Adaptive genetic algorithm for occupancy grid maps merging

Abstract

Multi-robot system can improve the efficiency of mapping and exploration. One of the key problems is when and how to merge the partial maps acquired by robots independently to share environmental information between robots. This paper studies the problem of fusing two partial maps without common reference frames and relative position information of robots. On the basis of the similarity metric, the paper applies an adaptive genetic algorithm for finding the overlapping region between the partial occupancy grid maps to realize map merging. The algorithm adjusts the crossover and mutation probability adaptively and nonlinearly with the similarity metric to avoid such disadvantages as premature convergence, low convergence speed and low stability. The experiment results show that the genetic algorithm based map merging does not get stuck at a local optimum, and is robust and can provide fast convergence for the optimal overlapping partial maps.