An adaptive multi-crossover population algorithm for solving routing problems

Throughout the history, Genetic Algorithms (GA) have been widely applied to a broad range of combinatorial optimization problems. Its easy applicability to areas such as transport or industry has been one of the reasons for its great success. In this paper, we propose a new Adaptive Multi-Crossover Population Algorithm (AMCPA). This new technique changes the philosophy of the basic genetic algorithms, giving priority to the mutation phase and providing dynamism to the crossover probability. To prevent the premature convergence, in the proposed AMCPA, the crossover probability begins with a low value, and varies depending on two factors: the algorithm performance on recent generations and the current generation number. Apart from this, as another mechanism to avoid premature convergence, our AMCPA has different crossover functions, which are used alternatively. We test the quality of our new technique applying it to three routing problems: the Traveling Salesman Problem (TSP), the Capacitated Vehicle Routing Problem (CVRP) and the Vehicle Routing Problem with Backhauls (VRPB). We compare the results with the ones obtained by a basic GA to conclude that our new proposal outperforms it.