should introduce yourself to the speakers. They will provide you with electronic copies of their presentations to be
loaded onto the seminar room computer.
• Uploading presentations: When you arrive at the seminar room you should login to the seminar room computer using the
username and password issued to you at registration. You should then upload each speaker’s presentation onto the desk
top ready for the stream to begin.
• Your main role will be to ensure that the stream runs to time. The speaker has 20 minutes for presentation followed by 5
minutes of questions and answers. Each talk is followed by a 5 minute break for the comfort of the audience and to allow
for movement between streams.
10
SCOR 2012
Smartphone App
• If a speaker fails to show for their talk, advise the audience to attend a talk in an alternative seminar room. Please, do not
move the next talk forward.
• Before each speaker presents, you should introduce them and remind the audience that all interruptions and questions are
to be reserved until the scheduled 5 minute question and answer session following each presentation.
• During each presentation, please use the ‘time remaining’ cards to indicate to the speaker when 5 minutes and then 1
minute of their presentation remains.
• Should a speaker overrun, you must politely but firmly stop their presentation and move on to the question and answer
section of the time slot.
• After each talk, thank the speaker, encourage applause, and open the floor to questions.
Smartphone App
SCOR 2012 has gone mobile using Guidebook!
We encourage you to download our mobile guide to enhance your experience at SCOR. You will be able to plan your day with a
personalised schedule and browse maps and other important information. The free app is compatible with iPhones, iPads, iPod
Touches and Android devices. Windows Phone 7 and Blackberry users can access the same information via our mobile site at
m.guidebook.com.
To get the guide, choose one of the methods below:
1. Download “Guidebook" from the Apple App Store or the Android Marketplace
2. Visit http://guidebook.com/getit from your phone’s browser
3. Scan the following image with your mobile phone (QR-Code reader required, e.g. ’Red Laser’, ’Barcode Scanner’)
The guide can be activated with the Redeem Code “scor2012" under Download Guides.
SCOR 2012
11
Conference Sponsors
Conference Sponsors
The OR Society, www.theorsociety.com
Automated Scheduling, Optimisation and Planning (ASAP) research group, University of Nottingham,
www.asap.cs.nott.ac.uk
12
SCOR 2012
Conference Sponsors
Prospect Recruitment, www.prospect-rec.co.uk
Gower Optimal Algorithms Ltd., www.goweralg.co.uk
Tata Steel, www.tatasteel.com
Banxia Software, www.banxia.com
SCOR 2012
13
FREE STUDENT MEMBERSHIP
FOR FIRST SIX MONTHS*








Free access to O.R. Journals and case studies
Free O.R. publications
Quick literature search
Help with O.R. techniques – www.theorsociety.com has a selection
of learning aids available only to Members.
An opportunity to join Regional Societies and Special Interest Groups
Access to document repository – an online facility to exchange
documents, presentations and opinions with other members
Help in finding a great job in O.R.
…and on successful completion of an O.R. based course, Members can
apply for CandORS accredited status
Share and contribute to the combined knowledge of over 2,500
other O.R. professionals and support other OR Society activities
such as marketing the O.R. profession to the business community
(www.scienceofbetter.co.uk) and promoting O.R. in schools
(www.LearnAboutOR.co,uk)
For more details, and to join on-line, visit
us at www.theorsociety.com
www.theorsociety.com
*If you sign up with Direct Debit, payments will start after six months
HERE’S WHAT YOU GET:
SCOR Committee
SCOR Committee
Stefan Ravizza
University of Nottingham
[email protected]
Conference Chair
Penny Holborn
Cardiff University
[email protected]
Conference Vice-Chair
Michael Clark
University of Nottingham
[email protected]
Emily Cookson
Lancaster University
[email protected]
Magdalena Gajdosz
University of Strathclyde
magdalena.[email protected]
Pablo Gonzalez Brevis
University of Edinburgh
[email protected]
Izabela Komenda
Cardiff University
[email protected]
Urszula Neuman
University of Nottingham
[email protected]
Martin Takáč
University of Edinburgh
[email protected]
Alessia Violin
Universit Libre de Bruxelles
[email protected]
SCOR 2012
15
PROGRAMME
SCOR 2012
10:00 - 11:00
11:00 - 11:30
11:30 - 12:30
12:30 - 13:30
13:30 - 15:00
15:00 - 15:30
09:00 - 10:30
10:30 - 10:45
10:45 - 11:45
11:45 - 12:00
12:00 - 13:00
13:00 - 14:00
14:00 - 15:30
15:30 - 16:00
16:00 - 17:30
17:30 - 19:30
19:30 - 21:30
12:15 - 13:15
13:15 - 13:30
13:30 - 13:45
13:45 - 14:35
14:35 - 14:45
14:45 - 16:15
16:15 - 16:30
16:30 - 18:00
18:00 - 19:30
19:30 - 22:30
Transport I
Graphs/Networks
Transport II
Supply Chain Management II
A26
Supply Chain Management I
Multicriteria Decision Analysis I
Saturday, 21st April 2012
A24
A25
Optimisation II
Heuristics/Metaheuristics I
Break
Reliability/Risk Assesment
Heuristics/Metaheuristics II
Coffee Break
Plenary Talk: Tony O'Connor (GORS) - A25
Lunch
Optimisation III
Heuristics/Metaheuristics III
Coffee Break
Optimisation IV
Scheduling/Timetabling I
Free Time
Dinner @ Peachy Keens
Free Time
Dinner @ Fat Cat and Social Evening
Coffee Break
LT2
A08
Multicriteria Decision Analysis II
Sunday, 22nd April 2012
A24
A25
A26
Optimisation V
Scheduling/Timetabling II
Decision Support
Coffee Break
Plenary Talk: Vincent Knight (Cardiff University) and Louise Orpin (The OR Society) - A25
Lunch
Simulation/System Dynamics
Optimisation VI
Scheduling/Timetabling III
Neural Networks/Machine Learning
Award and Closing Speech: Stefan Ravizza (SCOR) - A25
Healthcare
Stochastic Modelling III
Stochastic Modelling II
A08
Stochastic Modelling I
Friday, 20th April 2012
Welcome Reception and Lunch
Welcome Speech: Stefan Ravizza (SCOR) - LT2
Plenary Talk: Gavin Blackett (The OR Society) - LT2
Plenary Talk: David Buxton (DSE Consulting Ltd.) - LT2
Break
Mathematical Programming
Optimisation I
LT1
Programme Overview
17
Friday, April 20, 2012
12:15–13:15
Welcome Receptions and Lunch
13:15–13:30
Welcome Speech: Stefan Ravizza (LT2)
13:30–13:45
Gavin Blackett (The OR Society) (LT2)
13:45–14:35
David Buxton (DSE Consulting Ltd.) (LT2)
14:35–14:45
Break
14:45–16:15
Optimisation I (LT1)
Chair: Alessia Violin
1. Comparisons between observable and unobservable M/M/1 queues with respect to optimal customer behaviour Rob Shone, Vincent Knight and Janet Williams
2. Optimization of electricity trading using linear programming Minja Marinović, Lena Ðord̄ević, Dragana Makajić-Nikolić and Milan Stanojević
3. Performance measurement and trade-offs in UK higher education institutions Carolyn Booker, John Quigley and Lesley Walls
14:45–16:15
Transport I (LT2)
Chair: Penny Holborn
1. Optimal toll enforcements on motorways Elmar Swarat, Guillaume Sagnol and Ralf Borndörfer
2. The effects of release levels and trajectory lengths on the aircraft arrival sequence Stanislava Armstrong, Jason Atkin, Geert De
Maere and Edmund Burke
3. Optimisation of traffic signal settings at arterials during variable conditions Cesar Velandia-Brinez, Ruibin Bai, Graham Kendall
and Jason Atkin
16:15–16:30
Coffee Break
16:30–18:00
Mathematical Programming (LT1)
Chair: Pablo Gonzalez Brevis
1. Nurse rostering in a Danish hospital Jonas Baeklund
2. A bilevel model for valves location in water distribution systems Andrea Peano
3. The primal-dual column generation method: theory and new applications Pablo González-Brevis, Jacek Gondzio and Pedro Munari
16:30–18:00
Multicriteria Decision Analysis I (LT2)
Chair: Izabela Komenda
1. A nice use for MCDA in public health: potential new approaches to decision making in the national institute for health and
clinical excellence Brian Reddy
2. Compromise solutions Kai-Simon Goetzmann, Christina Büsing, Jannik Matuschke and Sebastian Stiller
3. Lower bound improvements of penalty parameters for discrete - continuous linear bilevel problems Renato Mari
18:00–19:30
Free Time
19:30–22:30
Dinner @ Fat Cat and Social Evening
18
SCOR 2012
Saturday, April 21, 2012
09:00–10:30
Stochastic Modelling I (A08)
Chair: Emily Cookson
1. Stabilizing policies for employment portals Hanyi Chen and Burak Buke
2. Assortment planning with substitution effects Jochen Schurr
3. Decomposition techniques for stochastic unit commitment problems Tim Schulze, Andreas Grothey and Kenneth I.M. McKinnon
09:00–10:30
Optimisation II (A24)
Chair: Pablo Gonzalez Brevis
1. The traveling visitor problem and algorithms for solving it Milan Djordjevic, Marko Grgurovic and Andrej Brodnik
2. A dynamic programming heuristic for the quadratic knapsack problem Franklin Djeumou Fomeni and Adam Letchford
3. Compact formulations of the Steiner traveling salesman problem Saeideh D. Nasiri, Adam Letchford and Dirk O. Theis
09:00–10:30
Heuristics / Metaheuristics I (A25)
Chair: Penny Holborn
1. Heuristics for a split vehicle routing problem with backhaul Michela Lai, Massimo Di Francesco and Paola Zuddas
2. Periodical vehicle routing problem due to driver familiarity Matthew Soulby and Jason Atkin
3. Dynamic vehicle routing problems with pickups, deliveries and time windows Penny Holborn, Jonathan Thompson and Rhyd Lewis
09:00–10:30
Supply Chain Management I (A26)
Chair: Magdalena Gajdosz
1. The impact of on time delivery on supply chain management: are third party logistics providers worth the investment? Yolanda
Silvera and Rebecca DeCoster
2. Forecasting ARMA demand processes: the impact of temporal aggregation Bahman Rostami Tabar, Mohammad Zied Babai, Aris
Syntetos and Yves Ducq
3. The impacts of material convergence on the post-disaster humanitarian logistic operation Nha-Nghi Huynh, Nha-Nghi Huynh,
Sandra Transchel, Maria Besiou and Luk Van Wassenhove
10:30–10:45
Break
10:45–11:45
Stochastic Modelling II (A08)
Chair: Emily Cookson
1. Time-dependent stochastic modelling for ambulance demand prediction and scheduling Julie Williams, Harper Paul, Gillard
Jonathan and Knight Vincent
2. Point process models for assessing the reliability of desalination plant equipments subject to failures due to red tide events
Ahmed Al Hinai and B. M. Alkali
10:45–11:45
Reliability / Risk Assessment (A24)
Chair: Magdalena Gajdosz
1. Reliability assessment of subsea systems in ultra-deepwater oil and gas developments Anietie Umofia, Stephen Okonji and Shaomin
Wu
2. Modelling the impact of human and organisational factors on the safety risk over time Magdalena Gajdosz, Susan Howick and
Tim Bedford
10:45–11:45
Heuristics / Metaheuristics II (A25)
Chair: Michael Clark
1. An improved choice function heuristic selection for cross domain heuristic search John Drake and Ender Ozcan
2. Hyper-heuristic to construct magic squares Ahmed Kheiri and Ender Ozcan
10:45–11:45
Supply Chain Management II (A26)
Chair: Alessia Violin
1. Markovian analysis of stochastic serial multi echelon supply chains Despoina Ntio and Michael Vidalis
2. Modeling and evaluating a tandem supply chain with multiple stages Michail Geranios and Michail Vidalis
11:45–12:00
Coffee Break
12:00–13:00
Tony O’Connor (GORS) (A25)
13:00–14:00
Lunch
14:00–15:30
Stochastic Modelling III (A08)
Chair: Izabela Komenda
1. Hybrid lateral transshipments in multi-item inventory networks Sandra Rauscher and Kevin Glazebrook
2. A chance constrained model for VRPTW with uncertain demands and travel times Pinar Dursun and Erhan Bozdağ
14:00–15:30
Optimisation III (A24)
Chair: Martin Takac
1. Product form of the inverse revisited Péter Tar and István Maros
2. Parallel coordinate descent method for composite objective Martin Takáč and Peter Richtárik
SCOR 2012
19
Saturday, April 21, 2012
14:00–15:30
Heuristics / Metaheuristics III (A25)
Chair: Penny Holborn
1. Evolutionary techniques for an order acceptance problem Simon Thevenin, Nicolas Zufferey and Marino Widmer
2. Meteheuristics for optimal control of discrete systems Lena Djordjevic, Slobodan Antic and Minja Marinovic
3. Simple heuristics for on-line scheduling of operation theatres Nor Aliza Abd Rahmin and Chris N. Potts
14:00–15:30
Transport II (A26)
Chair: Urszula Neuman
1. Vehicle routing with dependent vehicles Edward Kent and Jason Atkin
2. Capacity planning for motorail transportation Pascal Lutter
3. Integration aspects of the gate allocation problem Urszula Neuman, Jason Atkin and Edmund Burke
15:30–16:00
Coffee Break
16:00–17:30
Healthcare (A08)
Chair: Izabela Komenda
1. Allocating Welsh emergency medical services to maximise survival Leanne Smith, Paul Harper, Vincent Knight, Israel Vieira and
Janet Williams
2. Optimising the use of resources within the district nursing service Elizabeth Rowse, Paul Harper, Janet Williams and Mark Smithies
3. Queueing theory accurately models critical care units Izabela Komenda, Jeff Griffiths and Vincent Knight
16:00–17:30
Optimisation IV (A24)
Chair: Emily Cookson
1. Best next sample Sergio Morales Enciso
2. Optimization modeling of distributed energy systems for a smart grid Pedro Crespo Del Granado, Stein W. Wallace and Zhan Pang
3. Air cargo revenue management Emily Cookson, Kevin Glazebrook and Joern Meissner
16:00–17:30
Scheduling / Timetabling I (A25)
Chair: Stefan Ravizza
1. Scheduling games and referees in football: the last song of a Red Hot Chilean Rocker Mario Guajardo, Fernando Alarcón and
Guillermo Durán
2. Over-constrained airport baggage sorting station assignment problem Amadeo Ascó and Jason Atkin
3. Probabilistic airline reserve crew scheduling model Christopher Bayliss, Jason Atkin and Geert De Maere
16:00–17:30
Graphs / Networks (A26)
Chair: Michael Clark
1. The design of transportation networks: a multi objective model combining equity, efficiency and efficacy Maria Barbati
