Predictive Friction Compensation for Control of Pneumatic Actuators

Pneumatic actuators possess several attractive qualities: high power and force density, potentially adaptable compliance, and clean, safe, and low cost actuation. However, control of pneumatic actuators has proven difficult, limited by inherent compliance of the actuator, nonlinear and discontinuous third order dynamics, and friction. Stiction and compliance lead to a sandwiched nonlinearity that causes stick-slip and can cause significant tracking error and even instability. A broadly applicable method of friction compensation is addition of a feedforward term updated from a friction estimate at each time step. Since pneumatic dynamics are slow, achievable compensation can be insufficient. In this work, friction is estimated over a prediction horizon and then input into a model-based predictive controller as an offset term, so that compensation is planned and optimal over the prediction horizon. The controller is tested in simulation. Results are compared to control using instantaneous compensation and are characterized based on performance.