A fast identification method for multi-joint robot load based on improved Fourier neural network

The joint robots are used in kinds of production processes. A fast time work load identification is the premise to ensure the safe operation of robots. However, in some special application scenarios, the work load can not be measured directly, which is often identified indirectly by dynamic methods. Because of the obvious nonlinearity and the uncertainty of model parameters, the accuracy and efficiency of load identification are not enough. Therefore, based on Fourier neural network, this paper proposes an improved model to realize load identification, in order to improve the prediction accuracy and timeliness of system load parameters. Compared with the dynamic model solution method, the proposed method has higher accuracy and faster calculation speed. It only needs to learn several alternate sample sets in the prediction range, and any results in the prediction range can be accurately identified with good generalization ability. The sensitive parameters of the network are also analyzed, and the performance is compared with that of mature neural network algorithm. In this method, two neural network models are combined to effectively identify different feature sets in high-dimensional data, so as to realize efficient load parameter identification and provide reference for parameter identification of complex nonlinear systems.