2. Organization of a public service through the solution of a districting problem Carmela Piccolo and Sabrina Graziano
3. Modelling the uncertainty of data and the robust shortest path Michael Clark and Andrew J. Parkes
17:30–19:30
Free Time
19:30–21:30
Dinner @ Peachy Keens
20
SCOR 2012
Sunday, April 22, 2012
10:00–11:00
Multicriteria Decision Analysis II (A08)
Chair: Martin Takac
1. A non-identical parallel machines scheduling problem with multi-objective minimization Jean Respen, Nicolas Zufferey and
Edoardo Amaldi
2. Decision analysis for cost effective maintenance of trunk roads Ena Orugbo and Alkali Babakalli
10:00–11:00
Optimisation V (A24)
Chair: Alessia Violin
1. A choice function based hyper-heuristic for multi-objective optimization Mashael Maashi, Graham Kendall and Ender Özcan
2. A hybrid encoding scheme for grouping problems Anas Abdalla Osman Elhag and Ender Özcan
10:00–11:00
Scheduling / Timetabling II (A25)
Chair: Urszula Neuman
1. An exact algorithm for the uncertain version of parallel and identical machines scheduling problem with interval processing
times and total completion time criterion Marcin Siepak
2. Optimizing real-world workforce scheduling problems Nico Kyngäs
10:00–11:00
Decision Support (A26)
Chair: Stefan Ravizza
1. Deriving priorities from fuzzy group comparison judgements in the fuzzy analytical network process (FANP) Tarifa Almulhim,
Tarifa Almulhim and Ludmil Mikhailov
2. Modelling retail sales for retail price optimization Timo P. Kunz
11:00–11:30
Coffee Break
11:30–12:30
Vincent Knight (Cardiff University) and Louise Orpin (The OR Society) (A25)
12:30–12:30
Lunch
13:30–15:00
Simulation / System Dynamics (A08)
Chair: Magdalena Gajdosz
1. The transition to an energy sufficient economy: a system dynamics model for energy policy evaluation in Nigeria Timothy Mbasuen and Richard C. Darton
2. On the Peter principle: an agent based investigation into the consequential effects of social networks and behavioural factors
Angelico Fetta, Paul Harper, Vincent Knight, Israel Vieira and Janet Williams
3. Empirical bayes methods for discrete event simulation Shona Blair, John Quigley and Tim Bedford
13:30–15:00
Optimisation VI (A24)
Chair: Pablo Gonzalez Brevis
1. The k-separator problem Mohamed Sidi Mohamed Ahmed, Walid Ben-Ameur and Jose Neto
2. A multi-dimensional multi-commodity covering problem with application in logistics Alexander Richter, Jannik Matuschke and
Felix König
13:30–15:00
Scheduling / Timetabling III (A25)
Chair: Urszula Neuman
1. Flexible mobile workforce scheduling and routing Jose Arturo Castillo Salazar and Dario Landa-Silva
2. A case study of investigating a highly constrained search space Lisa Taylor, Jonathan Thompson and Rhyd Lewis
13:30–15:00
Neural Networks / Machine Learning (A26)
Chair: Martin Takac
1. A meta-parameter analysis of boosting for time series forecasting Devon Barrow and Sven Crone
2. Long-term reserve warranty forecasting with neural network models Shuang Xia and Shaomin Wu
3. Inventory optimization under process flexibility assumption using approximate dynamic programming approaches Mustafa
Cimen, Kevin Glazebrook and Christopher Kirkbride
15:00–15:30
SCOR 2012
Awards and Closing Speech: Stefan Ravizza (A25)
21
ABSTRACTS
Abstracts
Friday
Optimisation I
Room: LT1 (14:45 - 16:15)
Chair: Alessia Violin
1. Comparisons between observable and unobservable M/M/1 queues with respect to optimal
customer behaviour
Rob Shone1,∗ , Vincent Knight1 and Janet Williams1
1
Cardiff University; ∗ [email protected]
It is a well-observed property of queueing systems in which individual behaviour is a factor that customers acting in their own interests
do not optimise the collective welfare of society as a whole. In other
words, selfish users do not exhibit socially optimal behaviour. In this
talk we consider a simple M/M/1 queueing system in which the queue
length may or may not be observable by a customer upon entering
the system. The “observable” and “unobservable” cases are compared
with respect to system properties and performance measures under
two different types of optimal customer behaviour, which we refer to
as “selfishly optimal” and “socially optimal”. It can be shown that,
under both types of optimal customer behaviour, the equality of average queue-joining rates for the observable and unobservable models can be induced by a particular relationship between the system
parameters. However, this equality of queue-joining rates results in
differences with respect to other performance measures, such as mean
waiting times and busy periods. It can also be shown that the two types
of model (observable and unobservable) cannot simultaneously share
the same selfishly and socially optimal queue-joining rates, regardless
of the choice of input parameters.
2. Optimization of electricity trading using linear
programming
Minja Marinović1,∗ , Lena Ðord̄ević1 , Dragana
Makajić-Nikolić 1 and Milan Stanojević1
1
∗
University of Belgrade, Faculty of Organizational Sciences;
[email protected]
In last two decades, the liberalization of the electricity markets have
been established in order to increase efficiency, harmonize and reduce
electricity prices, make a better use of resources, give customers the
right to choose their supplier and provide customers with a better service. This change made the electricity market competitive and introduced several new subjects. In this paper, we observe one of the
subjects: Electricity Trading Company (ETC) and its daily trading
process. We present linear mathematical model of total daily profit
maximization subject to flow constraints. It is assumed that the demand and supply are known and some of them are arranged. Possible
transmission capacities are known but also additional capacities can
be purchased. All trading, transmission prices and amounts are subject of auctions. First, we present energy trading problem as directed
multiple-source and multiple-sink network and then model it using
linear programming. Also, we provide one realistic example which is
slightly changed in order to save confidentiality of the given data.
3. Performance measurement and trade-offs in
UK higher education institutions
Carolyn Booker1,∗ , John Quigley1 and Lesley Walls1
1
University of Strathclyde; ∗ [email protected]
The purpose of this research is to examine trade-offs and conflicts
between performance measures in the UK Higher Education sector.
The performance measures under consideration are those which are
imposed on universities from outside, such as statutory performance
indicators and newspaper league tables, and which bring rewards in
SCOR 2012
the form of either status or funding. The existing literature provides
evidence that such measures are causing tension within institutions,
but there has to date been no attempt to examine that tension using the
tools of management science.
The main tool used here is Data Envelopment Analysis. A new
DEA model has been developed which extends Podinovski’s Tradeoff model to incorporate a weighted preference structure. This model
is used first to determine the production possibility set for a group of
universities and then to explore the options open to them. In many
kinds of performance measurement system the reward achieved by a
university, such as a “top ten” position or a share of a fixed amount
of funding, depends not only on that institution’s own decisions but
also on the strategic decisions of others. Game Theory provides a
range of structures which model such interactive decisions and can
aid a decision-maker in determining optimal strategies. The results of
the DEA model are therefore processed using a typical league table
construction and then evaluated through the lens of Game Theory.
Transport I
Room: LT2 (14:45 - 16:15)
Chair: Penny Holborn
1. Optimal toll enforcements on motorways
Elmar Swarat1,∗ , Guillaume Sagnol1 and Ralf
Borndörfer1
1
Zuse Institute Berlin; ∗ [email protected]
We will present the problem of computing optimal tours of toll inspectors on German motorways. The control is the responsibility of
the German Federal Office of Goods Transport (BAG). They are interested in improving the control planning by the use of optimization techniques. We build up an integrated model, consisting of a
tour planning and a duty rostering part. The tours should guarantee a
network-wide control whose intensity is proportional to given spatial
and time dependent traffic distributions. We model this using a spacetime network and formulate the associated optimization problem as a
Multi-Commodity Flow Problem in an integer programming (IP) formulation. The rostering part is needed, since we must assign a crew
to each tour. All duties of a crew member must fit in a feasible roster.
It is also modeled as a Multi-Commodity Flow Problem in a directed
acyclic graph, where specific paths correspond to feasible rosters for
one month. To the best of our knowledge there is no approach in
the literature that integrates vehicle routing and duty rostering (planning) into one model yet, and there is no model of Toll Enforcement
Optimization. With our approach all legal rules can be modeled. We
present computational results which document the practicability of our
proposal.
2. The effects of release levels and trajectory
lengths on the aircraft arrival sequence
Stanislava Armstrong1,∗ , Jason Atkin1 , Geert De
Maere1 and Edmund Burke2
University of Nottingham; ∗ [email protected] University of
Stirling;
1
In the literature, the arrival sequencing problem is commonly addressed without consideration for the control problem within the
stacks, or the control problem between the stacks and the runway.
When used at airports like Heathrow, where during busy times aircraft are arranged in stacks and are released predominantly from the
bottom level of each stack, such algorithms could lead to solutions
that achieve very good objective function values, but are not always
feasible in practice. The main goal of this work is to find feasible arrival sequences by considering the sequencing and control problems
23
Abstracts
Friday
simultaneously. An exact multi-objective dynamical programming algorithm was developed for this purpose. Its effect on the arrival sequence and some potential benefits that can be achieved at Heathrow
through slight tactical changes have been investigated. This presentation focuses in particular on the impact which varying the number of
release levels has on the arrival sequence, both in terms of runway utilisation and of waiting time per aircraft. Furthermore, we will consider
the effects of extending the lengths of the trajectories from the stack
to the runway for some of the aircraft. This allows the aircraft that
remain in the stacks to ladder down faster into the release levels and
as a consequence, may result in better sequences becoming feasible.
3. Optimisation of traffic signal settings at arterials during variable conditions
Cesar Velandia-Brinez1,∗ , Ruibin Bai2 , Graham
Kendall3 and Jason Atkin4
1
University of Nottingham Ningbo China;
[email protected] The University of
Nottingham Ningbo China;3 The University of Nottingham Malaysia
Campus;4 The University of Nottingham;
∗
Effective traffic signal timings are capable of providing unstopped vehicular movements at arterial intersections. However, during peak demand hours, such strategies are unable to provide continuous green
indications. They might worsen prevalent traffic conditions if not revised to avoid queue overflows.
Signal settings (cycle time, green times, phase sequences and offsets)
can be calculated with different objectives in mind. Offsets are used
essentially to synchronise signals. It is often considered that progression schemes (maximisation of throughput by co-ordination) are only
achievable during light traffic conditions, and the use of other approaches based on minimisation of disutility functions, e.g. delay,
number of stops, queue lengths, represent better congested road networks. A critical challenge is the staging of static queues and platoons
(incoming groups of vehicles) to avoid progression disruptions and
overflowing queues. Therefore, a suitable control method would be
able to cope with traffic outbursts, accommodating moving vehicles,
and discharging queues regularly.
The proposed methodology examines the interactions between platoons and queues, considering both progression and fair allocation
of green times. For this purpose, a set of experiments using a traffic micro-simulation are proposed to reproduce scenarios fluctuating
between low and high peaks of demand. Measures of effectiveness
will be analysed based on progression bandwidth, platoon composition, and queue interactions, while considering side road conditions.
Our approach studies this phenomena from the perspective of vehicle groups rather than network or individual performance indicators,
which may provide new insights to the problem of finding adequate
traffic signal settings at arterial roads.
Mathematical Programming
Room: LT1 (16:30 - 18:00)
Chair: Pablo Gonzalez Brevis
1. Nurse rostering in a Danish hospital
Jonas Baeklund1,∗
1
Aarhus University; ∗ [email protected]
This presentation will describe a nurse rostering problem from a ward
at a Danish hospital. The problem is highly constrained and comprises a large set of different constraints. A branch-and-price method
for solving the problem exactly is proposed. The master problem is
to assign schedules to the nurses, and its linear relaxation is solved
by means of column generation. The pricing sub-problem is to generate feasible schedules for the nurses and – as a couple of different
24
constraints including several special Danish regulations have to be observed – is solved by constraint programming. A number of specific
algorithms for handling these constraints are proposed.
