机械手轨迹规划的神经网络逆模控制

Neural Network Inverse Model Control for Robot Arm Trajectory Planning

针对二自由度机械手动力学模型的非线性和参数的不确定性,提出了一种神经网络与逆模控制相结合的控制策略。针对传统BP算法在神经网络训练后期收敛速度慢且容易陷入局部极小的缺点,提出一种快速启发式学习算法。采用所提出的快速启发式网络学习算法训练多层前馈神经网络,建立机械手的逆动力学模型,实现对机械手的非线性控制。仿真结果表明了所提出控制策略的有效性和快速启发式网络学习算法的快速收敛性。

To the non-linear characteristic and parameter uncertainty in 2-DOF robot arm control, a strategy that combines neural network and inverse model control is proposed. Because the traditional BP algorithm has some disadvantages of slow convergence and getting easily into local minima at the last time of training, a fast speed and heuristic algorithm is introduced to train a multiple-feed-forward neural network, and then the inverse model of robot ann is established. The simulation result shows the validity of control strategy and the fast speed convergence of the proposed algorithm.