Learning Ant Colony Algorithm for Solving Green Multi-Compartment Vehicle Routing Problem with Simultaneous Pickup and Delivery

To solve the Green Multi-Compartment Vehicle Routing Problem with Simultaneous Pickup and Delivery (GMCVRPSPD), a mathematical model is established with the objective of minimizing total costs. The model consists of vehicle operating costs and fuel consumption costs. A Learning Ant Colony Optimization algorithm (LACO) is proposed to solve GMCVRPSPD. Firstly, in the global search stage of LACO, three different heuristic methods are used to generate the initial solution. These solutions are used to initialize the pheromone matrix, providing a better starting point for the algorithm's subsequent search process. A pheromone update mechanism is designed to effectively enhance the convergence speed of the algorithm. Secondly, in the local search of LACO: a local learning strategy is designed, which selects the neighborhood operation suitable for the current search stage from the six neighborhood operations designed, and dynamically updates the selection probability value according to the contribution value generated by the neighborhood operation. In addition, a rapid evaluation strategy is designed to reduce the calculation time in the local search process. Finally, the effectiveness of LACO in solving GMCVRPSPD problem is verified by simulation experiment and algorithm comparison.