The method is very flexible regarding the rules a schedule should comply with, which is a key concern when creating solution methods for
nurse rostering problems.
Computational tests show that optimal solutions can be found for instances with a two weeks planning period in a reasonable amount of
computing time.
2. A bilevel model for valves location in water distribution systems
Andrea Peano1,∗
1
ENDIF - Università degli Studi di Ferrara;
[email protected]
Maintenance operations on the pipes of a water distribution system
(WDS) require to drain the leaking pipe before repairing. To drain
a pipe, some of the isolation valves already present in the network
are closed, in such a way that the smallest part of network containing
the damaged pipe is disconnected from each of the reservoirs feeding
the WDS. In this way, only that subnetwork is isolated and drained.
Such subnetworks are called network sectors and their shape is determined by valves positioning. Isolation valves are usually located at
the extremes of the pipes. Since valves are expensive, WDSs can be
equipped with just a limited number of valves, whose location poses
a challenging optimization problem. When any pipe is equally like
to require maintenance, the best valve location is the one which minimizes the maximum undelivered demand in case a sector needs to be
isolated. Undelivered demand accounts not only for the demand of the
sector which requires maintenance, which is drained on purpose, but
also of the demand of any other sector which gets disconnected from
the reservoirs as a secondary effect of isolating the first sector. This
issue makes this real-life problem much more complex than a pure
graph partitioning problem. We propose a bilevel mixed-integer linear programming model, where the outer level models valves location
and the inner level models water flow. We show how to reformulate
the bilevel model as a single level mixed-integer linear model, and test
our approach on a real life WDS.
∗
3. The primal-dual column generation method:
theory and new applications
Pablo González-Brevis1,∗ , Jacek Gondzio1 and
Pedro Munari2
1
University of Edinburgh;
[email protected] University of Sao
Paulo;
In this talk we introduce the primal-dual column generation method.
This method relies on well-centred and suboptimal dual solutions to
generate new columns in a column generation framework. Theoretical
support is provided to show that the method converges to an optimum
if such optimum exists. Additionally, computational comparisons with
the standard column generation and the analytic centre cutting plane
method are presented. The problems on which we have performed the
comparisons are the cutting stock problem, the vehicle routing problem with time windows and the capacitated lot sizing problem with
setup times. The method shows consistently reductions in the number of outer iterations as well as CPU time when solving the relaxations of these combinatorial optimization problems after applying a
Dantzig-Wolfe reformulation. The results are encouraging showing
that the method is an attractive general purpose column generation
method. Its dynamic adjustable optimality tolerance and its natural
stabilization features make it a competitive option in this context. The
performance of the method is enhanced when larger instances are considered.
∗
SCOR 2012
Abstracts
Friday
Multicriteria Decision Analysis I
Room: LT2 (16:30 - 18:00)
the ideal point by a super-ideal reference point. Compromise solutions
thus neatly fit with the concept of Pareto optimality.
Chair: Izabela Komenda
3. Lower bound improvements of penalty param1. A nice use for MCDA in public health: poten- eters for discrete - continuous linear bilevel probtial new approaches to decision making in the na- lems
tional institute for health and clinical excellence
Renato Mari1,∗
1,∗
Brian Reddy
1
∗
Università di Roma "Tor Vergata"; [email protected]
1
ScHARR, University of Sheffield;
∗
[email protected]
The National Institute for Health and Clinical Excellence (NICE) is
an agency of the NHS, providing guidance to the health services and
other relevant decision making bodies in England regarding new drugs
and technologies, broader clinical practice and public health. It combines evidence-based and health economic approaches to medicine
with equity concerns in order to adequately prioritise interventions,
attempting to reduce variations in levels of treatment between local
regions and encourage best practice across the health system.
However, for a number of reasons it is challenging to adequately quantify, model and describe the multiple effects of public health interventions. As a result there is uncertainty around their cost effectiveness;
this adds to the complexity of the decision making process.
Multi-criteria decision analysis (MCDA) relates to the methods and
procedures by which concerns about multiple conflicting criteria can
be formally incorporated into the management process and is used to
provide a framework to help decision makers structure complex decisions. Typically these problems would also include multiple ways
of judging these criteria, multiple objectives, multiple stakeholders
and/or high levels of uncertainty. Such situations are common in public health scenarios.
This presentation will describe some of the most common difficulties of measuring cost effectiveness in public health and explain why
MCDA techniques may be an appropriate method to overcome some
of these issues. It will then explain areas within the Centre for Public
Health Excellence in NICE where MCDA approaches are currently
being investigated and may be incorporated in the future.
Linear bilevel programming problems are the most studied class of
bilevel problems and have been widely used to represent real life applications in which there are hierarchical decisional structures. The
case in which a subset or all the variables are discrete represents another important class of problems on which we focused as they enable
to better describe applications and problems with an inner combinatorial nature. We consider a particular class of linear bilevel problems
in which the variables controlled by the leader are required to be discrete. It is a well known result that such a problem is equivalent to
a continuous linear bilevel problem in which the integrality requirements are relaxed and the leader’s objective function is modified including a concave penalty function weighted by a penalty parameter.
This equivalence is true for a sufficiently large value of this parameter. A valid lower bound for it is already known. We provide two
improvements of this existing lower bound and experiment the two
new lower bounds on a set of 120 test problems. The parameters proposed are better from a theoretical point of view because they assure a
reduction of the penalty used to force the integrality requirements on
the upper level variables. This improvement is also clearly shown by
the computational results: the CPU time spent to solve the problem is
on average less than 20% and 30% respectively for the two proposed
lower bounds. This implies the possibility to solve larger instances
increasing the applicability of bilevel models.
2. Compromise solutions
Kai-Simon Goetzmann1,∗ , Christina Büsing1 , Jannik
Matuschke1 and Sebastian Stiller1
1
TU Berlin; ∗ [email protected]
Applications of combinatorial optimization often feature several contradicting objectives, e.g., cost and duration of a transport. The basic
concept in multicriteria optimization is Pareto optimality: a solution
is Pareto optimal, if improving one objective is impossible without
worsening another. However, in general the number of Pareto optimal solutions is exponential. To choose a single, well-balanced Pareto
optimal solution, Yu (1973) proposed compromise solutions.
A compromise solution is a feasible solution closest to the ideal point.
The ideal point is the component-wise optimum over all feasible solutions in objective space.
Compromise solutions are always Pareto optimal. Using different
weighted norms, the compromise solution can attain any point in the
Pareto set. The latter does not hold for other concepts, e.g., optimization over different weighted sums of objectives. In particular, compromise solutions find well-balanced Pareto solutions that may escape
weighted sum methods.
Compromise solutions (and the slightly more general reference point
methods) are widely used in state-of-the-art software tools. Still, there
are very few theoretical results backing up these methods.
We establish a strong connection between approximating the Pareto
set and approximating compromise solutions. In particular, we show
that an approximate Pareto set always contains an approximate compromise solution. The converse is also true if we allow to substitute
SCOR 2012
25
Abstracts
Saturday
Stochastic Modelling I
Room: A08 (09:00 - 10:30)
Chair: Emily Cookson
1. Stabilizing policies for employment portals
Hanyi Chen1,∗ and Burak Buke1
1
University of Edinburgh; ∗ [email protected]
The portals, which match the people who provide a specific service
with the people who demand the service, are becoming increasingly
popular recently. In this talk, we concentrate on employment portals,
where the employers and employees are assumed to arrive with respect
to independent poisson processes with rate λ1 and λ2 respectively.
Each given employee can match with a specific employer with probability q independently. This system resembles assembly-like queueing
systems. However, with q < 1 it differs significantly from assemblylike systems and has not been studied in the literature. We model it
as a two-dimensional Continuous Time Markov Chain and analyze its
properties. Our main focus is on stabilizing policies for such portals.
To consider the stability of the system, we will first consider the case
when the matching probability q = 1, as it can be simplified to an onedimensional birth-death process, which is well-known that the system
is null-recurrent when the λ1 = λ2 and transient when λ1 6= λ2 . For
the case when 0 < q < 1, we will prove that the same conclusion as
the case of q = 1 holds.
Next, we will introduce the Accept the Shortest Queue policy, which
forces constraint on the acceptance of customers. The system will accept employees(employers) only when they have a smaller or equal
number than employers(employees). Under this policy, the system is
ergodic for any set of arrival rates, but the rate of rejecting customers is
exceptionally large. Therefore we consider how to modify it to enable
us to accpet more customers without losing stability.
2. Assortment planning with substitution effects
Jochen Schurr1,∗
1
University of Edinburgh; ∗ [email protected]
The generation unit commitment problem is to find the scheduling of
a set of electric power generation units and their power output over a
short term planning period – typically 24 to 72 hours ahead. In traditional formulations the objective is to minimise the cost for real power
output subject to technical constraints such as load balance, spinning
reserve and minimum up-and downtimes of individual units. Deterministic variants of this problem are well known and researched and
are be- ing solved successfully in the energy industry, mainly due to
good predictability of future electricity demand. However, due to increased demand uncertainty and fluctuating wind supplies it becomes
more and more important to incorporate uncertainty explicitly in the
problem formulation. Stochastic unit commitment problems are large
scale multistage mixed integer optimisation programmes which are
excruciatingly hard to solve. In this talk I will give an outline of the
problem formulation and describe the scenario decomposition techniques we use to make the problem more tractable. Our solver is
implemented in C++ and some numerical results are available to underline the importance of decomposition when solving this class of
problems.
1
Lancaster University; ∗ [email protected]
In this talk, we propose an assortment planning heuristic with active
learning that accounts for substitution effects.
Some subset selection problems on a retail level, arising for example
in internet advertising or in the apparel industry, face besides the obvious resource limitation further challenges that are due to high uncertainty in demand. Unlike conventional retailers, fast fashion retailers
like Zara or H&M have invested in highly efficient supply chains that
enable them to react promptly on additional information or changes in
customer preferences. This ability raises the question how one should
optimally allocate show room space to products in order to maximize
profit over the entire selling season.
We present a dynamic programming (DP) approach with Bayesian
learning to study assortment decisions in a limited horizon setting.
A particular challenge when accounting for substitution effects is the
fact that the DP is no longer weakly coupled and hence cannot be
decomposed into single-product sub-problems. Further, the Bayesian
learning scheme becomes incomparably more complex as compared
to existing models that assume independence among products. We
will briefly discuss the simplification approach via a one-step look
ahead of the DP and a Monte-Carlo integration scheme for the expected cumulated future profits. Further, computational results will be
provided that demonstrate the large-scale feasibility and performance
of the approach.
3. Decomposition techniques for stochastic unit
commitment problems
Optimisation II
Room: A24 (09:00 - 10:30)
Chair: Pablo Gonzalez Brevis
1. The traveling visitor problem and algorithms
for solving it
Milan Djordjevic1,∗ , Marko Grgurovic1 and Andrej
Brodnik1
UP FAMNIT; ∗ [email protected]
We consider new problem named the Traveling Visitor Problem
(TVP). Visitors start from a hotel with desire to visit all interesting
sites in a city exactly once and to come back to the hotel. Since, the
visitors use streets and pedestrian zones, the goal is to minimize the
visitor’s traveling. A new problem is similar to the Traveling Salesman Problem (TSP) with a difference that the traveling visitors, during its visit of sites, can’t fly over buildings in the city, instead visitors
have to go around these obstacles. That means that all Euclidean distances, like those in Euclidean TSP, are impossible in this case. The
tested benchmarks are combined from three real instances made using tourist maps of cities of Venice, Belgrade and Koper and two instances of modified cases from TSPLIB. We introduced and compared
two exact methods for solving the TVP. In all tested cases the Koper
Algorithm significantly outperforms the Naïve Algorithm for solving
the TVP - the difference in quality of solutions differs from 6.52% to
354.46%.
1
2. A dynamic programming heuristic for the
quadratic knapsack problem
Franklin Djeumou Fomeni1,∗ and Adam Letchford1
1
Lancaster University;
[email protected]
It is well known that the standard (linear) knapsack problem can be
solved exactly by dynamic programming in O(nc) time, where n is the
number of items and c is the capacity of the knapsack. The quadratic
knapsack problem, on the other hand, is NP-hard in the strong sense,
which makes it unlikely that it can be solved in pseudo-polynomial
time. We show however that the dynamic programming approach
to the linear knapsack problem can be modified to yield an effective
heuristic for the quadratic version.
∗
Tim Schulze1,∗ , Andreas Grothey1 and Kenneth I.M. 3. Compact formulations of the Steiner traveling
McKinnon1
salesman problem
26
SCOR 2012
Abstracts
Saturday
Saeideh D. Nasiri1,∗ , Adam Letchford2 and Dirk O.
Theis3
1
STOR-i DTC, Lancaster University;
[email protected] Management School,
Lancaster University;3 University of Magdeburg;
∗
The Steiner Traveling Salesman Problem (STSP) is a variant of the
Traveling Salesman Problem (TSP) that is particularly suitable when
dealing with sparse networks, such as road networks. The standard integer programming formulation of the STSP has an exponential number of constraints, just like the standard formulation of the TSP. On
the other hand, there exist several known compact formulations of
the TSP, i.e., formulations with a polynomial number of both variables and constraints. In this study, we present three of these compact
formulations specifically the commodity flow formulations and timestage formulations for the TSP and show how they can be adapted to
the STSP. We also compare these formulations in terms of the strength
of their resulting linear programming relaxation bound.
