基于改进遗传算法的智能体路径规划仿真

Path Planning for Agent Based on Improved Genetic Algorithm

将改进自适应遗传算法应用于自治智能体动态路径规划,选取一维路径编码,并利用领域知识和局部避障技术生成初始种群,设计了交叉、变异和平滑算子,提出了新的交叉概率和变异概率调节公式。上述调节公式不仅考虑了种群中个体适应度的区别,而且还从整体上考虑了种群多样性和收敛性等性能指标,克服了传统遗传算法和一般自适应遗传算法的早熟收敛问题,提高了进化效率。仿真结果表明,改进方法在收敛速度和输出全局最优解的概率相对于标准遗传算法和一般自适应遗传算法都有较明显的提高。

The paper applied an improved an adaptive genetic algorithm for autonomous agent dynamic path planning. We selected one -dimensional path encoding, used domain knowledge and local obstacle avoidance technology to generate initial population, designed crossover, mutation and smooth operator, and put forward a new crossover probability and mutation probability adjustment formula. The adjustment formula considerd not only the difference of individual fitness value, but also the population diversity and convergence of indicators as a whole, which overcame the premature convergence problem in standard genetic algorithm and adaptive genetic algorithm and improved the evolutionary efficiency. The simulation experimental results show that the method in convergence speed and the probability of global optimal solution compared with standard genetic algorithm and adaptive genetic algorithm have been obviously improved.