A hybrid algorithm of local search for the heterogeneous fixed fleet vehicle routing problem

Under consideration is the optimization problem for the routes of heterogeneous vehicles serving a given set of customers. It is assumed that the customers are represented by points in the plane, whereas the number of each type of vehicles is limited. To solve the problem, we developed a hybrid local search algorithm with coding the solutions as a sequence of customers. To decode the sequence, the corresponding NP-hard problem is solved by the method of Lagrangian relaxation. We propose new procedures for intensification and diversification of the search, as well as a new neighborhood of exponential size. The results of computational experiments are presented for the available test instances with the number of customers up to 255. For 15 instances we obtain new record values of the objective function.