Heuristics / Metaheuristics I
Room: A25 (09:00 - 10:30)
Chair: Penny Holborn
1. Heuristics for a split vehicle routing problem
with backhaul
Michela Lai1,∗ , Massimo Di Francesco2 and Paola
Zuddas3
1
Department of Mathematics and Computer Science, University of
Cagliari; ∗ [email protected] Department of Land Engineering,
University of Cagliari;3 Department of Mathematics and Computer
Science, University of Cagliari;
This research addresses a VRP motivated by a real case study. A carrier must serve two type of customers: the importers receive loaded
containers from a depot and return empty containers, whereas the exporters receive empty containers and ship loaded containers to the depot. Unlike classical drayage problems, what is original in this vehicle
routing problem is the impossibility to separate trucks and containers
during customer service and the opportunity to carry more than one
container in some trucks. Moreover, according to the carrier’s policy,
importers must be served before exporters, customers may demand
more than one container and may be visited more than once. In order
to address this special split backhaul problem, we propose a mathematical model based on a node-arc formulation. Nodes represent the
depot, importers and exporters. Arcs represent links between nodes,
but, due to precedence constraints, arcs from exporters to importers
are not considered. Since this problem is NP hard, this research investigates approximate algorithms and, whenever it is possible, compares
approximate solutions to exact ones. In this work we propose a two
phase approach. In the first phase, we separate importers and exporters
and determine separate routes using an efficient metaheuristic. In the
second phase, we propose several heuristics to merge these routes. Finally, we compare our solutions to the real decisions of the carrier who
has motivated this problem.
2. Periodical vehicle routing problem due to
driver familiarity
Matthew Soulby1,∗ and Jason Atkin1
1
University of Nottingham; ∗ [email protected]
Driver familiarity is often of great importance to delivery companies.
Not only will familiarity decrease service and travel times, due to
knowing the roads, building entrances, delivery points and so on, but
the rapport which can be built up between driver and customer also
SCOR 2012
has value. When driver learning is applied to the Vehicle Routing
Problem (VRP) a trade off occurs between route length and the driver
familiarity within the route. To accommodate this, formulations have
been applied in the past which group customers that consistently occur
in the same route into delivery areas. Determining whether the delivery areas should remain the same over time or be adapted to account
for the periodical nature of the orders needs to be considered. For
the normal VRP where driver familiarity is ignored the main focus
is usually upon the route length, so taking account of the periodical
nature and more predictable elements of the stochastic customer demand is of little consequence. Information obtained from the vehicle
routing software and consultancy company, Optrak, has allowed for
a more accurate, rich representation of the VRP encountered by delivery companies, considering the gain in driver familiarity and the
periodical nature of the customer demand.
3. Dynamic vehicle routing problems with pickups, deliveries and time windows
Penny Holborn1,∗ , Jonathan Thompson1 and Rhyd
Lewis1
Cardiff University; ∗ [email protected]
To solve the dynamic pickup and delivery problem with time windows (DPDPTW) we are investigating methods embedded in a rolling
horizon framework, thus allowing us to view the problem as a series of static ones. Initial research concentrated on the static variant
of the problem. We produced several variations of a general heuristic for constructing an initial feasible solution, and used neighbourhood search operators to make improvements. This was extended into
an implementation of a Tabu Search heuristic, varying reconstruction
heuristics and a Branch and Bound method to improve the results.
Analyses have been performed across a range of benchmark datasets,
including both clustered and random allocations of pickup and delivery locations. The approaches used give results which are competitive
with the state of the art. Our current research is dedicated to solving
the dynamic problem, where a time stamp is allocated to each request
and the request does not become known to the system until that time.
Investigations have been performed to identify when the algorithms
should be updated to incorporate the arrival of new requests. Datasets
with both varying degrees of urgency and proportion of dynamic requests have been examined along with various different waiting strategies and the design of new evaluation functions. Competitive results
have been achieved across a range of benchmark datasets.
1
Supply Chain Management I
Room: A26 (09:00 - 10:30)
Chair: Magdalena Gajdosz
1. The impact of on time delivery on supply chain
management: are third party logistics providers
worth the investment?
Yolanda Silvera1,∗ and Rebecca DeCoster1
Brunel University, London; ∗ [email protected]
The purpose of this research is to develop an analytical model as well
as a framework for the measurement of performance within supply
chain partnerships. Most existing research on supply chain performance incorporates the use of quantitative analysis, but this has been
found to not be entirely appropriate when it comes to the area of supplier relationships and supplier performance management. The major
focus of this research is on the area of on time delivery, and its effects
on the supplier relationship.
There has been the misconception that Supply Chain Management’s
main focus is on software and systems. It has long been thought that
all that is required for an effective supply chain is an investment in
technology; it was not felt that there was a requirement to do anything
1
27
Abstracts
Saturday
as the technology installed will do everything to improve efficiencies.
But with new technologies and globalisation of the markets, this way
of thinking has been proving to be wrong. The leading scholars in
the area of supply chain management have also been emphasising this
new reality, when HBR convened a panel of leading thinkers in the
field of supply chain management, technology was not the major issue
on their minds it was people and relationship management. This paper, through data collected emphasises the relevance and importance
of performance measurement in supply chains, especially those indicators affecting on time delivery. Data collected thus far indicates the
advantages and disadvantages of third party logistic providers to the
on time delivery process.
and non priority goods. The objective is to maximise the throughput
of high priority goods under capacity restriction. Simulation will be
applied to assess the performance of the model.
2. Forecasting ARMA demand processes: the impact of temporal aggregation
Julie Williams1,∗ , Harper Paul1 , Gillard Jonathan1
and Knight Vincent1
Bahman Rostami Tabar1,∗ , Mohammad Zied
Babai2 , Aris Syntetos3 and Yves Ducq4
1
University of Bordeaux1, Bordeaux Management School(BEM);
∗
[email protected] Bordeaux Management
School;3 University of Salford;4 University of Bordeaux1;
There are many strategies that may be used to reduce the demand
variability and thus to improve forecasting performance. An intuitively appealing such strategy is to aggregate demand in lowerfrequency “time buckets", In this paper, we investigate the impact
of non-overlapping temporal aggregation on forecasting performance.
We assume that the non-aggregated demand follows an ARMA-type
process and a Single Exponential Smoothing (SES) procedure is used
to estimate the level of demand. The theoretical forecast errors (both
their mean and variance) are derived for the aggregated and nonaggregated demand in order to contrast the relevant forecasting performances. The theoretical analysis is followed by experimentation with
real world data. The results indicate that performance improvements
achieved through the aggregation strategy are a function of the aggregation level, the smoothing constant value and the process parameter;
they also show that for high positive autocorrelation, aggregation does
not add any value.
3. The impacts of material convergence on the
post-disaster humanitarian logistic operation
Nha-Nghi Huynh1,∗ , Sandra Transchel1 , Maria
Besiou1 and Luk Van Wassenhove2
1
∗
Kuehne Logistics University;
[email protected] INSEAD;
Logistics is central to disaster relief and includes activities such as
transport, tracking and tracing, warehousing and last-mile-delivery.
However, there are crucial differences between the humanitarian and
the commercial logistics. One of the differences is that the conditions in which humanitarian relief organizations operate are extremely
chaotic and instead of minimizing inventory or transportation costs
the objective is rather to minimize lead times for aid items in order
to reduce suffer of disaster victims. One of the challenges humanitarian operations are faced with are the so-called unsolicited donations.
In-kind donations often do not meet the relief needs of an affected
population, but are nevertheless pushed into local warehouses of the
disaster areas. During the Haiti earthquake for example, the number of
flights to the Port-au-Prince airport increased from 13 up to 100 during the first three days of response. Thus, the airport was faced with
serious bottlenecks and unsolicited donations occupied very limited
warehouse capacity in the airport. Over the past years the number of
research contributions to humanitarian logistics has risen significantly.
However, methods of operations research have not yet widely been
applied in this field. In this work a queuing model will be presented
which illustrates the effects of unsolicited donations on the Haiti relief operation. Aid items are classified into three classes: high, low
28
Stochastic Modelling II
Room: A08 (10:45 - 11:45)
Chair: Emily Cookson
1. Time-dependent stochastic modelling for ambulance demand prediction and scheduling
1
Cardiff University; ∗ [email protected]
For patients requesting Emergency Medical Service (EMS) assistance
for a life-threatening emergency, the probability of survival is strictly
related to the quickness of assistance. As both demand for, and public
expectation of, EMS is escalating in the Western world, the provision
of an efficient and effective service is a significant challenge for many
nations. A particular difficulty for planners is to allocate often limited resources whilst managing increasing demand for services, in a
way to ensure high levels of geographical coverage and to improve
key performance targets.
We are working with the Welsh Ambulance Service Trust (WAST)
which is struggling to meet all of its response time targets. Our work
considers a novel time series approach to forecasting the daily demand exerted upon WAST using the model-free technique of Singular
Spectrum Analysis; and shows that in addition to being more flexible in approach, the predictions generated using this technique compare favourably to forecasts obtained from conventional methods. We
progress to use this technique to predict demand at a regional level,
and use a range of approximation and numerical time-dependent priority queueing theory techniques to obtain staffing level recommendations. Our research contributes to the ongoing study of time-dependent
multi-server queues through developing the techniques to cope with a
priority system where life-threatening emergencies are treated with
precedence. Ultimately we aim to develop a time-dependent and priority workforce capacity planning tool to optimise ambulance deployment strategies and generate rosters for crew members.
2. Point process models for assessing the reliability of desalination plant equipments subject to
failures due to red tide events
Ahmed Al Hinai1,∗ and B. M. Alkali1
1
∗
Glasgow Caledonian University;
[email protected]
This paper investigates the number of failures related to red tide events
and other environmental factors on a Reverse Osmosis (RO) Desalination Plant in Oman. There are clear indications of worsen seawater
quality during periods of red tides in the Sultanate of Oman. In this
study a failure mode and effect analysis (FMEA) is conducted on the
critical equipment in the Plant to identify the classes of failure modes
and their effects on the RO Plant smooth operation. The failure and
maintenance history data for the period 2006-2010 is evaluated and
more emphasis is focused on plant failures during the periods of the
Red tide incidents. A stochastic point process model is considered for
reliability analysis in this study. The history of the observed failure
process is assumed to follow a non-homogenous Poisson process, as
the inter-arrival times between the Plant’s failures vary with time. A
standard statistical approach is used for reliability analysis by fitting
the Weibull distribution to the data sets. The results obtained are presented on a distribution overview.
SCOR 2012
Abstracts
Saturday
The results presented here is to set a pace of further reliability modelling to schedule cost effective preventive maintenance actions on the
RO Desalination Plant equipment.
Reliability / Risk Assessment
Room: A24 (10:45 - 11:45)
Chair: Magdalena Gajdosz
used for post-mortem accident analysis and safety risk assessment.
This is accomplished by applying it to the analysis of accident reports
from hazardous industries and by performing a safety risk study for an
external organisation. A mixed-modelling approach, combining system dynamics and methods used in the risk and reliability field (e.g.
fault tree analysis that is used to model equipment reliability) has been
adopted. The output will be a set of guidelines that can support risk
analysts in their future risk studies.
1. Reliability assessment of subsea systems in
ultra-deepwater oil and gas developments
Heuristics / Metaheuristics II
Anietie Umofia1,∗ , Stephen Okonji1 and Shaomin
Wu1
1
Cranfield University; ∗ [email protected]
The exploration and production of oil and gas is globally witnessing a
very dramatic and rapid expansion into deep waters. Studies indicate
that deep water fields account for more than 25% of operator investments in offshore facilities and will rise to over 40% by the end of
the decade. This drive introduces a significant increase in the cost of
hydrocarbon search and also presents unprecedented challenges. The
challenges include safety, environment, flow assurance and equipment
reliability.
Deepwater conditions inherently dictate the development of these
fields by means of subsea production systems since traditional surface
facilities such as steel-piled jacket might be either technically unfeasible or uneconomical due to the water depth. Maintaining systems
located in the ultra deep-water requires specialised and expensive vessels, which need to be equipped with robotic devices due to the water depths. Any requirement to intervene or repair an installed subsea
system is thus normally very expensive and may result in considerable
economic production loss. High equipment reliability is therefore required in order to safeguard the environment and personnel, and to
make the exploitation of hydrocarbons with subsea technology economically feasible.
This paper investigates reliability assessment for subsea systems operated in the ultra-deepwater. With API 17N as a key focus, the paper
looks at the control system for this type of developments and the reliability assessment.
