Parameter identification of flexible robotic joint based on multi-sensor information

The flexible robotic joint is an integrated mechatronic device that consists of harmonic drive reducer, torque motor and multi-sensors in one system. The complex structure make it difficult to identify the system parameters such as torque constant, friction, stiffness and damping. In this paper, a set of methods to identify these parameters based on multisensor information are proposed. The experimental data from link side and motor side position sensors, joint torque sensor and motor current sensors are employed to achieve the torque constant and friction model. Mechanical impulse excitation is applied to test the joint elasticity and damping properties. A hysteresis loop which reflect the stiffness of joint system is obtained through the mapping between torsion angle and output torque data. Neural Network is introduced to precisely fit the stiffness. Experiment results validate the curve fitting coincide with the test data.