Combining VNS with Genetic Algorithm to solve the one-to-one routing issue in road networks

Abstract

Computing the optimal route to go from one place to another is a highly important issue in road networks. The problem consists of finding the path that minimizes a metric such as distance, time, cost etc. to go from one node to another in a directed or undirected graph. Although standard algorithms such as Dijkstra are capable of computing shortest paths in polynomial times, they become very slow when the network becomes very large. Furthermore, traditional methods are incapable of meeting additional constraints that may arise during routing in transportation systems such as computing multi-objective routes, routing in stochastic networks. Therefore, we have thought about using meta-heuristics to solve the routing issue in road networks. Meta-heuristics are capable of copying with additional constraints and providing optimal or near optimal routes within reasonable computational times even in large-scale networks. The proposed approach is based on a hybridization process done between a Genetic Algorithm (GA) that belongs to the population-based metaheuristics and a variable neighborhood search (VNS) that performs with one single solution. To evaluate our method, we made experimentations using random generated and real road network instances. We compare our approach with two exact algorithms (Dijkstra and Integer Programming) as well as with GA and VNS when they are executed solely. Experimental results have proven the efficiency of our approach in comparison with the other approaches.