Annealed neural networks approach to the vehicle routing problem with stochastic demands

Various vehicle routing problems (VRP) are encountered in many service systems such as delivery, customers pick-up, repair and maintenance services. Most existing VRP research has been concentrated in the case of deterministic demands. Because of practical needs, research on vehicle routing problem with stochastic demands is beginning to draw much attention at home and abroad. A modified version of mean field annealing algorithm (MFA) is derived to solve the VRP with some kind of stochastic demands, which is a combination of neural networks and simulated annealing approaches. The behavior of the algorithm is tested with numerical examples. The results show that the algorithm has good performance in both global and local search. It is believed that the concurrent and adaptive mechanisms of neural network make the algorithm so efficient.