Joint Dynamics and Adaptive Feedforward Control of Lightweight Industrial Robots

The use of lightweight strain-wave transmissions in collaborative industrial robots leads to structural compliance and a complex nonlinear behavior of the robot joints. Furthermore, wear and temperature changes lead to variations in the joint dynamics behavior over time. The immediate negative consequences are related to the performance of motion and force control, safety, and lead-through programming.This thesis introduces and investigates new methods to further increase the performance of collaborative industrial robots subject to complex nonlinear and time-varying joint dynamics behavior. Within this context, the techniques of mathematical modeling, system identification, and adaptive estimation and control are applied. The methods are experimentally validated using the collaborative industrial robots by Universal Robots.Mathematically, the robot and joint dynamics are considered as two coupled subsystems. The robot dynamics are derived and linearly parametrized to facilitate identification of the inertial parameters. Calibrating these parameters leads to improvements in torque prediction accuracy of 16.5 %-28.5 % depending on the motion.The joint dynamics are thoroughly analyzed and characterized. Based on a series of experiments, a comprehensive model of the robot joint is established taking into account the complex nonlinear dynamics of the strain-wave transmission, that is the nonlinear compliance, hysteresis, kinematic error, and friction. The steady-state friction is considered to depend on angular velocity, load torque, and temperature. The dynamic friction characteristics are described by an Extended Generalized Maxwell-Slip (E-GMS) model which describes in a combined framework; hysteresis characteristics that depend on angular position and Coulomb friction that depend on load torque. E-GMS model-based feedforward control improves the torque prediction accuracy by a factor 2.1 and improve the tracking error by a factor 1.5.An E-GMS model-based adaptive feedforward controller is developed to address ...