2. Modelling the impact of human and organisational factors on the safety risk over time
Magdalena Gajdosz1,∗ , Susan Howick1 and Tim
Bedford1
1
∗
Room: A25 (10:45 - 11:45)
1. An improved choice function heuristic selection for cross domain heuristic search
John Drake1,∗ and Ender Ozcan1
1
SCOR 2012
University of Nottingham; ∗ [email protected]
Hyper-heuristics are a class of high-level search technologies which
aim to solve computationally difficult problems. Unlike traditional
meta-heuristic techniques, a hyper-heuristic operates on a search space
of heuristics rather than directly on the search space of solutions. A
single-point based search selection hyper-heuristic framework relies
on two key components, a heuristic selection method and a move acceptance method. Operating on a single candidate solution until some
termination criteria is met, low-level heuristics are repeatedly selected
and applied producing a new solution, then a decision is made as to
whether to accept this new solution or not. The Choice Function is
an elegant heuristic selection method which scores heuristics based
on a combination of three different measures and applies the heuristic
with the highest rank. Each of these measures are weighted appropriately to provide sufficient intensification and diversification of the
heuristic search process. Different methods in the literature have been
proposed to manage these weightings, here we will describe a new
method loosely based on reinforcement learning. Using the HyFlex
software benchmark framework developed to support CHeSC 2011
(the first Cross-domain Heuristic Search Challenge) we have tested
and compared this new method for controlling such parameters to previous approaches.
2. Hyper-heuristic to construct magic squares
University of Strathclyde;
[email protected]
This research project is concerned with the analysis and management
of safety risks associated with the operation of hazardous systems such
as those used in nuclear power plants, chemical plants or transportation organisations.
Overall, there are two types of analyses that inform safety risk management of hazardous systems: safety risk assessment and postmortem accident analysis. The review of the literature suggests that
the methods currently used in these analyses need to be extended to
account for more recent understandings of the impact of people who
operate, maintain and control these systems. Also, most of the methods are static, sequential and usually ignore feedback structures within
which human decision-making operate. Thus, they are not robust
enough to capture the dynamic impact of human and organisational
factors on the safety risk.
A system dynamics approach, which has been used to analyse feedback systems and their behaviour over time, provides a range of structures that can help to overcome these problems. This research investigates how system dynamics can complement methods traditionally
Chair: Michael Clark
Ahmed Kheiri1,∗ and Ender Ozcan1
1
University of Nottingham; ∗ [email protected]
A magic square is a square matrix that contains distinct numbers in
which the summation of the numbers in each row, column and the two
diagonals has the same constant total known as the magic number.
Constructing the magic square has been considered as a hard computational problem domain. A class of hyper-heuristics aims to provide
solution across a range of problem domains by selecting and/or mixing
a fixed set of low level heuristics during the search process. There has
been a growing interest in the development of such general methodologies which searches the space of heuristics rather than the space of
the solutions. This study presents a methodology to construct magic
squares using a selection hyper-heuristic based on a random descent
heuristic selection method. The results show that the approach could
be an effective methodology to construct magic squares. The results
will be reported in more details at the conference.
29
Abstracts
Saturday
Supply Chain Management II
Sandra Rauscher1,∗ and Kevin Glazebrook1
1
Room: A26 (10:45 - 11:45)
Chair: Alessia Violin
1. Markovian analysis of stochastic serial multi
echelon supply chains
Despoina Ntio1,∗ and Michael Vidalis1
1
University of the Aegean Chios Greece; ∗ [email protected]
In this project the performance of stochastic serial multi echelon inventory systems is being analyzed. In contrast to my previous work
and other similar works, in this case a more complex inventory model
with four stages is developed. Furthermore, the Erlang distribution
allows the model to depict the replenishment process in a better way
compared to other methods that have been utilized in the past. This
extended model combined with the two and three stage inventory systems leads to the construction of a framework for the inventory decision making process. Specifically, three serial inventory systems
with two, three and four stage respectively have been researched. The
demand and the replenishment process are stochastic. Orders follow
(S,s) policy. The replenishment processes follow the Erlang distribution. The external demand is distributed by pure Poisson process
which means that the amount that each customer asks is one unit. Last
but not least the last upstream node is always considered saturated.
The supply networks are modeled as continuous Markov processes
with discrete states. The structures of the transition matrices of those
systems are explored and computational algorithms are developed to
generate them for different values of systems parameters. Mat lab
software is used for any computation. Various performance measures
derive from steady state probabilities such as fill rate, average inventory in systems and cycle time.
2. Modeling and evaluating a tandem supply
chain with multiple stages
Michail Geranios1,∗ and Michail Vidalis1
1
University of Aegean; ∗ [email protected]
In this work a serial supply chain with an arbitrary number of nodes
(retailer, wholesaler, manufacturer, supplier, etc.) is examined. Each
node supplies only one downstream node and is replenished by only
one upstream node. Each node, except for the most upstream, faces
supply uncertainty and the last node (retailer) additionally faces external demand that follows the pure Poisson distribution. Each node,except for the most upstream, which is saturated- follows a continuous
review ordering policy. If the upstream has insufficient stock, then the
orders are partially satisfied and the rest is lost. The system’s performance is determined by performance measures, such as Fill Rate,
Average Inventory. To calculate these measures the supply chain is
modeled as a continuous time Markov process with discrete states.
Our major task is to figure out how the performance measures are influenced when we expand our system, adding one node at a time. A
computational algorithm is developed to generate the transition matrices (different values of system characteristics) and by them the stationary probability distribution. The proposed algorithm is used as an
optimization tool to determine the optimal values of the system’s parameters. After determining the performance measures, we compare
the numerical results in order to reveal the optimal ordering policy.
Stochastic Modelling III
Room: A08 (14:00 - 15:30)
Chair: Izabela Komenda
1. Hybrid lateral transshipments in multi-item
inventory networks
30
Lancaster University Management School;
[email protected]
Saving costs in inventory systems can often only be accomplished
by reducing service levels. Allowing lateral transshipments in multilocation inventory networks permits lower levels of safety stock
thereby cutting costs while maintaining or improving service levels.
Usually, such movements of stock are either carried out in a reactive manner responding to a stock-out in the system, or preventively
to rebalance inventory levels. Recent results show that using a hybrid version of these two approaches can yield further improvements.
This lies in the fact that shipment costs in many cases consist of a
higher fixed and a lower variable part. The rebalancing of inventory
levels can thus be achieved at an often negligible additional cost if
a reactive transshipment is made. We extend this idea and implement it in a multi-item setup. Our model describes a multi-location
inventory network facing compound, non-homogeneous Poisson demand. Instances of demand have an underlying discrete, multivariate
distribution. We derive a quasi-myopic policy by applying a dynamic
programming policy improvement step to a no-transshipment policy.
Transshipment costs are modelled with a knapsack-like structure to
accommodate different types of items. We carry out an extensive simulation study to show the benefit of modelling multi-item transshipments against operating single item models in parallel.
∗
2. A chance constrained model for VRPTW with
uncertain demands and travel times
Pinar Dursun1,∗ and Erhan Bozdağ2
1
Istanbul Technical University Department of Industrial Engineering;
[email protected] Istanbul Technical University;
In this study a model is developed to solve the vehicle routing problem with time windows (VRPTW) with uncertain demands and travel
times. Although the distances are discrete between two nodes, the
travel times may not be discrete because the traffic has uncertain nature. It is important to delivery order to customers in specified time
window, so handling the uncertainty is more critical. Also the vehicles
are capacitated so they must be loaded below their capacity. Chance
constrained model based random key represented genetic algorithm is
developed and numerical experiments are solved to show effectiveness
of the proposed algorithm.
∗
Optimisation III
Room: A24 (14:00 - 15:30)
Chair: Martin Takac
1. Product form of the inverse revisited
Péter Tar1,∗ and István Maros1
University of Pannonia; ∗ [email protected]
Using the simplex method is one of the most effective ways to solve
linear optimization problems. The efficiency of the solver procedure
is crucial for solving large-scale problems. The solution is obtained
by an iterative procedure, where each iteration can be represented by
a basis of the linear equation system. During an iteration some vectors must be multiplied by the inverse of the actual basis. In order to
speed up these operations, proper basis handling procedures must be
applied.
Two methodologies exist in the state-of-the-art literature, the product
form of the inverse (PFI) and LU factorization. The majority of the
LU methods is widely used, because 120-150 iterations can be done
without the need of re-factorization, and the PFI can serve about 30-60
iteration without re-inversion in order to provide numerical stability.
In our work we revisited the PFI and implemented it in such a way that
hundreds or sometimes even few thousands of iterations can be done
1
SCOR 2012
Abstracts
Saturday
without losing accuracy. The novelty of our approach is in the processing of the non-triangular part of the basis, based on block triangularization algorithms. The resulting inverse of the modified algorithm
performs way better than those found in the literature. These results
can shed new light on the usefulness of the PFI.
“This publication/research has been supported by the TÁMOP4.2.2/B-10/1-2010-0025 project.”
2. Parallel coordinate descent method for composite objective
Martin Takáč1,∗ and Peter Richtárik1
1
The University of Edinburgh; ∗ [email protected]
In this work we show that randomized block (coordinate) descent
methods can be accelerated by parallelization when applied to the
problem of minimizing the sum of a semi-separable smooth convex
function and a simple block-separable convex function. The speedup,
as compared to the serial method, and referring to the number of iterations needed to approximately solve the problem with high probability, is equal to the product of the number of processors and a natural
and easily computable measure of separability of the smooth component of the objective function. In the worst case, when no degree of
separability is present, there is no speedup; in the best case, when the
problem is separable, the speedup is equal to the number of processors. Our analysis also works in the mode when the number of blocks
being updated at each iteration is random, which allows for modeling
situations with variable (busy or unreliable) number of processors.
Heuristics / Metaheuristics III
Room: A25 (14:00 - 15:30)
Chair: Penny Holborn
1. Evolutionary techniques for an order acceptance problem
Simon Thevenin1,∗ , Nicolas Zufferey1 and Marino
Widmer2
The problem of the optimal control of discrete systems occurs in many
areas and includes a large number of subproblems. The approach analyzed in this paper is grounded on a dynamic model of decision making, developed in spreadsheet environment, with clearly separated tables representing discrete controlled object: (1) the law of dynamics,
(2) control space and (3) performance criterion. Discrete processes
with different values of the performance criteria are obtained by varying of values of control variables. It is necessary to find the discrete
process that gives the minimum value to the performance criterion.
In this paper, we also develop metaheuristic method for solving described problem, which are based on tabu search and variable neighborhood search. Two well-known methods have been adapted and
implemented in accordance with the described problem, in order to
compare obtained results. An implementation of the methods are realized in Visual Basic for Application and combine with some results
of a simulation model in Excel spreadsheet.
3. Simple heuristics for on-line scheduling of operation theatres
Nor Aliza Abd Rahmin1,∗ and Chris N. Potts1
University of Southampton; ∗ [email protected]
A hospital is an institution for health care that treats patients with specialized staff and equipment. If we focus on the medical facilities of
the hospital, operating theatres form one of the most important and
expensive resources. Therefore, surgery is a critical process in a hospital, not only for the cost but also for the impact of a patient’s health
and quality perception. Disruptions of various types can prevent a
previously constructed plan from being executed. Waiting for treatment due to operating theatre unavailability can result in deteriorating
health and even worse can lead to death. In this paper, we focused on
an operating theatre scheduling problem for emergency and regular
patients. However, long operation times or a high number of emergency patients arriving can lead to a disruption of previous bookings.
We consider an on-line version of this problem where each day a new
schedule is created based on current information such as new patients
who need to be booked into a slot in the operating theatre and previously scheduled patients whose treatment could not be performed
because of operating theatre capacity constraints. We allocate the patients depend on their urgency using a simple heuristic method, updating the schedule every day. Computational results are reported.
1
1
HEC, University of Geneva;
[email protected] DIUF Decision Support &
Operations Research, University of Fribourg, Switzerland;
∗
We consider the order acceptance and scheduling problem, in a single
machine environment, with setups, and regular (i.e. not decreasing)
objective functions. Order acceptance problem is particularly relevant
in a make-to-order production environment, it has gained increasing
popularity these last decades. Given a set of n jobs, the problem is to
decide which jobs to accept, and to schedule them. The objective is
to minimize the penalties brought by rejected jobs and the cost associated with performed jobs. The problem is NP-Hard and there only
exists an exact method. We propose to tackle it with heuristics. We
present and compare local search methods with population based algorithms for the problem. The local search is a tabu search approach
which includes restriction and diversification procedures. Two evolutionary strategies are proposed, a genetic algorithm with local search
and an adaptive memory algorithm. Tests are performed using realistic
instances and show that the proposed heuristics are very competitive.
2. Meteheuristics for optimal control of discrete
systems
Lena Djordjevic1,∗ , Slobodan Antic1 and Minja
Marinovic1
1
∗
Faculty of Organizational Sciences;
[email protected]
SCOR 2012
Transport II
Room: A26 (14:00 - 15:30)
Chair: Urszula Neuman
1. Vehicle routing with dependent vehicles
Edward Kent1,∗ and Jason Atkin1
University of Nottingham; ∗ [email protected]
The vehicle routing problem has been studied extensively throughout academia, and many solution methods have been produced that
generate a number of independent routes for vehicles. We have been
working with Optrak (a vehicle routing software and consultancy company), and using real world data. Extra constraints arise from these
richer problems, resulting in vehicle routing problems that require
consideration of the dependencies between vehicles. In particular,
when product mixing rules are applied to the demanded products and
time windows are applied to these customers, more than one vehicle
may be needed to service the customer in a small space of time. Products that can’t be mixed may need to be delivered at almost the same
time for the benefit of the customer. Loading limitations at customers
and depots may also present extra constraints that limit the number of
vehicles that can be at a depot or a customer at one time. Other real
world constraints include routes between customers which may contain tunnels, causing vehicles to have to deliver hazardous goods first
before they can pass. A fully loaded vehicle that contains hazardous
1
31
Abstracts
Saturday
goods would then have to make other stops first, potentially causing
other dependent vehicles to have to wait. Together, these constraints
may create narrow time windows which need to be met. Thus, some
conventional assumptions, such as the single route per customer and
the independence of vehicles no longer apply.
