Hybrid Adaptive Memory Programming to optimize the multi-commodity many to many vehicle routing problem

Abstract

With the quick development of urban transport networks, the multi-commodity many to many variants of pickup and delivery vehicle routing problem (PDVRP) becomes more and more important. A critical issue is to solve this variant through optimisation techniques. We address a new variant of the multi-commodity many to many PDVRP (m-MMPDVRP). The m-MMPDVRP problem is when one or multi-commodities are collected from many sites to be transported to many destinations. In this problem, we assumed that all commodities share the same vehicle capacity during transportation. All vehicles are non-homogeneous and each commodity has to be stored separately during transportation. A new model is developed, based on multiple commodities. The objective is to generate an optimal path plan, ensuring that the demand for heterogeneous commodities can be satisfied by an arbitrary set of suppliers. We propose an adaptive memory-programming (AMP) technique based on the Scatter Search (SS). The solution quality of the suggested methodology is assessed and compared with the result presented in the previous works for the same instances. Numerical experimentation shows the distinction of the AMP with Scatter Search compared with other existing techniques; and establishing an efficient metaheuristic method for the m-MMPDVRP problem.