An Evolutionary Approach for the Split Pickup and Delivery Problem

The split pickup and delivery problem is to find the shortest route that can provide delivery nodes with commodities collected from pickup nodes, subject to the vehicle load constraint. In particular, the split feature allows multiple visits to nodes and thus enables the vehicle to load or discharge an arbitrary portion of commodities along the route. This problem concerns the practical situations of limited capacity and insufficient commodities left for subsequent service. To resolve the split pickup and delivery problem, this study develops a memetic algorithm (MA) based on genetic algorithm and the modified 2-opt operator. The proposed MA employs a fixed-length representation carrying both the information of visiting order and the portion of requests satisfied; moreover, the modified 2-opt operator reduces the transportation cost without destroying the feasibility of route. Experimental results validate the optimization efficacy of the proposed MA in arrangement of visiting order and demand of each customer. In addition, the utility of split feature as well as its effect are examined in the empirical study.