2. Capacity planning for motorail transportation
Pascal Lutter1,∗
Ruhr University Bochum; ∗ [email protected]
Loading cars and motorbikes on trains is a challenging task for motorail companies. Several MILP models exist to load containers on
trains. Although the general model structure is similar, motorail train
loading differs strongly from container loading regarding certain technical requirements. In contrast to recent work on auto-carrier loading
we consider route specific requirements and physical constraints in
more detail and additionally focus on order acceptance in a dynamic
environment. Particularly the integration of the booking process distinguishes our approach from previous ones. Orders are placed at different points of time and must be accepted or rejected immediately so
the problem has to be solved repeatedly with varying information. Optimal train utilization can be achieved by the suggested MILP model if
vehicle specifications are exactly known. During the booking process
decisions have to be made immediately for every arriving booking request. In practice, the frequent use of an optimization model is impossible because of long runtimes and missing the integration in current
booking systems. Hence, it is necessary to identify risks and chances
in advance to control available capacity. We propose a mixed-integer
linear programming model which determines maximum weights for
all possible vehicle specifications to optimally decide about the acceptance of orders. We compare our new approach with a simple heuristic
method by a simulation study. The proposed models can be solved using standard commercial software as well as open source software and
their performance is evaluated.
1
Leanne Smith1,∗ , Paul Harper1 , Vincent Knight1 ,
Israel Vieira1 and Janet Williams1
1
Response time targets for the Welsh Ambulance Service NHS Trust
(WAST) are not currently being met. In particular, the more rural
areas in South East Wales consistently perform poorly with respect
to the target of reaching 65% of the highest priority emergency calls
(category A) within 8 minutes, and are amongst the worst in the UK.
This research is concerned with developing mathematical models for
the ambulance service system, utilising methods drawn from location
theory, to help WAST make better decisions on locations, capacities
and deployments. Findings from a developed location/allocation optimisation model, the Maximal Expected Survival Location Model for
Heterogeneous Patients (MESLMHP), will be used to suggest suitable
allocations of vehicles at stations across the South East region (and
subsequently for other areas of Wales), with the aim of maximising
the overall expected survival probability of patients given a homogeneous or heterogeneous fleet. The model can incorporate both survival
functions and response time targets allowing for different emergency
categories to be considered. The model will be run under various scenarios of interest to WAST; changes will be made to the demand of
calls, number of available vehicles and shift patterns to see the impact
on the expected number of survivors. The allocation obtained may
also be used as input to a detailed discrete event simulation. Recommendations will be made to the Trust to help provide a more efficient
and effective ambulance service to their population, and to achieve the
targets as set by the Welsh Assembly Government.
2. Optimising the use of resources within the district nursing service
Elizabeth Rowse1,∗ , Paul Harper1 , Janet Williams1
and Mark Smithies1
1
3. Integration aspects of the gate allocation problem
Urszula Neuman1,∗ , Jason Atkin1 and Edmund
Burke2
University of Nottingham; ∗ [email protected] University of
Stirling;
Smooth and punctual operation is a priority for most airports in the
world. The better the operation, the easier it is for an airport to attract
airlines and, in this way, increase the number of passengers. Competition between airports is often not observed directly by passengers,
but it exists, is observed by airlines, and can be fierce in some regions.
Many aspects of airport operations (e.g. runway sequencing, taxiing,
gate allocation) are currently managed in isolation, i.e. without considering the effects of decisions upon other aspects. This is likely to
reduce efficiency of an airport as a whole. Integration of these various
aspects could potentially improve airport operations. Obviously, integrating runway sequencing with taxiing would reduce the time that
an aircraft needs before taking off (after leaving a gate) and to get to a
gate (after landing). Way in which the gate allocation process could be
improved to aid these two processes will be discussed, along with the
difficulties which occur in the integration process and how integration
may contribute to smooth airport operations.
1
Healthcare
Room: A08 (16:00 - 17:30)
Chair: Izabela Komenda
1. Allocating Welsh emergency medical services
to maximise survival
32
Cardiff University; ∗ [email protected]
Cardiff University; ∗ [email protected]
Recent statements from the UK government indicate that future provision of services within the National Health Service will involve the
transition of care from hospitals into the community. District nurses
play an important role in caring for housebound patients whilst alleviating some pressure on other primary care services. An increase
in the number and complexity of patients’ needs treated within the
community, coupled with the predicted decline in the number of district nurses poses a potential supply and demand problem. Working
closely with a district nursing service in Wales, the optimal size and
skill mix of district nursing teams to meet patient demand is investigated. A two-stage model is developed that uses Monte Carlo simulation to generate patient demand, and Linear Programming to find
an optimal team composition that meets this patient demand at minimum cost. This approach, novel to workforce planning in the district
nursing service, gives results that indicate significant cost savings if
district nursing teams are restructured for optimal skill mix.
3. Queueing theory accurately models critical
care units
Izabela Komenda1,∗ , Jeff Griffiths1 and Vincent
Knight1
1
Cardiff University; ∗ [email protected]
Random number of arrivals and random length of stay make the number of patients in a Critical Care Unit (CCU) behave as a stochastic
process. This makes the determination of the optimum size of the bed
capacity very difficult. The number of admissions per day, length of
stay and bed occupancy are key control parameters to find the optimal number of beds required. In this study queueing theory is used
to develop a new mathematical model of patients’ flow and is applied
to two CCUs in hospitals from the same Health Board. Predictions
SCOR 2012
Abstracts
Saturday
from the model are compared to observed performance of the Units,
and the sensitivity of the model to changes in Unit size is explored.
The model is also used to analyse the effect of transferring patients
and beds from one hospital to another. We conclude that cooperation
between both hospitals helps to achieve lower utilisation rates with a
smaller probability of rejection.
and medium size battery. The outcome of this research will provide
an insight on how to handle peak demands with unexpected low wind
generation at lowest possible cost. The OR viewpoint of this research
aims to shed new light into the design of energy systems for the future.
3. Air cargo revenue management
Emily Cookson1,∗ , Kevin Glazebrook1 and Joern
Meissner2
Optimisation IV
1
Room: A24 (16:00 - 17:30)
Chair: Emily Cookson
1. Best next sample
Sergio Morales Enciso1,∗
1
∗
The University of Warwick;
[email protected]
We consider the scenario of a memoryless market to sell one product,
where a customer’s probability to actually buy the product depends on
the price. We would like to set the price for each customer in a way
that maximizes our overall revenue. In this case, an exploration vs exploitation problem arises. If we explore customer response to different
prices, we get a pretty good idea of what customers are willing to pay.
On the other hand, this comes at the cost of losing a customer (when
we set the price too high) or selling the product too cheap (when we set
the price too low). The goal is to infer the true underlying probability
curve as a function of the price (market behaviour) while simultaneously maximizing the revenue. This paper focuses on learning the
underlying market characteristics with as few data samples as possible, exploiting the knowledge gained from both exploring potentially
profitable areas with high uncertainty and optimizing the trade-off between knowledge gained and revenue exploitation. The response variable being binary by nature, classification methods such as logistic
regression and Gaussian processes are explored. This makes the analytic calculations intractable, so a series of approximations are used,
while the knowledge gain optimization is approximated by a dynamic
program. A series of simulations of the evolution of the proposed
model and a statistical analysis are finally presented to summarize the
results.
∗
Lancaster University;
[email protected] Kuehne Logistics University;
The air cargo industry faces some unique challenges: Highly volatile
demand, weight/volume uncertainty through variable tendering, and
short booking cycles are some of the features. In the case of mixed
air carriers, the prioritisation of passengers and their baggage further
challenges the allocation of aircraft belly space. Our research considers these challenges and focuses on modelling a new dynamic programming formulation that an air cargo company could use to maximise its expected profit. In particular, the formulation differs from
previous ones in that capacity and demand uncertainty is incorporated into the model using probability distributions. The main focus
is on spot-sale bookings, making acceptance decisions using dynamic
prices formulated under uncertainty. The initial results suggest this
dynamic model may give significant improvement over static models.
The research is being conducted in collaboration with the cargo division of a major airline. Data analysis has produced interesting results
that have been incorporated into the model.
Scheduling / Timetabling I
Room: A25 (16:00 - 17:30)
Chair: Stefan Ravizza
1. Scheduling games and referees in football: the
last song of a Red Hot Chilean Rocker
Mario Guajardo1,∗ , Fernando Alarcón2 and
Guillermo Durán3
1
2. Optimization modeling of distributed energy
systems for a smart grid
Pedro Crespo Del Granado1,∗ , Stein W. Wallace1
and Zhan Pang1
1
∗
Lancaster University;
[email protected]
UK’s decarbonisation policies are increasing the share of renewable
resources in the energy generation mix to 20-30%. Wind power will
deliver most of the renewable output by 2020. Due to uncertainty in
availability of wind generation, this new energy mix creates new planning challenges to maintain a stable and reliable supply-demand balance. From an OR perspective, this research focuses on modeling the
flexibility and robustness of the energy infrastructure in the presence
of short term extreme events. Through a bottom-up approach the research is an understanding of how synergies of generating units work
in sub-energy systems, such as a house, or small communities. Which
combination of energy generation will cope with high wind energy input? Through a dynamic portfolio optimization model, we assessed
the energy mix for a typical UK house to understand the value of storage. We observed that battery storage in households can shift house
peak demands to less stressful periods for the grid. Results showed a
20-25% energy cost saving for end users. We also have modeled the
energy mix of a small community (University Campus) in the form of
a stochastic optimization model that considers wind generation, CHP,
SCOR 2012
NHH Norwegian School of Economics and Business
Administration; ∗ [email protected] University of
Chile;3 University of Chile, University of Buenos Aires;
The use of sports scheduling in the Chilean football has run for years,
as a result of collaboration between managers of the Football Association and researchers in OR. The problems in the project include
the scheduling of games of several tournaments and the assignment
of referees to the games. As part of the research group, in this talk
I will present an overview of the achievements of the project and I
will detail our most recent work on the referee assignment problem.
Given a schedule of games, the problem consists of assigning referees to the games, fulfilling a number of operational and fairness constraints. We propose an integer programming model, whose natural
formulation can be solved by using standard solvers. However, large
instances may lead to relatively long solution times, an undesirable
matter for practical use. We propose a two-stages solution approach
based on patterns. Firstly, we generate the patterns for each referee
by solving an IP model that considers some constraints of the problem. The patterns indicate the set of games to which a referee can be
assigned in each round. Secondly, we implement another IP model
that incorporates the remaining constraints and assigns the referees to
the games. With this approach, we reduce the solution time significantly. To our knowledge, this is the first work proposing a pattern
approach for referee assignment. Moreover, while the scheduling of
games has recently shown a massive development, the literature of
sports scheduling applied to real-world referee assignment problems
is still scarce.
33
Abstracts
Saturday
2. Over-constrained airport baggage sorting station assignment problem
Amadeo Ascó1,∗ and Jason Atkin1
1
University of Nottingham; ∗ [email protected]
Correct assignment of airport resources can greatly affect the quality of service which airlines and airports provide to their customers.
Good assignments can help airlines and airports to keep to published
schedules, by minimising changes in these schedules and reducing delays. Given the expected increases in civil air traffic, the complexities
of resource scheduling and assignment continue to increase. For this
reason, as well as the dynamic nature of the problems, scheduling and
assignment are becoming increasingly difficult.
The assignment of Baggage Sorting Stations to flights is one of the
resource assignment problems in an airport, and like many other real
world optimisation problems, it naturally has several objectives, which
conflict with each other. We present a model for the problem, look at
different approaches for obtaining good solutions and study these to
gain an insight into their qualities.
A network design problem consists in locating facilities (nodes and
arcs) that enables the transfer of flows (passengers and/or goods) from
given origin-destination pairs. The topic can have several applications within transportation and logistics contexts. As various criteria
can be assumed as objective functions, the problem can be formulated
through multi-objective models. In the literature of multi-objective
network design problems two fundamental bi-objective models can be
emphasized: the Maximum Covering Shortest Path (MCSP) model
and the Median Shortest Path (MSP) model. Both the models determine paths on a network considering a trade off between a measure of
efficiency (the path length) and a measure of attractiveness (the coverage and the users’ distance). In this work we propose multi-objective
models in which also balancing or equity aspects, i.e. measures of the
distribution of distances of users from the path, are considered. These
kinds of models can be used when there is the need to balance risks
or benefits among all the potential users deriving from the location of
the path to be designed. In particular we illustrate bi-objective models
whose optimization criteria are similar to the MCSP and MSP objectives, but including a certain balancing measure as constraint. The
application of the proposed models to a benchmark problem used in
literature to test these kinds of models, show that they are able to find
Pareto solutions characterized by significant level of equity.
