PRE2019 3 Group16

Group Members

Name	Student Number	Study	Email
Zakaria Ameziane	1005559	Computer Science	z.ameziane@student.tue.nl
Cahitcan Uzman	1284304	Computer Science	c.uzman@student.tue.nl
Efe Utku	1284290	Applied Physics	e.utku@student.tue.nl
Roel den Hoet	1248170	Computer Science	r.d.hoet@student.tue.nl
Venislav Varbanov	1284401	Computer Science	v.varbanov@student.tue.nl

Subject

What/How (AI) algorithms can be used by future self driving delivery cars to efficiently solve the pickup-and-delivery problem with time-windows, where a fleet of delivery vehicles must collect and deliver items according to the demand of customers and their opening hours.

Objectives

The objective of our project is to find out which (AI) algorithms can be used to optimize the planning of self-driving delivery cars and how. We will work with instances of the pickup-and-delivery problem with time-windows, where a fleet of delivery vehicles must collect and deliver items according to the demand of customers and their opening hours. The objectives are to minimize the fleet size and then to assign a sequence of customers to each vehicle of the fleet minimizing the total distance travelled. We aim to get comparable results with the Li & Lim benchmark for at least the smallest instances of the problem.

Users

The users that will benefit from those algorithms used by self driving delivery cars can be divided into:

- Primary user: Delivery companies that will use less vehicles and will use less fuel/energy and thus make more profit by delivering in a more efficient way.

- Primary user: The customer that will receive his order within the time window he has chosen.

- Primary user: Manufacturers that produce the product to be delivered.

- Secondary user: Other road users that are indirectly influenced, as the number of delivery cars will be reduced, and replaced by self-driving cars that are safer.

Approach

We will first research all papers of people that have beaten a record in the Li & Lim benchmark, then we will research other papers related to algorithms for this or similar problems. We will come up with ideas for achievable (within the length of the project) algorithms based on the read papers which we think would lead to best results. We will then implement and test these algorithms on all problem instances from the Li & Lim benchmark and compare them with the current world records. Finally, we will discuss what we have learned for the problem and the possible algorithmic approaches.

Planning

Week 1: Choosing a subject

Week 2: Planning subject, objectives, users, state-of-the-art, approach, planning, milestones, deliverables, who will do what

Week 3: Research of algorithms | Wiki: finalize subject; finalize objectives; introduction; state-of-the-art

Week 4: Research of algorithms | Implementation and testing of algorithms | Wiki: users; state-of-the-art

Week 5: Finalize research of algorithms | Implementation and testing of algorithms | Wiki: description of algorithms; finalize users; finalize state-of-the-art

Week 6: Finalize implementation and testing | Wiki: finalize description of algorithms; descriptions of results; discussion; future work

Week 7: Finalize wiki

Week 8: Finishing the final presentation and presenting

Milestones

Week 1: Subject chosen

Week 2: Initial planning finished

Week 3: Main research of algorithms done; Finalized subject and objectives on the wiki page, all other sections started

Week 4: Implementation and testing of algorithm 1

Week 5: Finalized research of algorithms; Implementation and testing of algorithm 2; Finalized users and state of the art on the wiki page

Week 6: Finalized implementation and testing (possibly new algorithm); Finalized description of algorithms and descriptions of results on the wiki page

Week 7: All sections on the wiki page finalized

Week 8: Finishing the final presentation and presenting

Deliverables

1. Implementation of one or more algorithms that work with the Li & Lim benchmark instances and produce valid solutions such that the number of vehicles is minimized and then the total distance is minimized as much as possible.

2. Wiki page that contains all the information about our project including the results of running the algorithm(s) on all instances from the Li & Lim benchmark and a comparison with the current records.

3. Final presentation that will explain our project and results.

Who will do what

Zakaria Ameziane

Cahitcan Uzman

Efe Utku

Roel den Hoet

Venislav Varbanov - Research, implementation, testing and description of algorithms

State of the Art

References (of the work of all people that have managed to beat an existing record):

BBM - Baldacci, Bartolini, and Mingozzi. An Exact Algorithm for the Pickup and Delivery Problem. Operations Research 59(2), pp. 414–426 (2011).

