基于遗传算法的机器人高精度运动复现方法

Robot high precision motion reproduction based on genetic algorithm

为了实现机器人高精度运动轨迹的复现,需要把双目视觉采集到的离散的采样数据作为示教数据,由于机器人数学模型的实际参数与其名义值存在偏差,导致由视觉测量获得的机器人末端位姿逆解得到的关节变量直接复现时与原目标轨迹有一定的偏差。文中采用遗传算法对数据进行处理,将机器人运动学模型各参数的偏差都归结到关节变量上,建立位姿误差模型,求出目标函数最小时的关节变量修正值并补偿到控制器中,控制机械手完成高精度运动轨迹的复现。该方法避免了复杂的轨迹规划和传统的较为复杂的机器人标定补偿方法,提高了机器人复现精度,并能应用于任意复杂运动。在自行研制的6R串联机器人上对该方法进行了验证。

To achieve robot high precision trajectory reproduction, discrete sampled data collected from binocular vision need used as the demonstration data, due to difference between actual parameter and nominal value of robot mathematical mod- el, there will has deviation compared to original target trajectory when reproducing with the joint variables obtained by inverse solution of robot-end pose collected through vision measuring. To address this question, genetic algorithm was used to process the data, attributing all the parameter deviations of kinematics model to joint variables, established pose error model, then cal- culated the correction values of angle variables corresponding to the minimum objective function and compensated them into controller, then high precision reproduction tasks was completed. This proposed method avoids complex curve fitting and tradi- tional complex robot calibration compensation methods, improves accuracy and can be applied to any motion, the method was validated in the 6R serial robot developed by ourselves.