Stribeck Friction Model Identification Based on Genetic Algorithm

Abstract

Aiming at the nonlinear problem of friction force when the robot is drag teaching, the stribeck model is used to describe the system friction, and a step-by-step identification method of friction coefficient based on parameter identification and genetic algorithm (GA) is proposed. First, the Newton-Euler method is used to establish the dynamic model, and the linear coulomb-viscous friction model is established. Then, the above-mentioned model parameters are identified by the high-speed dynamic parameter identification method, and the coulomb-viscous friction coefficient in the identification result is substituted into the stribeck friction. The model is used as a known quantity. Finally, the stribeck threshold and static friction coefficient are optimized by the GA, and the identification accuracy of the model is judged by the root mean square (RMS) of the theoretical value and the actual value of the friction torque. The experimental results show that, compared with the traditional coulomb-viscous friction model, the RMS value of the stribeck model decreases by 26% approximately at low speed. At the same time, compared with the RMS value of the GA to identify the full parameters of the stribeck model, the RMS value of the method proposed in this paper is reduced by 11%. This method not only improves the accuracy of the robot dynamics model but increases the flexibility of dragging during the drag teaching process, which fully demonstrates the effectiveness and feasibility of the method.