A Preliminary Study of Evolutionary Multitasking for Multiobjective Vehicle Routing Problem With Time Windows

Currently, most researches on multiobjective optimization (MOO) focus on solving only a single problem from the scratch during the search process, without the effective use of valuable knowledge from other related problems. It may lead to inefficient and repeated searches on similar problems. In recent years, with the rapid development of the cloud computing industry, evolutionary multitasking optimization (EMO) has attracted lots of attention and has shown its superior performance in solving multiple related problems simultaneously. Based on these considerations, we propose a novel EMO algorithm to solve multiple multiobjective vehicle routing problems with time windows (MOVRPTW) concurrently by combining the EMO framework with an off-the-shelf multiobjective optimization algorithm. The proposed algorithm is termed MMOVRPTW. In the proposed algorithm, a cooperation mechanism is designed to adaptively switch the search between the knowledge transfer process and the local search process. To evaluate the efficacy of the proposed algorithm, the preliminary study on the multitasking MOVRPTW benchmark problems by pairing the 45 real-world instances is carried out to compare MMOVRPTW with its single-task counterpart. The experimental results have demonstrated the advantages of the proposed algorithm for solving multiple MOVRPTW simultaneously.