Improving Ant Colony Optimization For The Multi-objective Multiple Travelling Salesman Problem In The E-commerce Delivery Scene

With the rise of online shopping,the e-commerce logistics industry flourishes and gradually penetrates into people's life.The "Last Mile" problem of express delivery is becoming increasingly prominent and has become the bottleneck of the efficiency of e-commerce logistics distribution.The key to this problem is how to reasonably deploy courier personnel,so that all the express delivery routes are as short as possible and roughly equivalent.This paper studies the above phenomenon,abstracts it into a multi-objective multiple traveling salesman problem,and defines a new evenness goal in combination with the actual demand.Its characteristic is that the loss caused by small deviations of travelers' path lengths will be tolerated,while the loss caused by large deviations will be amplified.Obviously,the revised model is more consistent with the needs of the actual scene and has a stronger practical significance.The multi-objective multiple traveling salesman problem has complex constraints and a wide range of feasible areas.Therefore,while the problem has long been concerned,there are few effective solutions.At present,the commonly used solutions are mainly improved by genetic algorithms.They do have strong global search capabilities,but the slow convergence rates and unclear search directions makes they useless when in practical.The ant colony algorithm uses the pheromone positive feedback mechanism to guide the algorithm to converge to the optimal solution with heuristic information.It is an iterative search algorithm with fast convergence and good effect.However,the classical ant colony algorithm is not suitable for solving multi-objective multiple traveling salesman problems due to lack of certain heuristic information and restrictions of tabu list.Based on such situation,this paper introduces the multi-traveler tabu list to the ant colony algorithm framework,make it efficient to solve the multiple traveling salesman problem.On this basis,this paper proposes the strategies of random initialization,state transfer rebound,solution recombination and multiple pheromone update,which effectively improves the performance of ant colony algorithm on the uniformity target,and makes the ant colony algorithm better take into account the targets of uniformity and total length.In addition,the above mentioned strategies are combined by multi ant colony hybrid search,which greatly improves the global search ability of ant colony algorithm.Finally,we design a number of test examples and evaluate the above-mentioned improved ant colony algorithm on the distribution situation of express delivery in Guangzhou Higher Education Mega Center.Compared with the related research results of genetic algorithms and differential evolution algorithms in recent years,the experimental results show that the improved ant colony algorithm has obvious advantages in efficiency,and is superior in multiple evaluation indicators of multi-objective optimization.