Identification for Industrial Robots Using Bedplate Force Sensors with Physics-Based Simulation Models

This paper proposes an identification method for industrial robots by installing a bedplate force/torque sensor under the robot. Bedplate sensor calibration and dynamic identification are realized in one step. A physics-based model is constructed in a simulation environment. Both the real and simulated data are used through a close-looped output error method to solve the dynamic parameters of the robot and the coupling matrix of the force/torque sensor. An experiment is conducted on an industrial robot in which the excitation trajectory for identification is designed and executed. The excitation trajectories minimize the condition number of the regression matrix. Cross-validation trials are used to assess the identification precision, and the bedplate wrench's prediction performance is approximately 97.5^%.