Unifying Robot Trajectory Tracking with Control Contraction Metrics

摘要

Design of tracking controllers for robotic systems is a classical problem, as old as robotics itself. Most industrial robots are fully-actuated, i.e. for each configuration coordinate (degree of freedom) there is an associated actuator. Additionally, many walking robots are controlled so as to remain in a fully-actuated state, i.e. at least one flat foot firmly planted on the ground (see, e.g., [1] and references therein). For fully-actuated systems, there are several generally applicable methods for tracking control: independent joint control using PID, computed torque, sliding control, and passivity/energy-based methods, to name but a few [2, 3].