Iterative Machine Learning for Precision Trajectory Tracking with Series Elastic Actuators

摘要: When robots operate in unknown environments small errors postions can lead to large variations the contact forces, especially with typical high-impedance designs. This potentially damage surroundings and/or robot. Series elastic actuators (SEAs) are a popular way reduce output impedance of robotic arm improve control authority over force exerted on environment. However this increased forces lower comes at cost positioning precision and bandwidth. article examines use an iteratively-learned feedforward command position tracking when using SEAs. Over each iteration, responses system quantized inputs used estimate linearized local models. These estimated models obtained complex-valued Gaussian Process Regression (cGPR) technique then, generate new input based previous iteration's error. illustrates iterative machine learning (IML) for two degree freedom (2-DOF) arm, demonstrates successful convergence IML approach