This research presents a framework to develop robust control methods for bio-inspired electrohydraulic soft robotic systems. While soft robotic sensors, actuators and materials are prevalent in research, soft robotic systems are sparse in practical settings. A major contributing factor in confining soft robots to research labs is their dependence on a controlled environment with minimal external disturbances. Therefore, we suggest a framework which will promote the use of soft technologies in real-world applications by leveraging feedback control. Specifically, we apply methods from linear system theory and robust control theory to control electrohydraulic soft systems equipped with redundant actuators and distributed sensors - two mechanisms observed in nature.

Within this work, we employ hydraulically amplified self-healing electrostatic (HASEL) actuators to drive the benchtop systems. Given their fast dynamic responses and their precise actuation, HASEL-driven systems can exploit highly performant control schemes. Initially, we develop data-driven state space models on a HASEL-driven multiple input single output (MISO) system analogous to the human biceps-triceps muscular system. We use the empirically-derived dynamic plants to synthesize model-based control schemes. The resultant closed loop mechanism demonstrates high precision and excellent disturbance rejection.

We extend this work by performing uncertainty analyses on a similar HASEL-actuated multiple input multiple output (MIMO) platform. We build on the theory by developing multiple empirically-derived linear models in various operational regions of the system. These models result in a collection of linear plants that can be used in an uncertainty analysis of the closed loop system to ensure robust stability. Additionally, we utilize ’H Infinity’ controller synthesis techniques to determine an optimal controller to meet performance requirements for the nominal plant. The resultant system demonstrates a balance of excellent dynamic performance and minimal control- input energy. We extend this framework to an electrohydraulic robotic shoulder joint, a system possessing three degrees of freedom. We demonstrate the developed controller synthesis framework on this shoulder-inspired platform, and thus, demonstrate its transferrability to other systems with more degrees of freedom.

Lastly, we design a sensing strategy by fusing bio-inspired distributed sensory networks with these soft electrohydraulic actuators without compromising their exceptional dynamic performance. We demonstrate a distributed light sensing solution which can be tightly integrated with HASEL actuators. These light sensors enable the HASEL-driven systems to be used with feedback control outside of the motion capture area.

We hope the demonstrations in this work inspire others to confidently apply aspects of the methodology to other electrohydraulically-actuated multivariable soft robotic systems. Better yet, we hope others expand upon the framework in aims to promote the use of soft robots in real-world applications.