Soft robots are gaining traction due to their adaptability in various applications. Pneumatic soft actuators, known for large deformations, simple designs, and safety, are particularly promising. However, a critical challenge lies in incorporating real-time sensing for accurate feedback control in dynamic environments. Current approaches often involve integrating heterogeneous sensing devices, leading to increased complexity, higher costs, and poor interfacial stability. This research addresses these limitations by designing an intelligent pneumatic soft actuator with an inherently integrated proprioceptive sensor. This integrated design simplifies the system, improves manufacturing efficiency, enhances stability, and lowers costs compared to previous methods involving separate sensing components. The key innovation is using existing structural components of the actuator to serve a dual purpose, specifically employing two parallel conductive microfibers as both radial expansion limiting fibers and the electrodes for a capacitive bending sensor. This integrated approach reduces complexity, fabrication steps, and cost while simultaneously improving durability and reliability.

Numerous studies have explored the development of soft actuators using various mechanisms, including pneumatic, hydraulic, electrical, magnetic, thermal, and ionic approaches. Pneumatic actuators are favored for their large-scale deformations and simple designs. However, achieving reliable feedback control in dynamic environments requires embedded sensing. Previous attempts at self-sensing pneumatic soft actuators have involved embedding or attaching heterogeneous sensing components such as resistive, capacitive, magnetic, inductive, and optical strain sensors. While these methods offer self-sensing capabilities, they often increase structural complexity, manufacturing costs, and reduce interfacial stability between the sensor and the actuator. This inherent incompatibility between soft actuators and rigid sensors hinders broader applicability and usability. The present work aims to address these limitations by developing an inherently integrated sensing solution within the actuator itself, thereby enhancing simplicity, manufacturability, and reliability.

The researchers designed and fabricated a proprioceptive soft actuator consisting of an elastomeric chamber, axial strain-limiting fiber, and radial expansion-limiting fibers. Two parallel conductive microfibers were incorporated as part of the radial expansion limiting fibers, functioning simultaneously as a capacitive bending sensor. The fabrication involved a four-step process: molding the elastomeric chamber using 3D-printed molds, incorporating the axial strain-limiting fiber, helically winding the radial expansion-limiting fibers (including the conductive microfibers), and encapsulating the assembly with silicone elastomer. The conductive microfibers' capacitance changes with the actuator's bending angle, providing proprioceptive feedback. The bending angle was measured using image analysis. Capacitance was measured using an LCR meter. Mechanical characterization involved evaluating bending angle response to pressure, isometric and isotonic force measurements, hysteresis analysis, and durability testing. Finite element method (FEM) simulations and mathematical analyses were conducted to model the actuator's mechanical and capacitive behavior. A two-degree-of-freedom (TDOF) motion controller, incorporating a disturbance observer and PID control, was implemented for closed-loop feedback control. The actuator's performance was evaluated in soft robotic demonstrations, such as a soft gripping system and a soft prosthetic hand, to showcase its capabilities in real-world scenarios.

The inherently integrated proprioceptive soft actuator demonstrated significant advantages over prior art. The actuator achieved a bending angle of up to 240 degrees with high sensitivity (1.2 pF rad⁻¹). Mathematical modeling and FEM simulations accurately predicted both the mechanical actuation and the capacitive sensing behavior. The actuator exhibited negligible hysteresis and high durability, maintaining its performance even after 10,000 cycles of testing. The actuator's response and recovery times were fast (71 ms and 38 ms, respectively, at 50 kPa). The TDOF controller successfully maintained the desired bending angle despite external disturbances, highlighting the actuator's robustness. Furthermore, the fabricated soft gripping system demonstrated stable grasping of diverse objects and a significant increase in load-bearing capacity compared to the open-loop system. The soft prosthetic hand successfully performed various hand gestures and object recognition tasks with high accuracy (approximately 91%). The cost of the actuator is low, estimated at approximately $0.92 per unit.

This research successfully addressed the limitations of previous self-sensing soft actuators by integrating the sensing functionality directly into the actuator structure. The inherently integrated design simplifies system complexity, significantly reduces manufacturing costs, and improves the long-term reliability and stability of the system. The actuator's high sensitivity, low hysteresis, and rapid response time make it well-suited for dynamic environments and real-time feedback control. The demonstrated applications – the soft gripping system and the soft prosthetic hand – successfully showcase the potential of this technology in various robotic systems. These results demonstrate the potential for creating robust, cost-effective, and highly versatile soft robotic systems.

This study presents a novel design for an inherently integrated proprioceptive soft actuator. The actuator's simplicity, high performance, and low cost represent a significant advancement in soft robotics. Future work will focus on further enhancing the actuator's output force through material optimization, refining the analytical model, and improving the long-term stability of the sensor. Exploring applications in more complex robotic systems and incorporating advanced control algorithms is also planned.

While the actuator showed excellent performance, future work could explore improvements in the adhesive properties between the microfibers and the elastomeric matrix to prevent potential microfiber drift and further enhance the sensor's stability. Additionally, the current object recognition accuracy of the prosthetic hand could be improved by applying more advanced machine learning techniques. The influence of environmental factors on the sensor's performance requires further investigation.