This paper presents a study on an adaptive version of particle swarm optimization (PSO) algorithmfor solving vehicle routing problems (VRPs). Recently, PSO has been showing promising results in solvingmany optimization problems include VRP. There are some parameters that need to be set in order to obtain agood performance of the PSO algorithm. However, finding the best set of parameters that is good for allproblem cases is not an easy task. Many experiments must be performed to set the parameters and yet there isno guarantee that the best obtained parameter set will provide consistently good algorithm performance whenit is applied to a new problem cases. Hence, a novel idea to have a self-adaptive PSO, that can adapt itsparameters automatically whenever it is applied to solve a problem instance, is an alternative way toovercome this situation. The adaptive version of PSO proposed in this paper has additional capability to selfadaptits inertia weight (w), one of the key PSO parameter, based on the velocity index of the swarm, thesearching agents in PSO. The inertia weight is controlled so that the balance between exploration andexploitation phases of the swarm is maintained, since a better balance of these phases is often mentioned asthe key to a good performance of PSO. The performance of this adaptive PSO is evaluated for solving VRPinstances and is compared with the existing application of PSO for VRP. The computational experiment showsthat the adaptive version of PSO is able to provide better solution than the existing non-adaptive PSO withslightly slower computational time.
展开▼
