Real-Time Learning Control for Quadruped Robot Velocity Tracking

This research presents a control framework for quadruped robots that achieves accurate velocity tracking by combining Proportional-Derivative (PD) control, Iterative Learning Control (ILC), and Gaussian Process Regression (GPR). The PD controller provides real-time feedback using inverse kinematics, while ILC learns feedforward torques from repetitive gait cycles to compensate for unmodeled dynamics. GPR then generalizes these learned torques across different velocities, eliminating the need for relearning at each speed.