Neuro-controller for antagonistic bi-articular muscle actuation in robotic arms based on terminal attractors

The conventional limitations of the robotic actuation mechanisms have led to many researchers needing to explore novel biomimetic motor mechanisms as the antagonistic human motor system. In this way, it is of interest to understand the inherent adaptive stiffness, or compliance, and modulation, in different alternative actuation architectures such as the antagonistic bi-articular (AbA) system. These novel AbA actuation mechanisms are characterized by resembling the efficient tendon and muscle build-up over our skeletal structure. In this paper, we propose a Cartesian neuro-controller for a robot manipulator actuated by a simplified adaptive viscoelastic linear AbA system. It is shown that the adaptive closed-loop system enforces terminal attractors, induced by a continuous model-free sliding mode control, simultaneously with a learning algorithm to compensate parametric uncertainties of AbA system through a low dimensional neural network. Numerical simulation results exhibit the feasibility of this approach.