Sei sulla pagina 1di 19

Sirindhorn International Institute of Technology

VEHICLE ROUTING
PROBLEM USING AI
SYED MUHAMMAD TARIQ SHAH (6222300060)

ADVISOR: ASSOC. PROF. MONGKUT PIANTANAKULCHAI

Research Progress
1st Semester, PhD Student, School of Civil Engineering and Technology
Sirindhorn International Institute of Technology, Thammasat University

1
Introduction

Different Variants of
VRP

Algorithms Used in
VRP

Artificial
Intelligence in VRP

Conclusion and
Future Plan

2
INTRODUCTION
• The process of selecting “best” routes in a network of links and nodes.

• Under study since 1959.

• Initially solved with linear Programing. Later on extended to Integer Programing.

3
TRADITIONAL VRP VARIANTS

Capacitated VRP
(Since 1959)

• First introduced by Dantzig


and Ramser (1959).

• Euclidean distance were


considered.

• Integer programming was


used to solve.

• Branch and Bound


algorithm was used.

4
TRADITIONAL VRP VARIANTS
Time-dependent VRP
(since 1966)
• Actual conditions such as urban
congestion were considered.

• Mixed integer linear programming


was used to solve. Cooke and
Halsey (1966).

• Extended to TDVRP with Time


Windows.

• Iterative Route Construction and


Improvement algorithm (IRCI)
Kritzinger et al. (2012).

5
TRADITIONAL VRP VARIANTS

Pickup and Delivery Problem


(since 1967)
Similar to

• VRP with backhauls.

• VRP with dial-a-ride


problem, etc.

• Integer programming was


used to solve.

• Metaheuristic tabu search


was also used. Glover and
Laguna (1997)

• Ant colony optimization.


Wade and Salhi (2003)

6
TRADITIONAL VRP VARIANTS
Multi-depot VRP
(since 1969)

• Contains more than one depot

• Each customer is visited by a


vehicle

• Extended to MDVRPTW

• Integer programming was used


to solve.

• Genetic Algorithm. Mibrabi et al


(2016)

7
TRADITIONAL VRP VARIANTS
Dynamic VRP
Stochastic VRP (since 1976)
(since 1969)
• May be • Consider real time dynamic
• Stochastic customer environment.
• Stochastic demands
• Stochastic times • Ant Colony Optimization.
Yu et al. (2009)
• Tabu Search was used to
solve. Tas et al. (2012) • Genetic algorithm. Hanshar
and Ombuki-Berman (2007)
• Extended to SVRPTW.
• Particle swarm optimization.
Okulewicz and Mańdziuk
(2013)

8
TRADITIONAL VRP VARIANTS
Periodic VRP
(since 1974)

• Used in municipal waste collection,


industrial gas distribution, grocery
industry, picking up raw materials from
suppliers etc.

• Feasible routing over the time horizon


(week) is minimized.

• Extended Multi-depot PVRP

• PVRP with Time Windows

• Improved ant colony optimization


(IACO)

9
TRADITIONAL VRP VARIANTS
Fleet Size and Mix Vehicle Routing Problem
(since 1984)

• Consider the most economic


combination of vehicles.

• Consider heterogeneous vehicles with


different capacities and traveling cost.

• Extended to FSMVRP with Time


Windows. Bräysy et al. (2008)

• Savings heuristic

• Mixed Integer Programming

10
TRADITIONAL VRP VARIANTS
Generalized VRP
(since 1984)

• Customers are partitioned into clusters and


vehicles are obligated to visit only one customer
in each cluster (i.e. each cluster should be visited
exactly once).

• Integer programming was used to solve this


problem. Pop et al. (2011)

• Local search (LS) Pop et al. (2012)

11
TRADITIONAL VRP VARIANTS
Split-delivery VRP Green Vehicle Routing Problem
(since 1989) (since 2006)
• Deals with the optimization of energy
• When customer demand exceeds the vehicle consumption of transportation.
capacity. Customer is serviced by more than one
vehicle. • Mixed integer linear program. Erdogan & Miller
(2011)
• Tabu Search Algorithm was used to solve it.
Archetti et al. (2006)

12
ALGORITHMS USED

Exact Solution
(1) Direct tree search methods.
(2) Dynamic programming. Tackle problems of a relatively small scale.
(3) Integer linear programming.

Approximate Solution
Heuristics Metaheuristics
(1)Saving algorithms (1)Local search. The main local search based
(2)Sequential improvement algorithms metaheuristics for VRP include:
(3)Sweep algorithms (1)Tabu Search (TS)
(4)Petal algorithms (2)Simulated Annealing (SA)
(5)Fisher and Jaikumar two-phase (3)Greedy Randomized Adaptive Search
algorithms Procedure (GRASP)
(4)Variable Neighborhood Search (VNS)
Solve large problems with (5)Large Neighborhood Search (LNS)
(2) Population search
acceptable computational time.
(1)Genetic Algorithms (GA)
(2)Ant Colony Optimization (ACO).

13
AI IN VRP
Reinforcement learning
(since 2018)

• Nazari et al. (2018) used Recurrent Neural Network (RNN) coupled with attention
mechanism (to choose feasible destinations).

• Markov Decision Process (MDP) was used, in which the optimal solution was treated as a
decision.

• Better then classical heuristics and Google OR-Tools.

14
AI IN VRP
Reinforcement learning
(since 2018)

• TraCI tool was used for performing Reinforcement Learning(RL) in the SUMO simulator.

• RL optimizes its policy by repeating the simulation and observing the reward.

• Vehicles get smarter in selecting the route with less congestion.


Sang et al. (2018)

15
Conclusion and Future Plan

• It can be concluded that various variants of VRP cover a specific problem and ignore most
of the associated actual road characteristics.

