机械手逆运动学神经网络算法研究

提出一种基于模糊遗传算法的机械手逆运动学神经网络建模方法。该方法采用3层前向神经网络建立机械手逆运动学模型，应用模糊遗传算法训练神经网络的权系数。此算法可根据种群进化情况，对交叉概率和变 异概率进行在线模糊控制，加快了算法的搜索过程，有效地避免了简单遗传算法中容易出现的初期收敛问题。仿真结果表明，本方法提高了求解精度和收敛速度，不但有效克服了简单遗传算法常出现的初期收敛和BP算法求解精度低、容易陷入局部极小等缺点，而且避免了计算Jacobian矩阵的伪逆，结构简单、容易实现。

Study on Manipulator Inverse Kinematics Neural Network Algorithm

An inverse kinematics neural network modelling method of manipulator based on fuzzy genetic algorithm is proposed. The inverse kinematics model is built by three-layer forward neural network. A fuzzy genetic algorithm is used to select weights of the network. The crossover probability and the mutation probability are changed by on line fuzzy control for expediting search process of algorithm.The premature convergence is effectively conquered, and the convergence speed is improved. The simulation results show the precision of the solutions of inverse kinematics and the convergence speed is higher than that of simple genetic algorithm and BP algorithm. It gets over premature convergence of simple genetic algorithm and shortcomings of BP algorithm, such as the low precision of solutions and stacking easily into the local minimum points. The computation of the pseudoinverse of Jacobian matrix is avoided. This method is simple and easy to be realized.