6R机器人逆运动学求解与运动轨迹仿真

中文摘要: 针对如何提高6R机器人逆运动学求解的精度和效率问题，提出一种基于动态模糊神经网络进行求解的方法。根据6R机器人逆运动学方程组具有高维非线性、求解复杂的特点，对动态模糊神经网络进行改进，使其能够应用于多输入多输出系统，建立运动学逆解预测模型。通过正运动学方程获取工作空间位姿样本，以工作空间的位姿作为预测模型的输入变量，以关节空间中的关节角作为输出变量，用样本数据对逆解预测模型进行训练。最后，运用该模型对KR16-2机器人进行复杂运动轨迹仿真，并与RBF和BP神经网络模型的求解效果进行比较，结果显示，基于动态模糊神经网络的6R机器人运动学逆解预测模型具有精度高、鲁棒性优和泛化能力强的特点，证明了该方法的可行性和有效性。
Abstract: A new method of solving 6R robot inverse kinematics equations based on dynamic fuzzy neural networks（D-FNN）was presented to improve its accuracy and efficiency. In view of the high-dimensional nonlinearity of 6R robot inverse kinematics equations and the complexity of solving these equations, the D-FNN was improved to fit for multiple-input multiple-output system, and also to establish inverse kinematics solution prediction model of 6R robot. Both position and orientation samples in work space were obtained through forward kinematics and were regarded as input variables of prediction model, the output variables of which were joint angles in joint space. Inverse kinematics solution prediction model was trained by sample data. At last, this prediction model was applied to complex movement trajectory simulation of KR16-2 robot, and the results of the prediction were compared with those of prediction models based on radial basis function (RBF) and back propagation(BP) neural networks. The comparison showed that the D-FNN prediction model of solving 6R robot inverse kinematics equations featured high accuracy, optimal robustness and strong generalization ability, and was proved to be feasible and effective.