This paper proposes a learning-based iterated local search algorithm for the asymmetrical prize collecting vehicle routing problem, which is a new variant of VRP where the objective is a linear combination of three objects: minimization of total distance, minimization of vehicles used, and maximization of customers served. Some benchmark problem instances are taken as the experiment data and the computational results show that our approach can yield about 4.05% average duality gap compared to the lower bound.
展开▼