Dissertation Title: Crosstalk-free sensing-actuation integrated soft robots for intelligent operations

Date: 2025/07/30 – 2025/07/30

Dissertation Title: Crosstalk-free sensing-actuation integrated soft robots for intelligent operations

Speaker: Shoulu Gong, Ph.D. candidate in Shanghai Jiao Tong University

Time: July 30 from 8:30 a.m. to 10:30 a.m., 2025 (Beijing Time)

Location: Room 403, Longbin Building

Abstract

Soft robots, due to their inherent compliance, demonstrate significant potential for applications such as object manipulation, human-robot interaction, and environmental exploration, surpassing traditional rigid robot designs. Soft robots can actively alter their geometry and dimensions to better perform tasks in complex environments or passively deform to accommodate environmental changes. These attributes grant superior adaptability and safety for soft robots. Current research efforts focus on integrating flexible sensors and other sensory units into the surfaces or internal structures of soft robotic bodies. This approach results in integrated sensing-actuation systems, significantly enhancing the robots’ perceptual abilities and environmental adaptability. Moreover, such integration paves the way for developing intelligent soft robots capable of learning, adapting, and making autonomous decisions by leveraging the synergy between sensing and actuation. However, while sensing-actuation integration has been widely adopted in soft robot design and intelligent operations, several challenges remain. These include feedback signal crosstalk, low control precision, and the complexity of integrating various system components.

This dissertation addresses these challenges by proposing crosstalk-free sensor-actuator integration strategies for soft robots and demonstrating their applications in three representative soft robot configurations:

(1) To mitigate the crosstalk effects caused by large deformations in soft grippers on the feedback

signals of the sensing unit, a highly stretchable but strain-insensitive flexible force sensor is designed. This sensor is built on a strain-gradient elastomer substrate prepared through a silicone-induced poisoning and reconnection process, which produces locally variable mechanical properties. By positioning the sensing unit in a high-stiffness region of the substrate, strain interference is effectively shielded. The resulting flexible force sensor demonstrates exceptional strain insensitivity, remaining unaffected under tensile strain up to 96%, bending deformation up to 56 degrees, and torsional deformation up to 33 degrees. It also exhibits an external pressure detection limit below 20 Pa. When integrated into a soft gripper, this strain-insensitive sensor provides crosstalk-free tactile perception. Experimental results illustrate that the integrated soft gripper can accurately measure contact forces on objects of varying sizes without being influenced by its own bending motions.

(2) To address the low control accuracy of soft continuum robots, a sensing-actuation integrated design incorporating human drag teaching capability is proposed. This approach uses self-sensing cables to achieve precise feedback and enhanced control in a two-section soft continuum robot (with a body length of 230 mm). The robot features a pneumatic – cable-driven antagonistic actuation mechanism, inspired by the motion of antagonistic muscle pairs. Cables are connected to servos via tension sensors, allowing the integration of sensing and actuation functions within a single system. A kinematic model and a coupled mechanics-geometry model have been developed to analyze the relationship between cable length variation, tension, posture, and external contact. Using this coupled model, the robot can accurately map its movement and interaction state based on real time feedback from cable tension and length. Additionally, a drag-teaching-based collaborative motion programming method is introduced, enabling intuitive programming through direct human guidance and significantly enhancing absolute positioning accuracy, achieving an error of less than 1 mm.

(3) To address the perception system redundancy in small-scale soft locomotion robots, an integrated

strategy is proposed to construct compact soft locomotion robots with self-sensing adaptability. This

approach utilizes an elastic spine made of piezoelectric fiber materials, which integrates both sensing and actuation functions. The elastic spine serves as both an auxiliary actuator and flexible sensors, with crosstalk between these two functions effectively eliminated through a self-decoupling circuit. Based on this strategy, high-performance small-scale soft crawling robots and soft amphibious robots have been developed. Machine learning techniques are employed to enable the robots to learn feedback signals from the elastic spine, allowing them to recognize different environments. As a result, the robots can actively adjust their motion behavior based on environmental recognition, achieving optimal motion efficiency. Soft robots with self-sensing adaptation demonstrate a 39.5% to 80% improvement in motion speed across various terrains, enabling them to perform intelligent operations such as self-obstacle detection and avoidance.

The design methodology proposed in this dissertation greatly enhances the synergistic efficiency between sensing and actuation units in integrated soft robots. These approaches can be further applied to the development of intelligent soft robots for various applications.

Biography

Shoulu Gong received the B.S. degree in mechanical engineering from Shandong University in 2020. He is currently pursuing the Ph.D. degree in mechanical engineering in Shanghai Jiao Tong University, supervised by Prof. Lei Shao. His current research interests include crosstalk-free sensing-actuation integrated soft robots and their intelligent operations.