A Model-Free Redundant Manipulator Control Scheme Based on Tikhonov Regularization Cerebellar-Inspired Network

In practical engineering scenarios, the motion control accuracy is affected by measurement noise arising from sensor errors and model uncertainties due to mechanical abrasion and structural variability, limiting the robustness and adaptability of the existing model-based schemes. To address these issues, in this article, inspired by the motor coordination of the human cerebellum, a novel model-free Tikhonov-regularized cerebellum-inspired (MFTRC) control scheme is proposed. This scheme mitigates the effects of measurement noise while avoiding reliance on the accurate model of the redundant manipulator, thereby achieving strong robustness and adaptability in practical engineering environments. Specifically, a Tikhonov-regularized cerebellum-inspired (TRC) network is developed to achieve model-free adaptive control, while a neural dynamics (NDs) solver is employed to address kinematic constraints and generate reference signals for adaptive learning of the TRC network. The TRC network is integrated with the ND solver to formulate the MFTRC control scheme, thus providing velocity-level control for trajectory-tracking tasks. The simulations and physical experiments verify the feasibility and effectiveness of the proposed scheme, which is more accurate and robust than the state-of-the-art methods.