针对传统的求逆运动学方法相当复杂,一般的神经网络收敛速度慢,精度不高的缺陷,提出一种由3个并行的BP (Back Propagation)神经网络组成的系统来解决运动学逆问题,输人数据分别通过3个并行的BP神经网络,再对输出分别求正运动学解,然后计算误差,最后选择误差最小的作为系统的输出,仿真表明,该方法可以有效地解决运动学逆问题,使用3个并行的BP神经网络可以使整个系统的误差更小,BP神经网络使用Levenberg-Marquardt训练方法,可以使学习收敛速度更快.%Many traditional solutions are usually complex and general neural networks have slow convergence velocity and low precision, for the solution of this problem a neural network for inverse kinematics solution approach has been presented in this paper. The structure of the proposed method is based on using three parallel BP (Back Propagation) neural networks. Input data goes through the three parallel BP neural networks separately. At the end of parallel implementation, the results of each network are evaluated by using direct kinematics e-quations to obtain the network with best result. Simulations show that inverse kinematics problem can be solved effectively. Using three parallel networks can minimize the error of the whole system. BP neural networks use Levenberg?Marquardt training algorithm which is a fast -learning algorithm.
展开▼