Abstract: In robot path planning algorithms based on genetic algorithm，it is difficult that the granularity of grid cells is controlled according to the robot’s environment，and population initialization have some shortages.Therefore an improved path planning algorithm is proposed.In the algorithm，the size of grid cells in robot’s field is divided again according to the information of obstacles in the environment，which makes the length of chromosome in the population is more suitable to the environment，so as to improve the efficiency and performance of algorithm，at the same time，a new method based on insurance matrix to generate initial population is proposed，it can improve initial population’s ergodicity and feasibility in the search space.The simulation shows that robot can not only plan a optimal path with rapid speed but also avoid collision safety using the algorithm in the complicated environment，and the results obtained are satisfactory.

Key words: genetic algorithm, second division of environment, recessive gene, dominant gene, insurance matrix

摘要： 针对已有基于遗传算法的机器人路径规划的栅格建模方法粒度难以控制及种群初始化等方面的不足，提出了根据障碍物启发信息对环境二次划分的方法，以使得种群染色体长度具有自适应环境的特点，从而有效地提高算法的优化效率和性能，同时，提出了基于保险矩阵初始化种群新方法，可提高初始种群在搜索空间的遍历性和有效性。仿真实验结果表明：应用该算法，机器人可在具有复杂障碍物的环境中快速规划出一条全局优化路径，且能安全避障，效果显著。

关键词: 遗传算法, 环境二次划分, 隐性基因, 显性基因, 保险矩阵