Abstract:Aiming at the problems of discontinuous trajectory control,high complexity and low comprehensive efficiency existing in the existing manipulator movement deviation control technology,a hybrid neural network control algorithm based on machine learning and deep learning was proposed.The spatial coordinate transformation relationship of each joint and link of the manipulator was analyzed,a hybrid neural network model based on RBF was constructed,the inverse multiple quadratic function was selected as the activation function of the model,and the weights of the middle hidden layer and the output layer were determined respectively.Introducing the LSTM algorithm,the structural design of the input gate,forgetting gate and output gate of the LSTM algorithm was used to suppress the gradient expansion problem occurred in coordinate data training,and the precise trajectory correction instruction was given.The simulation results show that the average deviation of sampling points is 0.02 mm,and the VARP value is close to 0.The hybrid neural network algorithm has better automatic control stability and higher control efficiency.