摘要
分析了6-DOF精密并联机器人末端位姿的误差来源及以往误差补偿方法的局限性。通过测量末端位姿,提出了基于BP神经网络在精密定位的局部工作空间内对机器人关节空间进行误差补偿的方法。确定了BP神经网络模型,建立了误差补偿的数据样本,并对数据样本进行了标准化。用实验对比的方法确定了隐层神经元的个数,同时对网络的推广能力进行了验证。经过误差补偿,6-DOF精密并联机器人的平移定位误差下降了80%,转角定位误差下降了60%。实验结果表明,基于BP神经网络的误差补偿方法对机器人局部工作空间的补偿具有明显的效果,能够满足精密并联机器人工作的精度要求。
The main error sources and the limitations of conventional error compensation for the 6-DOF precision parallel robot were discussed. An error compensation method based on Back Propagation (BP) neural network for the articulatory space of a parallel robot was presented in the local workspace of precision positioning by measuring the end pose. BP neural network model and datum sample of error compensation were established, and the datum sample was standardized. By the experiment, the numbers of node in hidden layer was obtained. In order to improve the generalization performance, the overfitting was prevented in the network training. After error compensation, the positioning error and the orientation error reduced by 80% and 60%, respectively. The experimental results show that the error compensation based on B1c neural network has an obvious effect on that of the articulatory space of parallel robot, which satisfies the accuracy requirement of the precision parallel robot.
出处
《光学精密工程》
EI
CAS
CSCD
北大核心
2008年第5期878-883,共6页
Optics and Precision Engineering
基金
国家自然科学基金资助项目(No.50605013)
教育部高等学校重点学科建设项目
上海市重点学科建设项目(No.Y0102andBB67)
上海大学创新基金资助项目
关键词
并联机器人
BP神经网络
定位误差
误差补偿
parallel robot
BP neural network
positioning error
error compensation