High performance pneumatics using Model Predictive Control

High force density, compact form factor and muscle like compliance properties makes pneumatics actuators quite desirable for robotic applications. However, the compressibility of the air not only throttles their bandwidth but also makes them harder to control. In this work, we leverage online trajectory optimization techniques to design high performance controllers for pneumatic actuators of ADROIT manipulation system [1]. These controllers use future predictions to act in anticipation. We show that the proposed controllers prepares the system ahead of time in order to achieve better performance and use less controls while doing so. Hardware results are presented on the pneumatic actuators of the ADROIT platform. Controller’s robustness is evaluated using hardware variants.