Data-Driven Inverse Control of Pneumatic Soft Robotic Actuators with On-Off Valves: Simulated Performance

Abstract

This article describes a data-driven control method for fast and accurate pressure regulation of pneumatic soft robotic actuators with on-off valves. Due to the non-linearity and binary control input, linear controllers are not ideal for pressure regulation as they must be tuned conservatively, which results in a slow transient response. Alternatively, non-linear methods can provide a faster response time but are complicated by the need for an accurate model. This article proposes an inverse control approach, where the control action is parameterized by the valve on-time, and the system output is the cumulative change in pressure after the control action is complete and all system transients have decayed. This approach eliminates the need for a dynamic model and results in a small space of input-output combinations that can be measured from a calibration experiment. The result is a non-linear inverse controller that does not require a dynamic model. The proposed method is compared in simulation to a hysteresis controller with three different valve orifice diameters. Significant improvements to the pressure regulation and valve lifetime are observed. This method is expected to find applications in soft robotics where fast step changes and dynamic trajectories are required.