Matrix-based Genetic Algorithm for Mobile Robot Global Path Planning

In this paper, a new variant of genetic algorithm, matrix-based genetic algorithm (MGA), is proposed, which represents the population of genetic algorithm by matrix, and achieves evolution by matrix operation. Applying it to the 2D global path planning problem of robot, MGA has better accuracy and running speed than the basic genetic algorithm. Specifically, the average path length obtained after convergence of MGA is closer to the actual shortest path length than that of the basic genetic algorithm, and, due to the parallelism of the MGA in matrix operations, the MGA runs in 1/2 the time of the traditional genetic algorithm when using the NumPy library. Two experiments comparing shortest path lengths demonstrate that MGA has better stability and can cope with complex environments. In order to enable the algorithm to have better accuracy, this paper also explores the effect of whether to use elite strategy on the accuracy of the shortest path of the algorithm. In addition, the effect of different pathfinding algorithms on the computing time of the algorithm is explored to speed up the operation of the algorithm.