High-Order Control Barrier Functions-Based Adaptive Robust Safety Control for Manipulator Under Model Uncertainty and Disturbance

Safety is a critical issue for robotic manipulators under model uncertainties and external disturbances. To address this, we propose an adaptive robust control framework based on an echo state network (ESN) and high-order control barrier functions (HoCBFs). Unlike traditional CBF methods that only ensure end-effector safety, our HoCBF considers the shortest distance between each link and obstacles, automatically focusing on the joint of highest-risk to provide full manipulator protection. We also design an environment-dependent extended class k function that adjusts gain parameters according to operational conditions, offering greater flexibility in dynamic settings compared to static-gain approaches. Utilizing the ESN’s efficient training and dynamic memory, we approximate unknown nonlinear dynamics and integrate it into the safety controller. Under strong disturbances, our controller demonstrates higher robustness and safety than conventional HoCBF-QP, while being less conservative and offering a larger feasible region than classical robust one thus achieving a better safety-conservatism balance. Simulations and experiments on a UR5e robot confirm the framework’s effectiveness and superiority.