Path Planning Based on Two-layer Adaptive Genetic Algorithm

Path planning for autonomous agent is an important issue in artificial intelligence, its purpose is to find a reasonable path by following certain optimization criteria, such as the length of path is shorter, the path can avoid collision, and the path is smooth, etc. This paper proposed a two-layer adaptive genetic algorithm (TAGA) and applied to path planning. On the one hand, we use one of genetic algorithm to find the optimal path (called as pathfinding GA), and adopt fuzzy logic to adjust crossover probability and mutation probability. On the other hand, we use another genetic algorithm to optimize the fuzzy reasoning rules and type of membership functions (called as self-learning GA). These two GA work cooperatively, self-learning GA provide the optimal individual for pathfinding GA, that means the optimal fuzzy reasoning rules and type of membership functions, at the same time, the selection of the optimal individual in self-learning GA need recur to pathfinding GA. The proposed TAGA method shows efficiency in path planning, and we demonstrate this point by applying it to the static and dynamic environments. Experimental results show that the proposed TAGA method overcomes premature convergence of the standard genetic algorithm (SGA), speed up convergence, and enhanced the application scope of the adaptive genetic algorithm (AGA).