• Dynamic VRP can be considered as most efficient in terms of considering the various road
characteristics. But its application to the other variants of VRP is very difficult to apply.
(Need more literature review.)

• Reinforcement Learning can be a substitute of the conventional methods. Nazari et al.


(2018)

• The same concept as used by Sang et al. (2018) can be extended to any real world problem
and can be applied to any of the variants of the VRP.

• In future more literature review related to dynamic VRP and use of artificial intelligence in
VRP is required.

• Machine Learning using Python is also required. (I have already started related online
course.)

16
Conclusion and Future Plan

• I have already stated working on solution of basic VRP variants using Python or machine
learning.

• Data related to VRP is collected in simplified form. But for considering dynamic VRP more
data is required.

• In future I will try to apply use RL to solve some basic variants of VRP and extend it to
dynamic VRP.

• The results will be compared to the other most efficient methods of solving VRP.

17
References
• G. B. DANTZIG & J. H. RAMSER. "The Truck Dispatching Problem." (1958).
• Cooke, Kenneth L., and Eric Halsey. "The Shortest Route through a Network with Time-Dependent Internodal Transit Times." Journal of
Mathematical Analysis and Applications 14, no. 3 (1966/06/01/ 1966): 493-98. https://doi.org/https://doi.org/10.1016/0022-
247X(66)90009-6. http://www.sciencedirect.com/science/article/pii/0022247X66900096.
• Figliozzi, Miguel. "The Time Dependent Vehicle Routing Problem with Time Windows: Benchmark Problems, an Efficient Solution
Algorithm, and Solution Characteristics." Transportation Research Part E: Logistics and Transportation Review 48 (05/01 2012): 616–
36. https://doi.org/10.1016/j.tre.2011.11.006.
• Glover, Fred, and Manuel Laguna. Tabu Search I. Vol. 1, 1999. doi:10.1287/ijoc.1.3.190.
• Wade, Anne, and Said Salhi. "An Ant System Algorithm for the Mixed Vehicle Routing Problem with Backhauls." In Metaheuristics:
Computer Decision-Making, edited by Mauricio G. C. Resende and Jorge Pinho de Sousa, 699-719. Boston, MA: Springer US, 2004.
• Piroozfard, Hamed, Kuan Yew Wong, and Adnan Hassan. "A Hybrid Genetic Algorithm with a Knowledge-Based Operator for Solving
the Job Shop Scheduling Problems." Journal of Optimization 2016 (2016): 13 7319036. https://doi.org/10.1155/2016/7319036.
http://dx.doi.org/10.1155/2016/7319036.
• Kaviani, Mohamad, Mehdi Abbasi, Bentolhoda Rahpeyma, and Mohamad Yusefi. "A Hybrid Tabu Search-Simulated Annealing Method
to Solve Quadratic Assignment Problem." Decision Science Letters 3 (05/09 2014): 391–96. https://doi.org/10.5267/j.dsl.2014.2.004.
• Yu, Hualong, Guochang Gu, Haibo Liu, Jing Shen, and Jing Zhao. "A Modified Ant Colony Optimization Algorithm for Tumor Marker
Gene Selection." [In eng]. Genomics, proteomics & bioinformatics 7, no. 4 (2009): 200-08. https://doi.org/10.1016/S1672-
0229(08)60050-9. https://www.ncbi.nlm.nih.gov/pubmed/20172493
• Hanshar, Franklin, and Beatrice Ombuki-Berman. "Dynamic Vehicle Routing Using Genetic Algorithms." Appl. Intell. 27 (06/18 2007):
89-99. https://doi.org/10.1007/s10489-006-0033-z.
• Okulewicz, Michał, and Jacek Mańdziuk. "Application of Particle Swarm Optimization Algorithm to Dynamic Vehicle Routing Problem."
Paper presented at the Artificial Intelligence and Soft Computing, Berlin, Heidelberg, 2013// 2013.
• Bräysy, Olli, Wout Dullaert, Geir Hasle, David Mester, and Michel Gendreau. "An Effective Multirestart Deterministic Annealing
Metaheuristic for the Fleet Size and Mix Vehicle-Routing Problem with Time Windows." Transportation Science 42 (08/01 2008): 371-
86. https://doi.org/10.1287/trsc.1070.0217.
• Horvat-Marc, PetricA C. Pop and Andrei. "Local Search Heuristics for the Generalized Vehicle Routing Problem." (2012).
• Archetti, C., M. Grazia Speranza, and Alain Hertz. "A Tabu Search Algorithm for the Split Delivery Vehicle Routing Problem."
Transportation Science 40 (11/30 2006): 64-73. https://doi.org/10.1287/trsc.1040.0103.

18
References
• Erdoğan, Sevgi, and Elise Miller-Hooks. "A Green Vehicle Routing Problem." Transportation Research Part E: Logistics and
Transportation Review 109 (01/31 2012): 100–14. https://doi.org/10.1016/j.tre.2011.08.001.
• Mohammadreza Nazari, Afshin Oroojlooy, Lawrence V. Snyder, Martin Takáč. "Reinforcement Learning for Solving the Vehicle Routing
Problem." (2018).
• Koh, S. S., B. Zhou, P. Yang, Z. Yang, H. Fang, and J. Feng. "Reinforcement Learning for Vehicle Route Optimization in Sumo." Paper
presented at the 2018 IEEE 20th International Conference on High Performance Computing and Communications; IEEE 16th
International Conference on Smart City; IEEE 4th International Conference on Data Science and Systems (HPCC/SmartCity/DSS), 28-30
June 2018 2018.

19

Potrebbero piacerti anche