The Improvement of Simulated Annealing Algorithm on the Penalty Function in Multi-agent Traveling Salesman Problem

Abstract

To solve the MTSP and discover the efficiency of different methods, this paper compares Monte Carlo with Simulating Annealing Algorithm by testing them in three sizes of map samples from small to large. The Monte Carlo can just solve the simple TSP problems and become useless due to its time and space complexity when the number of cities goes large. Therefore, we adopt a heuristic algorithm, Simulated Annealing. The Simulated Annealing Algorithm can solve the MTSP, although it is not the most optimal path, especially when an agent faces a group of clustered cities. To get a better result, this paper describes how to alternate the penalty function to set some limitations in the SAA and provide a better way to solve the MTSP when facing clustered group locations, which helps optimize the path in practice