3. Probabilistic airline reserve crew scheduling
2. Organization of a public service through the
model
solution of a districting problem
Christopher Bayliss1,∗ , Jason Atkin1 and Geert De
Carmela Piccolo1,∗ and Sabrina Graziano1
Maere1
1
University of Nottingham; ∗ [email protected]
This paper introduces a probabilistic model for airline reserve crew
scheduling. The model can be applied to any schedules which consist
of a stream of departures from a single airport. Reserve crew demand
can be captured by a single probability (probability of crew absence)
for each departure. The aim of our model is to assign some fixed
number of available reserve crew in such a way that the overall probability of crew unavailability in an uncertain operating environment
is minimised. A comparison of different probabilistic objective functions, in terms of the most desirable simulation results, is carried out,
complete with an interpretation of the results. Based on the best objective function, a sample of heuristic solution methods are tested and
compared to the optimal solutions on a set of problem instances. The
current model can be applied in the early planning phase of reserve
scheduling, when very little information is known about crew absence
related disruptions. The main conclusions include the finding that the
probabilistic objective function approach gives solutions whose objective values correlate strongly with the results that these solutions will
get on average in repeat simulations. Minimisation of the sum of the
probabilities of crew unavailability was observed to be the best surrogate objective function for reserve crew schedules that perform well
in simulation. A list of extensions that can be made to the model is
then given, followed by a conclusion that summarises the findings and
important results obtained.
Graphs / Networks
Room: A26 (16:00 - 17:30)
Chair: Michael Clark
1. The design of transportation networks: a multi
objective model combining equity, efficiency and
efficacy
Maria Barbati1,∗
1
∗
University of Naples Federico II;
[email protected]
A districting problem consists in subdividing a given region into a
certain number of sub-regions (districts) on the basis of an objective
function to optimize. This kind of problem can occur in different fields
of application when a service (public or private) has to be provided in
a region. The decision about the number of districts and the allocation of the potential users to the district affects the management of the
service in terms of costs and accessibility of the service. In order to
solve these problems, various models have been proposed with different characteristics in dependence of the service type, the objective
and the constraints to be considered. After a general description of the
models able to represent districting problems, we propose a mathematical formulations to define appropriate districts when a public service
has to be organized and managed on a territory. An application of the
model to the case of a school districting problems is described and the
results of a real case are analyzed and discussed.
3. Modelling the uncertainty of data and the robust shortest path
Michael Clark1,∗ and Andrew J. Parkes1
1
University of Nottingham; ∗ [email protected]
Most optimisation problems rely on perfect information in order for
the algorithms to perform well. Often in real world situations this
information is not available. Modifying an algorithm to handle such
situations can lead to complications and in many cases is not possible.
Changing a model to handle uncertain data increases the complexity of
the problem. As a result of this measuring the quality of the solution
is no longer a trivial process leading to a need for new approaches.
Here we will describe different ways to model uncertainty and discuss their application to the robust shortest path problem. We will
discuss different ways to calculate the cost of robustness and show a
simple algorithm, which is based on the A* Search Heuristic. We will
explain how graphs are generated, allowing different characteristics
to be modelled. We will show the experiments run on the generated
graphs and discuss the results.
1
University of Naples Federico II - Department of Engineering
Management (DIEG); ∗ [email protected]
34
SCOR 2012
Abstracts
Sunday
Multicriteria Decision Analysis II
Room: A08 (10:00 - 11:00)
Chair: Martin Takac
1. A non-identical parallel machines scheduling
problem with multi-objective minimization
Jean Respen1,∗ , Nicolas Zufferey1 and Edoardo
Amaldi2
1
∗
HEC, University of Geneva;
[email protected] Politecnico di Milano;
Multi-objective production problems are becoming nowadays a regular faced issues for decision-makers who need to take the more accurate decision at the right moment. As exact methods are unable
to tackle this kind of problems in a fair amount of time, researchers
started to develop new and quick methods, called metaheuristics,
which are able to find near-optimal solutions in reasonable amount of
time. In this context, we propose a multi-objective scheduling problem based on non-identical parallel-machines. Makespan, setup costs
and times, as well as smoothing issues are considered, and eligibility constraints are fulfilled. To tackle this complex problem, we propose different metaheuristics, such as tabu search and adaptive memory techniques, as well as an exact method (for comparison purposes
on the small instances). We show that our algorithms are fast, efficient, and robust, even for big instances where exact methods cannot
be considered, due to exponential computation times. Current results
for the considered problems are presented and future works conclude
the presentation.
Hyper-heuristics have drawn increasing attention from the research
community in recent years, although their roots can be traced back to
the 1960’s. They perform a search over the space of heuristics rather
than searching over the solution space directly. Research attention
has focussed on two types of hyper-heuristics: selection and generation. A selection hyper-heuristic manages a set of low level heuristics
and aims to choose the best heuristic at any given time using historic
performance to make this decision, along with the need to diversify
the search at certain times. In this study, we propose a choice function based hyper-heuristic for multi-objective optimization that controls and combines the strengths of three well-known multi-objective
evolutionary algorithms (NSGAII, SPEA2, and MOGA), which are
utilised as the low level heuristics. A choice function acts as the high
level strategy, which adaptively ranks the performance of three lowlevel heuristics, deciding which one to call at each decision point. “All
Moves” is employed as an acceptance strategy, meaning that we accept the output of each low level heuristic whether it improves the
quality of the solution or not. Four performance metrics (Algorithm
effort (AE), Ratio of non-dominated individuals (RNI), Size of space
covered (SSC) and Uniform distribution of a non-dominated population (UD)) act as an online learning mechanism to provide knowledge
of the problem domain to the high level strategy. The experimental
results demonstrate the effectiveness of this hyper-heuristic approach
when tested on the Walking Fish Group test suite, a common benchmark for multi-objective optimization.
2. A hybrid encoding scheme for grouping problems
Anas Abdalla Osman Elhag1,∗ and Ender Özcan1
1
2. Decision analysis for cost effective maintenance of trunk roads
Ena Orugbo1,∗ and Alkali Babakalli1
1
Glasgow Caledonian University; ∗ [email protected]
This paper investigates the dynamics of road network category1 defect
accumulation and failure process. Category1 defects are failings that
significantly accelerate structural deterioration and present hazards to
road users. The study establishes appropriate insight into hazards
related failures and interrelationship between trunk road sub-assets.
Higher incidence of defects due to up-comings such as growing traffic
has led to increased maintenance works. This works impede efforts
to meet ever increasing demands for safe and reliable journeys. Reliability Centred Maintenance (RCM) and a Delphi expert survey are
conducted on XYZ trunk road network. A Multiple Criteria Decision
Analysis (MCDA) is to be conducted using Analytical Hierarchical
Process (AHP) to set the pace for the development of a Road Defect
Maintenance Model (RDMM). The results from the RCM, downtime
and sub-asset history failure data analysis are presented. Further investigation and modelling that could support cost-effective trunk road
maintenance decisions is on-going.
Optimisation V
Room: A24 (10:00 - 11:00)
Grouping problems represent a family of combinatorial optimization
problems where the task is to partition a single set of objects into a collection of mutually disjoint subsets such that each object is in exactly
one subset. Linear Linkage Encoding and the Genetic Grouping Algorithm Representation are two different encoding schemes that have
been used to solve the grouping problems. In this study, a hybrid encoding scheme that combines the benefits from both representations
is investigated. Additionally, crossover and mutation operators based
on the hybrid encoding scheme are described. These operators can be
used in any appropriate algorithmic framework for solving grouping
problems. The initial results will be provided at the conference using
a selection hyper-heuristic framework for graph colouring as a case
study.
Scheduling / Timetabling II
Room: A25 (10:00 - 11:00)
Marcin Siepak1,∗
Chair: Alessia Violin
1. A choice function based hyper-heuristic for
multi-objective optimization
Mashael Maashi1,∗ , Graham Kendall1 and Ender
Özcan1
University of Nottingham; ∗ [email protected]
SCOR 2012
Chair: Urszula Neuman
1. An exact algorithm for the uncertain version of
parallel and identical machines scheduling problem with interval processing times and total completion time criterion
1
∗
1
University of Nottingham; ∗ [email protected]
Wroclaw University of Technology;
[email protected]
An uncertain version of parallel and identical machines scheduling
problem with the total completion time criterion is considered. It is
assumed that the execution times of tasks are not known a priori but
they belong to the intervals of known bounds. Such a way of uncertainty description is useful in the cases where no any historical data
is available regarding the imprecise parameters, which would be required in order to obtain the probability distribution and apply the
35
Abstracts
Sunday
stochastic approach and also when there is lack of experts opinions
which would be a source of other representations of uncertain execution times, e.g. in the form of membership function for the fuzzy
approach. The absolute regret based approach for coping with such
an uncertainty is applied. This problem is known to be NP-hard and
a branch and bound algorithm (B&B) for finding the exact solution
is developed. The results of computational experiments show that for
the tested instances of the uncertain problem – B&B works significantly faster that the exact procedure based on a simple enumeration.
The algorithm proposed has application for further research of quality
evaluation for planned to develop heuristic and approximate solution
approaches for the considered problem - in order to check how far
from the optimality are solutions generated by them. It also allows to
obtain the exact solution for small instances of the uncertain problem
faster than an algorithm based on a simple enumeration.
2. Optimizing real-world workforce scheduling
problems
Nico Kyngäs1,∗
1
∗
Satakunta University of Applied Sciences;
[email protected]
The process of constructing optimized work timetables for the personnel is an extremely demanding task, hence the use of decision support
systems for workforce scheduling has become increasingly important
for both the public sector and private companies. Good rosters have
many benefits for an organization, such as lower costs, more effective
utilization of resources and fairer workloads. The workforce scheduling process includes four phases: 1) shift generation is the process of
determining the shift structure, 2) in preference scheduling employees’ wishes are fulfilled as well as possible, 3) days-off scheduling
deals with the assignment of rest days between working days, 4) staff
rostering deals with the assignment of employees to shifts. The paper
presents a process and a method for optimizing real-world workforce
scheduling instances. My current research deals with shift generation
and large-scale staff rostering, hence those topics will be given emphasis in the presentation. A population-based local search heuristic
called PEAST is used to solve the workforce optimization phases. The
generated software is in real-world use.
Tarifa Almulhim1,∗ and Ludmil Mikhailov1
∗
Manchester Business School;
[email protected]
A non-linear extension of the fuzzy preference programming (FPP)
method for deriving group priorities in the Fuzzy Analytical Network
Process (FANP) is proposed. The proposed method considers the different important weights for multiple decision makers (DMs). Additionally, it allows for representation DMs preferences as fuzzy numbers rather than exact numerical assessments in order to tackle the
uncertainty and imprecision in human thinking. The proposed method
also transfers the group prioritization problem in the FANP into a nonlinear program. Unlike the known fuzzy prioritization techniques, the
efficacy of the proposed method is demonstrated for deriving crisp
weights from incomplete and inconsistency fuzzy set of comparison
judgements. Moreover, it doesn’t require additional aggregation producers and provides a consistency index for measuring the inconsistency of the DMs’ uncertainty judgements. Numerical examples are
36
Timo P. Kunz1,∗
1
Lancaster University; ∗ [email protected]
In recent years, Revenue Management has found its way into retail
industry practice, where previously high counts of sales outlets and
products, and the resulting masses of data had limited the proliferation of the revenue management idea. However, price optimization
theory and practice are still largely disconnected. While the relevant
contributions in the area of forecasting and demand modelling are usually driven by empirical data but rarely address the pricing problem,
the price optimization and revenue management literature tends to focus on the theoretical optimization problem paying little attention to
the empirical relevance and applicability of the underlying demand
model. Our research aims to fill this gap by taking an integrated look
and highlighting the interplay of the individual components of the optimization system, namely the interaction between optimization and
demand models along with the empirical data that is used for the estimation and the calibration of the later. The findings up to this point
suggest that the most pressing issue from an operational perspective
are to be found in the applicability and empirical validation of the
prevalent market share modelling approach via attraction/choice models. Further the work aims to highlight the relevance of the category’s
competitive structure for the demand modelling and the optimization
problem alike.
Simulation / System Dynamics
Room: A08 (13:30 - 15:00)
Chair: Magdalena Gajdosz
1. The transition to an energy sufficient economy:
a system dynamics model for energy policy evaluation in Nigeria
1
Chair: Stefan Ravizza
1. Deriving priorities from fuzzy group comparison judgements in the fuzzy analytical network
process (FANP)
1
2. Modelling retail sales for retail price optimization
Timothy Mbasuen1,∗ and Richard C. Darton1
Decision Support
Room: A26 (10:00 - 11:00)
illustrated the implement of the proposed method and compared to the
existing fuzzy prioritization method.
University of Oxford; ∗ [email protected]
Nigeria is an energy-rich developing country with a huge energy resource base. The country is currently the largest reserves holder and
largest producer of oil and gas in the African continent. However, despite this, only 40% of the country’s 158 million people have access
to modern energy services. About 80% of its rural dwellers depend
almost wholly on traditional biomass for their energy needs. Several
attempts by the government; to improve this situation have failed to
produce the desired results. This paper presents an overview of ongoing research being undertaken to examine energy policies in Nigeria,
with the aim of identifying and quantifying the barriers of sustainable
energy development in that country. System dynamics modelling is
shown to be a useful tool to map the interrelations between critical
energy variables with other sectors of the economy, and for understanding the energy use dynamics within the economy. It is found in
our preliminary analysis that the critical factors are lack of energy sector capacity utilisation, inadequate energy financing/investment, lack
of trained and suitably qualified manpower, as well as inconsistencies and policy reversals. These remain the key challenges hampering
Nigeria’s smooth transition from energy poverty to an energy sufficient economy.
