This paper presents a novel Population Control Genetic Algorithm (PC-GA) to address the Travelling Salesman Problem (TSP), incorporating techniques such as population grouping based on fitness and diversity, elite selection, and dynamic local search improvements. These strategies enhance solution quality and maintain population diversity, increasing the probability of finding the global optimum. The PC-GA is tested on benchmark TSP instances (berlin52, pr76, lin105) and compared with traditional heuristics and classic GA. Results show that PC-GA consistently outperforms other methods with minimal divergence from the optimum, though its execution time increases with problem size. This trade-off highlights the algorithm’s effectiveness in high-dimensional problems.
Mardešić et al. (Thu,) studied this question.