基于改进适应度函数组合法的机器人逆运动学求解

提出基于分离-重构方法获得智能算法的适应度函数,并与解析法相结合求解一般结构多自由度机器人逆运动学,一般结构满足至少一组相邻关节轴线交于一点。利用分离-重构的方法构建分离-重构适应度函数,使用智能算法验证分离-重构适应度函数,求解出部分关节角,再结合解析法求出其余关节角。以一般结构的FANUC CRX-10iA型协作机器人为算例,使用改进的CMA-ES算法验证分离-重构适应度函数,仿真结果表明,在相同条件下求解的点对点运动中,单次平均求解时间为0.0040s,适应度函数值稳定在10-7数量级,且迭代收敛次数稳定在25次左右;在空间中对两种连续轨迹进行跟踪时,单次平均耗时分别为0.0068s和0.0102s,位置误差均稳定在10-7m数量级,且各关节运动曲线平滑,提升了多自由度机器人逆运动学分析的效率和精度。以REBot-V-6R型六自由度机器人为实验对象,基于VC++6.0创建MFC实验平台,进行连续的空间轨迹跟踪实验,仿真与实验结果验证了该方法的有效性。

Solution of Robot Inverse Kinematics with Combination Method of Improved Fitness Function

Robot inverse kinematics analysis is the basis of many researches, but the process of multi-degree-of-freedom robot with a general structure inverse kinematics is complicated, and the analytical method may have no solution. Combining the improved fitness function of intelligent algorithm with the analytical method, a method for solving the inverse kinematics of multi-degree-of-freedom robot with a general structure was proposed, which improved the efficiency and accuracy of multi-DOF inverse kinematics analysis. The general structure satisfied at least one set of adjacent joint axes intersect at one point. Using the method of disconnection and reconnection to construct the fitness function of disconnection and reconstruction, and intelligent algorithm was used to verify the fitness function of disconnection and reconstruction and solve some joint angles, and then the analytical method was combined to find the other joint angles. Compared with the fitness function constructed by other methods, this method simplified the process of establishing the fitness function and reduced the algebraic dimensionality of the fitness function. Taking the FANUC CRX-10iA collaborative robot as an example and using the improved CMA-ES algorithm to verify the fitness function of disconnection and reconstruction, the simulation results showed that in the solution of point-to-point inverse kinematics under the same conditions, the single average solution time was 0.0040s, the fitness function value was stable at the order of 10-6, and the iteration convergence times was stable at about 25 times. In order to verify the effectiveness of this method in environments with different levels of complexity, continuous trajectory tracking was performed in two-dimensional and three-dimensional spaces, the average time for a single time was 0.0068s and 0.0102s, respectively, the position error was stable at the level of 10-7m, and the motion curve of each joint was smooth. Taking REBot-V-6R robot of six-DOF as the experimental object, the MFC experimental platform was created based on VC++6.0, and continuous space trajectory tracking experiments were carried out. Simulation and experimental results verified the effectiveness of the method.