基于多目标遗传算法的机器人避障研究

提出了在势场法斥力场函数中引入目标点和障碍物的距离、障碍物和机器人的速度矢量及最小安全距离的一种新的动态势场(NDAPF)方法，建立了考虑运动路径长度、移动步数、动态相对距离的避障评价函数，进一步提出了根据遗传算法个体适应度不同情况下变异概率自适应调整的一个改进算法，利用多目标遗传算法(MOGA)对NDAPF。方法进行寻优．在一个移动机器人仿真平台中分别对传统的静态势场(TAPF)方法、无参数优化NDAPF和利用MOGA参数优化的NDAPF方法(MOGANDAPF)进行仿真实验，实验结果证明，MOGANDAPF方法可进一步应用于足球机器人避碰模型研究中．

Motion planning of robot collision avoidance based on multi-objective genetic algorithm

A new dynamic artificial potential field (NDAPF) is proposed to solve motion planning problem, in which distance between target and obstacle, velocity of robot and obstacle and the minimum safe distance are introduced to the repulsion function. Following the proposition of the NDAPF, an evaluating function of collision avoidance is presented within which the motion path lengths, motion steps and dynamic relative distance are considered. Furthermore, a modified self-adaptive mutation operator is put forward for the Multi-objective Genetic Algorithm (MOGA), which is used to search for the optimal parameters of NDAPF function. Simulation results show that MOGANDAPT is feasible and can be used in robot collision avoidance.