Evolutionary algorithms for the pursuit problem

In this paper we propose new variants of the Evolutionary Algorithms, called Multi-Virus Evolutionary Algorithm (MVEA) and Multi-Virus Evolutionary Annealing Algorithm (MVEA2) for the following pursuit problem. Given a set of mobile agents and a set of mobile targets in a map as an input instance, the goal of the agents is to pursue and capture as many mobile targets by cooperating with other agents as possible, and furthermore as quickly as possible. When we use the multipoint search algorithms such as Genetic Algorithms (GAs) and Virus Evolutionary Algorithms (VEA) for optimization problems, it is important to keep the diversity of the search points. To improve the diversity, MVEA selects several viruses of inferior individuals with a certain probability, and moreover, MVEA2 blends the simulated annealing heuristic (SA) with MVEA. In this paper, we show the detailed experimental comparisons among the performances of a traditional GA, GA with SA, MVEA, and MVEA2 on the number of captured mobile targets and the number of the required steps for agents to capture all the mobile targets in the pursuit problem. The comparisons show that (i) MVEA and MVEA2 behave in similar ways as GA and GA with SA, respectively, and (ii) GA with SA and MVEA2 does not work better in the early stages than GA and MVEA, but especially in the final stages, the former algorithms can capture the larger number of mobile targets than one of the latter ones.