An Enhanced Ant Colony Algorithm with Variable Neighborhood Descent for Multi-compartment Vehicle Routing Problem with Time Limits

The multi-compartment vehicle routing problem (MCVRP) is applicable in many areas, such as fuel transportation, garbage collection and food distribution. Ant colony (ACO) algorithm has been recognized as a prominent evolutionary methodology due to its excellent performance of global exploration. In this paper, an enhanced ant colony algorithm, namely EACO, is presented to minimize the total distance of the MCVRP with time limits (TLs), which is a typical NP-hard combinatorial optimization problem with strong practical application background. Firstly, a new pheromone concentration matrix (NPCM) is developed to learn both the information of two successive customers and the customer sequence. Secondly, to improve the quality of initial solutions, three heuristic rules are adopted to initialize the NPCM and the population. Thirdly, after the ACO-based exploration, a problem-dependent local search with the technique of variable neighborhood descent is presented to perform exploitation. Simulation results on benchmark data sets demonstrate the effectiveness of the proposed EACO.