基于多种群遗传算法的一般机器人逆运动学求解

Solution of Inverse Kinematics for General Robot Manipulators Based on Multiple Population Genetic Algorithm

几何结构不满足Pieper准则的机器人被称为一般机器人,其逆运动学运算不能采用封闭解法,而采用数值解法又需要庞大的计算量,且存在奇异位置无法求解的问题。为此,将多种群遗传算法应用于运动学逆解运算,提出一种适用于一般机器人的高精度并行求逆算法。为避免机器人位姿收敛精度不同,该算法将目标函数分解为位置和姿态函数,同时引入适应度函数权值系数来平衡两函数收敛速度;为避免局部收敛,该算法采用多点交叉和均匀交叉相结合的交叉算子,并逐步增大均匀交叉概率来抑制短子串偏差,使搜索趋于稳健;为提高收敛速度,该算法采用动态变异率的变异算子,以及种群替代和个体替代相结合的移民算子来克服全局收敛的盲目性。以封闭解法和数值解法无法求逆的6R一般机械臂为对象,开展与单种群遗传算法的对比试验,结果表明:该算法可在避免局部收敛的基础上保证算法稳定性,且能够大幅提升收敛精度和速度。

If one robot does not meet the Pieper criterion,it is then called a general robot.In such case,the closed-form methods cannot be applied to solve the inverse kinematics problem,while the numerical methods may cause considerable computational load.To solve these issues,a multiple population genetic algorithm based inverse kinematics method is proposed.To achieve the same convergence accuracy between the position and the posture,the proposed method decomposes the objective function into position function and posture function,and introduces weight coefficients to balance the convergence rate of the two functions.To avoid local convergence,a crossover operator is applied,which combines multi-point crossover with uniform crossover.To accelerate the convergence rate,a mutation operator of dynamic mutation rate and a migration operator is utilized to overcome the blindness of global convergence.Taking a general 6R robotic manipulator as an example,experiments are conducted by using the single population genetic algorithm and the proposed method.The results indicate that the proposed method can not only guarantee the stability,avoid local convergence,but also improve the convergence accuracy and rate significantly.