A Robust Method for the VRPTW with Multi-Start Simulated Annealing and Statistical Analysis

Vehicle routing problems have been extensively analyzed to reduce transportation costs. More particularly, the vehicle routing problem with time windows (VRPTW) imposes the period of time of customer availability as a constraint, a very common characteristic in real world situations. Using minimization of the total distance as the main objective to be fulfilled, this work implements an efficient hybrid system which associates non-monotonic simulated annealing to hill climbing with random restart (multi-start). Firstly, the algorithm is compared to the best results published in the literature for the 56 Solomon instances. Then, it is shown how statistical methods - analysis of variance and linear regression - can be used to determine the significance degree of the system´s parameters to obtain an even better and more reliable performance