Pneumatic actuators are frequently selected for use in machines intended for human interaction because of their clean operation and natural compliance. However, the compliance, coupled with friction, can also make motion control difficult, leading to the use of more aggressive controllers, such as high-gain PID or sliding mode control, which result in stiff closed-loop system behavior. Model-based options are needed to obtain behavior that provides a better trade-off of compliance and accurate position control. In particular, Model Predictive Control (MPC) is suggested; through the use of constrained optimal control, it offers a framework for minimizing tracking error while enforcing force constraints that ensure low impedance behavior.
This paper assesses the suitability of controllers for pneumatic systems to positioning applications in which human-machine interaction is anticipated. MPC is compared against commonly-used alternatives for such scenarios: sliding mode, PID, and impedance control. Results are shown in simulation, and use spectral analysis of the impedance and closed loop tracking to characterize the balance of compliance and accuracy for each of the controllers.
