Output-Prediction Based Nonlinear Control of a Class of Neuro-Musculoskeletal Systems

Motion control of neuro-musculoskeletal systems constitutes a challenging problem. The presented work proposes an approach exploiting the use of a prediction-based control technique to address some of the issues involved. In particular, the skeletal actuators (muscle-tendon complexes) are replaced with a virtual system, emulating the corresponding input/output behavior in the control design process. A control law, exploiting this predictor, prescribes the rate of change of system input (descending signals). It is shown to guarantee uniform ultimate boundedness of the tracking errors. To illustrate efficacy of the approach, the control law is applied to a two degree of freedom skeletal system, actuated by a set of five muscle-tendon complexes, activated by a simple neural model representative of a range of spinal functions.