A comparative study on pheromone updating strategies of MP2AS for VRP

This paper proposes an improved ant system, which is of multiple evolution phases and multiple transition preferences (MP²AS). Considering the critical role of pheromone updating, we compare five existing strategies: ant-quantity, ant-density, ant-cycle, ant colony system (ACS) and ranked ant system (ASrank). The algorithms are applied to conquer four VRP benchmark problems of different sizes and performances are evaluated. Numerical results validate that ASrank statistically find better solutions with less computation time and lower deviation.