Iterative motion learning with stiffness adaptation for multi-joint robots

This paper proposes an iterative motion learning method with stiffness adaptation for energy saving in multi-joint robots. The method iteratively updates both a desired motion and joint stiffness using time-series data of the actuator torque obtained in a control trial with the desired periodic motion. The iterative method is designed to realize the convergence of the desired motion to an energy-efficient motion. Although the trajectory of the desired motion is updated iteratively, we can specify the boundary conditions of the desired motion. The advantages of the updating process are that it does not require exact physical parameter values of the robots or elaborate numerical calculations. Therefore, if the control trials are accomplished without exact parameter values or elaborate numerical calculations, the whole process works without them. We mathematically analyzed the motion convergence, and showed that the updating process improves the desired motion. The simulation results illustrate the effectiveness of the proposed method.