A New ACO Approach for Solving Dynamic Problems

International audienceMany real-world optimization problems evolve in a dynamically changing environment: the problem to solve changes through time, and no information about changes and problem evolution is available ahead of time. When solving such problems, we need an optimization technique which is able not only to converge around optimal solutions but also to follow their shifting through time as closely as possible. In this paper, we propose a new ACO (Ant Colony Optimization) approach for solving dynamic problems. Like in PACO, every time the problem changes, instead of restarting from scratch, pheromone trails are re-computed at each time w.r.t a set of already found solutions. However, in order to avoid the algorithm to be misguided, the new pheromone trails are first tested. The goal is to study the relevancy of the re-computed pheromone trails w.r.t the changes and consequently to adjust/correct their values. The proposed algorithm is tested on Dynamic TSP (Traveling Salesman Problem) instances