Analysis And Modeling For Flexible Joint Hysteresis Behavior For Light Industrial Robots

In industrial production,the application requirements of industrial robots are becoming more and more extensive,and the requirements for their operating speed and accuracy are becoming higher and higher.Flexible joints are one of the keys to determining the positioning accuracy of industrial robot.Harmonic reducer is one of the main components of flexible joints of industrial robots,and its stable accuracy has become the key to restrict the development of high-precision robot industry.The flexible joint containing the harmonic reducer exhibits a non-smooth strong nonlinear hysteresis characteristic,which inevitably affects the transmission accuracy.From the angle of the information compensation,the compensation of model is to improve the conversion accuracy of harmonic drive.Aiming at the flexible joints of industrial robots containing harmonic reducers,a non-smooth strong nonlinear hysteresis characteristic is exhibited.This paper proposes the following two modeling methods:(1)Online Sequential Extreme Learning Machine-based Hybrid ModelIn order to reflect the basic characteristics of the hysteresis of the flexible joint,a hysteresis like operator is constructed.By connecting the hysteresis-like operator and Online Sequential Extreme Learning Machine(OS-ELM)in series,an Extreme Learning Machine hysteretic hybrid model of the flexible joint hysteresis behavior for industrial robots is proposed.In the hysteretic hybrid model,the adoption of OS-ELM with the high learning efficiency and generalization ability can effectively avoid the problems of the slow training speed and local minimum in learning the model parameters by using the gradient descent method for improving the accuracy of the model.(2)Quantum RBF Neural Network Hybrid Structure ModelThe hysteresis characteristics of flexible joints have a multi-value mapping relationship.The flexible joint hysteresis characteristic has a multi-valued mapping relationship.Under different input frequencies,the equivalent inputs have different outputs,that the output has uncertainty.Qubit have the characteristics of state memory and entangled coherent states.A qubit can represent two different results,namely,the output uncertainty of qubit corresponds to the output uncertainty of the hysteresis.Aiming at the hysteresis characteristic of flexible joints,a hybrid structure hysteresis model of quantum RBF neural network is proposed.It consists of a quantum RBF neural network and a quantum RBF neural network that completes nonlinear mapping in series.A quantum RBF neural network models the hysteresis of the harmonic drive to obtain the hysteresis characteristics of the flexible joint,then the second quantum RBF neural network is connected in series,which mainly completes the non-linear mapping and improves the generalization ability of the hysteresis model.In the quantum RBF neural network hybrid structure-based hysteresis hybrid model,Adam optimization algorithm is introduced for improving the learning speed of the model.The model is validated by the experimental data under the different input conditions.Quantum RBF neural network hybrid structure hysteresis model has better prediction error effect than online sequence extreme learning machine-based hybrid model,but the model structure is relatively complicated.Two models can effectively model the hysteresis characteristics.By using different computing power platforms,two models can effectively model the hysteresis characteristics.