SCOR 2012
Abstracts
Sunday
2. On the Peter principle: an agent based investigation into the consequential effects of social networks and behavioural factors
Angelico Fetta1,∗ , Paul Harper1 , Vincent Knight1 ,
Israel Vieira1 and Janet Williams1
1
Cardiff University; ∗ [email protected]
The Peter Principle is a theory that provides a paradoxical explanation for job incompetence in a hierarchical organisation. It argues that
should staff be competent at a given level, their competence may not
be implicit at higher levels due to the differences in the skill set required. Furthering the work of a recent investigation into the Peter
Principle utilising Agent Based Simulation, this paper explores external factors upon varying promotion strategies to assess efficiency.
Through additional elements of social networks, behavioural dynamics and social capital, a more representative view of workplace interaction is presented.
Results from the simulation found that although the Peter Principle
affects efficiency, it may not be to the levels previously suggested and
could be influenced by social network topology. Furthermore promotion on merit provided the most favourable maximum and minimum
efficiency margins, given the absence of clear evidence pertaining to
the existence of the Peter Principle.
3. Empirical bayes methods for discrete event
simulation
Shona Blair1,∗ , John Quigley1 and Tim Bedford1
1
∗
University of Strathclyde; [email protected]
Discrete event simulation (DES) is widely utilized in OR applications
for the design, analysis and improvement of complex, dynamic and
stochastic real-world systems. One of the key advantages of discrete
event simulation is its ability to incorporate a “realistic” level of system complexity into the analysis process, as opposed to the more rigid
assumptions of alternative modelling techniques. This, however, frequently results in simulation models which are large-scale, structurally
complex and computationally expensive to run. As such, careful statistical analysis of experimental results is necessary to ensure efficient
use of DES models. Empirical Bayes (EB) procedures offer a structured and theoretically sound framework for the pooling of data obtained across a set of populations to support inference concerning the
parameters of an individual population. This often enables more efficient inference in situations which feature a repeated structure, providing that sufficient “similarity” exists between component elements.
It seems intuitively reasonable that such an approach may be of benefit in DES model experimentation, owing to the underlying similarity
between model configurations. In light of the computational expense
involved in executing simulation models, such increased efficiency in
estimation would likely prove highly advantageous in practice. Despite this potential, EB has so far been neglected in the simulation
literature. In this talk, the results of a pilot study into the use of EB
procedures in the estimation of DES performance measures will be
presented. In addition, the practical significance of the results, and
directions for further research will also be discussed.
Chair: Pablo Gonzalez Brevis
1. The k-separator problem
Mohamed Sidi Mohamed Ahmed1,∗ , Walid
Ben-Ameur1 and Jose Neto1
SCOR 2012
Telecom Sudparis;
[email protected]
Let G = (V;E;w) be a vertex-weighted undirected graph and k be a
positive number. We want to compute a minimum-weight subset of
vertices S whose removal leads to a graph where the size of each connected component is less than or equal to k. Let us call such a set a
k-separator. If k = 1 we get the classical vertex cover problem. The
case k = 2 is equivalent to compute the dissociation number of a graph
(in the case of unit weights). This problem is NP-hard even if the
graph is bipartite. The k-separator problem has many applications. If
vertex weights are equal to 1, the size of a minimum k-separator can
be used to evaluate the robustness of a graph. Intuitively, a graph for
which the size of the minimum k-separator is large is more robust.
Unlike the classical robustness measure given by the connectivity, the
new one seems to avoid to underestimate robustness when there are
only some local weaknesses in the graph. The minimum k-separator
problem has also some applications in the context of networks. A classical problem consists in partitioning a graph into different subgraphs
with respect to different criteria. For example, in the context of social
networks, many approaches are proposed to detect communities. By
solving a minimum k-separator problem, we get different connected
components that may represent communities. The k-separator vertices
represent persons making connections between communities.
∗
2. A multi-dimensional multi-commodity covering problem with application in logistics
Alexander Richter1,∗ , Jannik Matuschke1 and Felix
König1
TU Berlin; ∗ [email protected]
In this talk, we study a multi-commodity multi-dimensional covering
problem which we encountered as a subproblem in optimising large
scale transportation networks in logistics. The problem asks for a selection of containers for transporting a given set of commodities, each
commodity having different extensions of properties such as weight or
volume. Each container is specified by a fixed charge and capacities
in the relevant properties and can be selected multiple times. The task
is now to find a cost minimal collection of containers and a feasible
assignment of the demand to all selected containers.
From theoretical point of view, by exploring similarities to the well
known set cover problem, we derive NP-hardness and see that the
non-approximability result known for set cover also carries over to
our problem.
For practical applications we need very fast heuristics to be integrated
into a meta-heuristic framework that—depending on the context—
either provide feasible near optimal solutions or only estimate cost
value of an optimal solution. Thus, in a second part we develop and
analyse a flexible family of greedy algorithms that meet these challenges. In order to find best-performing configurations for different
requirements of the meta-heuristic framework, we provide an extensive computational study on random and real world instance sets obtained from our project partner 4flow.
We outline a trade-off between running times and solution quality
and conclude that the proposed methods achieve the accuracy and efficiency necessary for serving as a key ingredient in more complex
meta-heuristics.
1
Scheduling / Timetabling III
Optimisation VI
Room: A24 (13:30 - 15:00)
1
Room: A25 (13:30 - 15:00)
Chair: Urszula Neuman
1. Flexible mobile workforce scheduling and
routing
Jose Arturo Castillo Salazar1,∗ and Dario
Landa-Silva1
37
Abstracts
University of Nottingham; ∗ [email protected]
In times where employees need to be more flexible and mobile regarding the types of jobs they perform, a range of problems arise.
Such problems, like home health care and technicians scheduling, require employees travelling using different transportation modalities.
Employees move across many diverse locations to do work related
tasks. Tasks are activities which have specific starting time and duration. Moreover, employees necessitate appropriate skills related to
their performing tasks. Skills vary for every employee, resulting in a
widely diverse workforce. Some tasks need to be synchronised and
appropriately sequenced. The motivation of this study is to gather the
most common characteristics of such problems in order to develop
an algorithm to solve them. In the context of this study, we refer to
these problems as flexible mobile workforce scheduling and routing
(FMWSR). The FMWSR problem is a hybrid one, combining concepts from vehicle routing with time windows and employee scheduling. The outcome is a set of typical features from the literature: time
windows, transportation modality, start and end locations, skills and
qualifications, service time, connected activities, teaming and clusterisation, the last two being optional. Solving FMWSR problems has
many objectives such as reducing employees’ travel time, guaranteeing tasks are performed by qualified personnel and reducing costs.
Devon Barrow1,∗ and Sven Crone1
1
1
Lancaster University; ∗ [email protected]
Since its origins in the mid 1990s, boosting has been successfully
applied in over 2500 studies emerging as one of the best ensemble
algorithms in classification and regression. In contrast a recent survey noted only 15 papers applying boosting to forecasting time series
data. None of these investigate the interaction between the algorithm
meta-parameters and their impact on forecast performance. To close
this gap we describe based on original AdaBoost, a generic algorithm
whose components correspond to different meta-parameter choices.
We evaluate the influence of combination method, loss function, stopping criteria and a new meta-parameter based on the choice of loss
update model. Within this framework well known AdaBoost.R2 and
AdaBoost.RT are easily identified as a choice of these four factors.
Boosting is applied to Multilayer Perceptron (MLP) networks to forecast the NN3 competition real world dataset of 111 time series containing long and short, seasonal and non-seasonal time series. A multifactorial analysis of variance (MANOVA) on forecast errors provides
empirical evidence that different meta-parameter choices have a significant impact on forecast accuracy, with certain parameter choices
proving superior to standard boosting approaches adopted in classification and regression.
2. A case study of investigating a highly constrained search space
2. Long-term reserve warranty forecasting with
neural network models
Lisa Taylor1,∗ , Jonathan Thompson1 and Rhyd
Lewis1
Cardiff University; ∗ [email protected]
In this research we have focussed on a post enrolment-based problem
(International Timetabling Competition 2007), looking particularly at
maximising the connectivity of the search space. When dealing with
timetabling problems that are subject to both hard and soft constraints,
a common strategy is to make use of a two-stage optimisation process.
Stage one consists of satisfying the hard constraints whilst stage two
is to minimise the soft constraint violations.
A highly constrained problem can mean that a search space is disconnected; that is, feasible regions are separated by an infeasible region under certain neighbourhood operators. Our research focuses on
searching highly constrained solution spaces and means of navigating
between feasible regions.
In stage one, we looked at two strategies. The first strategy schedules
all events using saturation degree. The events were then re-arranged
using neighbourhood operators. The second strategy allows unplaced
events and inserts them using a type of maximum matching algorithm.
The second strategy works much better because the first strategy unnecessarily constrains the problem.
Stage two entails optimising the feasible solution in terms of a soft
constraint penalty. We found a strong correlation between the number
of moves that can be performed whilst maintaining feasibility and the
improvement made to the soft constraint penalty. We can use this information to achieve further improvements. For instance, adding more
neighbourhood operators to increase the number of moves that maintain feasibility; and temporarily removing events from the timetable
making more spaces available for the scheduled events to move to.
Shuang Xia1,∗ and Shaomin Wu1
1
Neural Networks / Machine Learning
Room: A26 (13:30 - 15:00)
Chair: Martin Takac
1. A meta-parameter analysis of boosting for time
series forecasting
38
1
Cranfield University; ∗ [email protected]
Forecasting future warranty reserves is vitally important for warranty
suppliers. Many techniques in this area exist in the literature. However, little work has been done on long-term warranty reserve forecasting, while it is an important issue in making fiscal plan. This paper
develops two neural network models to forecast warranty reserves. It
compares the two techniques based on both artificially generated data
and data collected from an electronics product manufacturer.
3. Inventory optimization under process flexibility assumption using approximate dynamic programming approaches
Mustafa Cimen1,∗ , Kevin Glazebrook1 and
Christopher Kirkbride1
1
Lancaster University; ∗ [email protected]
Classical inventory problem requires optimizing two decisions: when
and how much to produce. Even though process flexibility (being able
to produce multiple products in each factory) provides the ability to
react rapidly to changes in or results of stochastic environment, inventory decisions are even more complex under this assumption, as decision maker also needs to give decisions of how much to produce each
product in each factory, which depends on each other. In this study,
we use Machine Learning (known also as Approximate Dynamic Programming (ADP) in OR literature) approaches to cope with this complex system. We define a small-sized stochastic inventory problem under capacity constraint and process flexibility assumptions, and solve
the decision problem by several look-up table ADP approaches and
step-size parameter selections. Two main goals of these computations
are (i) showing how capable ADP is to cope with this kind of problems, and (ii) making a comparison of performances of various ADP
algorithms.
SCOR 2012
Index of Authors
Index of Authors
Abd Rahmin, Nor Aliza, 31
Al Hinai, Ahmed, 28
Almulhim, Tarifa, 36
Armstrong, Stanislava, 23
Ascó, Amadeo, 34
Baeklund, Jonas, 24
Barbati, Maria, 34
Barrow, Devon, 38
Bayliss, Christopher, 34
Blair, Shona, 37
Booker, Carolyn, 23
Castillo Salazar, Jose Arturo, 37
Chen, Hanyi, 26
Cimen, Mustafa, 38
Clark, Michael, 34
Cookson, Emily, 33
Crespo Del Granado, Pedro, 33
D. Nasiri, Saeideh, 27
Djeumou Fomeni, Franklin, 26
Djordjevic, Lena, 31
Djordjevic, Milan, 26
Drake, John, 29
Dursun, Pinar, 30
Peano, Andrea, 24
Piccolo, Carmela, 34
Rauscher, Sandra, 30
Reddy, Brian, 25
Respen, Jean, 35
Richter, Alexander, 37
Rostami Tabar, Bahman, 28
Rowse, Elizabeth, 32
Schulze, Tim, 26
Schurr, Jochen, 26
Shone, Rob, 23
Siepak, Marcin, 35
Silvera, Yolanda, 27
Smith, Leanne, 32
Soulby, Matthew, 27
Swarat, Elmar, 23
Takáč, Martin, 31
Tar, Péter, 30
Taylor, Lisa, 38
Thevenin, Simon, 31
Umofia, Anietie, 29
Velandia-Brinez, Cesar, 24
Elhag, Anas Abdalla Osman, 35
Williams, Julie, 28
Fetta, Angelico, 37
Xia, Shuang, 38
Gajdosz, Magdalena, 29
Geranios, Michail, 30
Goetzmann, Kai-Simon, 25
González-Brevis, Pablo, 24
Guajardo, Mario, 33
Holborn, Penny, 27
Huynh, Nha-Nghi, 28
Kent, Edward, 31
Kheiri, Ahmed, 29
Komenda, Izabela, 32
Kunz, Timo P., 36
Kyngäs, Nico, 36
Lai, Michela, 27
Lutter, Pascal, 32
Maashi, Mashael, 35
Mari, Renato, 25
Marinović, Minja, 23
Mbasuen, Timothy, 36
Mohamed Ahmed, Mohamed Sidi, 37
Morales Enciso, Sergio, 33
Neuman, Urszula, 32
Ntio, Despoina, 30
Orugbo, Ena, 35
SCOR 2012
39