Comprehensive modeling and identification of nonlinear joint dynamics for collaborative industrial robot manipulators

Abstract For collaborative robots, the ability to accurately predict the actuator torques required to realize the desired task is highly important. This will improve and guarantee the safety, motion and force control performance, and smooth lead-through programming experience. Thus, this paper presents the investigation towards comprehensive modeling and identification of nonlinear joint dynamics for collaborative robots. The proposed joint dynamics model and identification describes the most dominant dynamic characteristics of robot joints that comprise strain-wave transmissions, such as nonlinear friction, nonlinear stiffness, hysteresis, and kinematic error. Position-dependent backlash characteristics is observed and quantified using our proposed identification method and the Generalized Maxwell-Slip friction model is extended to describe the observed phenomena. The developed dynamic modeling and identification procedures provides insightful guidance for the design and model-based control of collaborative robots.