Data-based controller for a soft robot actuated by ultrasonic atomization with unknown dynamics and uncertainty

Soft robots are increasingly indispensable in engineering and scientific applications due to their ability to navigate unstructured environments, interact with humans, and emulate biological systems. These tasks pose challenges for rigid robots. Traditional actuation methods for soft robots rely on pneumatic or hydraulic systems, which may face limitations in miniaturized soft robots. Here, a soft robot featuring an actuation system based on vibrating mesh atomization and ethanol evaporation is introduced. However, the physical process of this actuator introduces noise, non-linearity, and hysteresis during displacement and force cycles. Thus, the inherent uncertainty and lack of a dynamic model pose challenges for a classical model-based control. Existing data-based controllers typically prioritize minimizing cost functions to address this issue. In this study, a novel data-based fuzzy network controller for the discussed soft actuation is proposed, integrating an input error function within a sliding mode framework. The controller, along with its learning law, is specifically designed to tackle time-varying parameters resulting from the hysteresis and nonlinear characteristics of soft robots. Rigorous analysis is provided using the Lyapunov method to validate the proposed approach. Numerical experiments were performed using an approximate discretized nonlinear model with noise to describe the data reported in the literature. Experimental validation involves measuring the voltage of a micro-force sensor to track blocking force as a control signal to the plant. Both numerical and experimental results validate the efficacy of the proposed controller in force tracking under external perturbations.

Soft robots play an essential role in engineering and scientific applications due to their adaptability in unstructured environments, safe human interaction, and biomimetic capabilities—challenges that rigid robots face. Traditional actuation methods, such as pneumatic and hydraulic systems, exhibit limitations in miniaturized soft robots. This study presents a soft robot actuated by vibrating mesh atomization and ethanol evaporation. However, the actuator introduces noise, nonlinearity, and hysteresis, complicating model-based control. To address these challenges, a data-driven fuzzy network controller is proposed, integrating an input error function within a sliding mode framework. The controller and its learning law are designed to accommodate time-varying parameters associated with hysteresis and nonlinear behavior. Stability is rigorously analyzed using the Lyapunov method. Numerical experiments utilize a discretized nonlinear model with noise to approximate reported data. Experimental validation measures the voltage of a micro-force sensor to track blocking force as a control signal. Both numerical and experimental results confirm the controller’s effectiveness in force tracking under external perturbations.