Shape Self-sensing Pneumatic Soft Actuator Based on the Liquid-metal Piecewise Curvature Sensor

The shape estimation technique can help solve the end positioning or grasping control of soft robots. However, there is a lack of sensing and modeling techniques for accurate deformation estimation and soft robots with axial elongation, e.g., pneumatic soft actuators (PSAs). This paper presents a shape-self-sensing pneumatic soft actuator (SPSA) with integrated liquid-metal piecewise curvature sensors (LMCSs). Two types of LM composite (Ga-In-Sn/Ga2O3 composites for the sensor and Ga-In-Sn/NdFeB/Ni for the electronic wire) were used to build the strain sensor network. Furthermore, a piecewise variable curvature (PVC) model was developed to predict the bending deformation of the soft actuator. A two-SPSAs-based gripper was built to test the identification performance of LMCSs. The results indicate that SPSA could perform contact and size identification using the PVC model. In addition, the K-nearest neighbors (KNN) algorithm was used to classify the shape of the targets. Finally, the circular, triangular, and square targets were identified with an accuracy rate of 93.3%. The work was expected to be applied to the size and shape perception and deformation planning of soft robots.