BVH - Bent, R. and Van Hentenryck. P. A Two-Stage Hybrid Algorithm for Pickup and Delivery Vehicle Routing Problems with Time Windows. In Principles and Practice of Constraint Programming (2003).

CLS - Curtois, T., Landa-Silva, D., Qu, Y. and Laesanklang, W., 2018. Large neighbourhood search with adaptive guided ejection search for the pickup and delivery problem with time windows. EURO Journal on Transportation and Logistics, 7(2), pp.151-192.

CVB - Christiaens J. and Vanden Berghe G. A Fresh Ruin & Recreate Implementation for Capacitated Vehicle Routing Problems. To be submitted.

CVB2 - Christiaens J. and Vanden Berghe G. Preliminary title: Slack Induction by String Removals for Vehicle Routing Problems.

DK - Dirk Koning. Using Column Generation for the Pickup and Delivery Problem with Disturbances, Technical Report, Department of Computer Science, Utrecht University, 2011.

EOE - Eirik Krogen Hagen, EOE Koordinering DA. Exploring infeasible and feasible regions of the PDPTW through penalty based tabu search. Working paper.

EMIF - Evgeny Makarov & Ilya Fiks (swatmobile.io).

H - Keld Helsgaun, Working paper, Roskilde University, Denmark (2016).

K – Richard Kelly: Hybrid Ejection Chains and Adaptive LNS for the PDPTW. Working paper.

Li & Lim - Li H. and A. Lim: A MetaHeuristic for the Pickup and Delivery Problem with Time Windows, In Proceedings of the 13th International Conference on Tools with Artificial Intelligence, Dallas, TX, USA, 2001.

MFS - Evgeny Makarov, Ilya Fiks, Eugene Sorokhtin (swatmobile.io). Unpublished.

NB1 - J. Nalepa and M. Blocho. "Enhanced Guided Ejection Search for the Pickup and Delivery Problem with Time Windows", Intelligent Information and Database Systems: Proc. 8th Asian Conference, ACIIDS 2016, pages 388–398. Springer, Heidelberg, 2016.

Q - Quintiq. http://www.quintiq.com/optimization-world-records.aspx.

R - Ropke S. Heuristic and exact algorithms for vehicle routing problems. (2005) . Ph.D. thesis, Computer Science Department, University of Copenhagen (DIKU), Copenhagen

RC - Ropke S. and J.-F. Cordeau. Branch and cut and price for the pickup and delivery problem with time windows. Transportation Sci. 43(3)267–286 (2009).

RP - S. Ropke & D. Pisinger, An Adaptive Large Neighborhood Search Heuristic for the Pickup and Delivery Problem with Time Windows, Technical Report, Department of Computer Science, University of Copenhagen, 2004.

SAM::OPT - Hasle G., O. Kloster: Industrial Vehicle Routing Problems. Chapter in Hasle G., K-A Lie, E. Quak (eds): Geometric Modelling, Numerical Simulation, and Optimization. ISBN 978-3-540-68782-5, Springer 2007.

SB - Carlo Sartori, Luciana Buriol. A matheuristic approach to the PDPTW (to be submitted).

SCR - Piotr Sielski (psielski@emapa.pl), Piotr Cybula, Marek Rogalski, Mariusz Kok, Piotr Beling, Andrzej Jaszkiewicz, Przemysław Pełka. Emapa S.A. www.emapa.pl "Development of universal methods of solving advanced VRP problems with the use of machine learning", unpublished research funded by The National Centre for Research and Development, project number: POIR.01.01.01-00-0012/19. "Optimization of advanced VRP problem variants", unpublished. Computing grant 358 funded by Poznan Supercomputing and Networking Center.

Shobb - http://shobb.narod.ru/vrppd.html

TS - TetraSoft A/S: MapBooking Algoritm for Pickup and Delivery Solutions with Time Windows and Capacity restraints.

WM - Ganzhong Luo (luoganzhong@water-mirror.com), Lei Gao (gaolei@water-mirror.com), Zhixin Liu, Yaning Li, Mingxiang Chen, Qichang Chen, Nuoyi Zhu. "New Algorithms for VRPTW & PDPTW", unpublished result of WATERMIRROR AI.

Other references: