Abstract: Considering the congested roads situation of urban central areas in China, a speed calculation method based on vehicle driving sections is designed, and a model for the Green Multi-Depot Vehicle Routing Problem with Total Fuel-Consumption cost Criterion (TFC-GMDVRP) is established, considering simultaneously the vehicle travel distance, load, and speed factors. A learning ant colony optimization algorithm (LACO), combining a knowledge model and an ant colony optimization algorithm (ACO), is proposed for solving the TFC-GMDVRP. In order to improve the performance and robustness of the algorithm's global search, the parameter knowledge that contains the different parameter combinations of ACO and the selection probability of each parameter combination is designed to adjust the ACO's parameters for each generation. In order to enhance the ability of algorithm's local search, the local operation knowledge that contains the contribution ratio of each neighborhood operation is designed to determine the execution times of each neighborhood operation for each generation. Simulation experiments on different instances and comparisons of algorithms show the effectiveness of the proposed algorithm.

Key words: learning ant colony optimization, green multi-depot, vehicle routing problem, knowledge model, neighborhood search