Scatter Search Particle Filter to Solve the Dynamic Travelling Salesman Problem

This paper presents the Scatter Search Particle Filter (SSPF) algorithm and its application to the Dynamic Travelling Salesman Problem (DTSP). SSPF combines sequential estimation and combinatorial optimization methods to improve the execution time in dynamic optimization problems. It allows obtaining new high quality solutions in subsequent iterations using solutions found in previous time steps. The hybrid SSPF approach increases the performance of general Scatter Search (SS) metaheuristic in dynamic optimization problems. We have applied the SSPF algorithm to different DTSP instances. Experimental results have shown that SSPF performance is significantly better than classical DTSP approaches, where new solutions of derived problems are obtained without taking advantage of previous solutions corresponding to similar problems. Our proposal reduces execution time appreciably without affecting the quality of the estimated solution.