Vous êtes sur la page 1sur 17

Study of Vehicle Routing

Problem and Methods to Solve


By,
Abhinav Sharma,
154270011

Introduction
The Vehicle Routing Problem (VRP) is defined as the optimal design of routes from depot to
customers so that constraints are satisfied and objective function is achieved.
A typical VRP has a set of customer, set of vehicle and depot. The road network used for VRP is
described through a graph, whose vertices represents customers/cities and edge represents road.

Every edge is associated with a cost.


In CVRP following assumption are made,
Customer demands are deterministic and known in advance.

The vehicles are identical and based at a single central depot.


Capacity restrictions for the vehicle are imposed

Introduction (contd.)

Introduction (contd.)
Given a graph with a finite set of nodes (V), a set of arcs (A), a nonnegative cost cij associated with
each arc i, j A, find a collection of S routes, where S V with minimum cost, such that:
Each route visit the depot vertex, V0 .
Each node is visited exactly once by a route.
The sum of demands of nodes visited by a route does not exceed vehicle capacity Q
A bound for minimum number of vehicles is given by

d(S)
r(S)
C

(1)

Vehicle Flow Model


Subjected to

Vehicle Flow Model (Contd.)

Directed Formulation

Subjected to

Vehicle Flow Model (Contd.)

One Index formulation


Subjected to

Vehicle Flow Model (Contd.)


Set Partitioning Formulation

Set Partitioning Formulation (Contd.)


Subjected to

Methods to Solve
1. Clarke and Wrights Savings Algorithm

Methods to Solve (Contd.)


Step 1. Calculate the savings for all pair of
customers and then sort the routes in
descending order of savings.
Step 2. Select a pair of node and assign it to
a route. Now select next pair of node and
check if that node can be added to existing
route by taking permutation of nodes and
without disturbing neighbourhood of
previous pair of node and at the same time
satisfying capacity constraint. If criteria is
not satisfied, move to next node.
Step 3. Repeat step 2 for remaining pair of
nodes.

Methods to Solve (Contd.)


2. Sweep Algorithm
Consists of two parts:
1.

Clustering

2.

Route Generation

Methods to Solve (Contd.)


n

Clarke and Wright

Sweep

Best known solution

50

578.56

531.90

524.61

75

888.04

884.20

835.26

100

878.04

846.34

826.14

150

1128.24

1075.38

1028.42

199

1386.84

1396.05

1291.45

Observations
As the number of nodes increases, number of branches in branch and bound
increases exponentially with n, where n is number of nodes. This increases
computational time and the solver seems to stagnate at some point.
Problem can be overcome by employing random number search to eliminate
some branches and further solving remaining branches.
Heuristic solutions doesnt guarantee optimal solution but only gives near
optimal solution. To have the best solution, do a local search to get best solution
after several runs of heuristics. Another method is exploring a large
neighbourhood of the current solution by shuffling the nodes in route.
Heuristics seems to be best option for solving large instance of problems but are
complex in computation.

Observations (Contd.)
Sweep algorithm is proven to minimize the distance of existing route but does not
generate new route.
New hybrid algorithm can be developed by clubbing iterative search algorithm
and meta-heuristic. As meta-heuristic algorithm requires an initial seed or guess,
it can be provided by solving given problem with iterative search algorithm and
then selecting a best solution after limited number of iterations and then feed to
meta-heuristic algorithm.
Formulations should have less number of variables and constraint to get tight
formulation and strong LP relaxation.
VRP is NP-hard.
VRP with customers with no demand is TSP.
VRP with one depot and K number of vehicles and no customer demand is one
origin, multiple travelling salesman problem.

References
Clarke, G. U., and John W. Wright. "Scheduling of vehicles from a central depot to a number of delivery
points." Operations research 12.4 (1964): 568-581.
Fisher, Marshall L., and Ramchandran Jaikumar. "A generalized assignment heuristic for vehicle routing."
Networks 11.2 (1981): 109-124.
Gendreau, Michel, Gilbert Laporte, and Ren Sguin. "Stochastic vehicle routing." European Journal of
Operational Research 88.1 (1996): 3-12.
Kumar, Suresh Nanda, and Ramasamy Panneerselvam. "A survey on the vehicle routing problem and its
variants." Intelligent Information Management 4.3 (2012): 66.
Laporte, Gilbert, et al. "Classical and modern heuristics for the vehicle routing problem." International
transactions in operational research 7.45 (2000): 285-300.
Laporte, Gilbert. "The vehicle routing problem: An overview of exact and approximate algorithms." European
Journal of Operational Research 59.3 (1992): 345-358.
Shin, Kwangcheol, and Sangyong Han. "A centroid-based heuristic algorithm for the capacitated vehicle
routing problem." Computing and Informatics 30.4 (2012): 721-732.
Toth, Paolo, and Daniele Vigo, eds. Vehicle routing: problems, methods, and applications. Vol. 18. Siam, 2014.

Vous aimerez peut-être aussi