A Holistic Indirect Contact Identification Method for Soft Robot Proprioception

Area of Science:

• Robotics
• Materials Science
• Control Systems

Background:

• Soft robots offer advantages but present control challenges due to their inherent compliance and back-drivability.
• Accurate perception of robot pose (position and orientation) is crucial for effective control, especially under perturbed conditions.
• Existing methods often struggle with the complex deformations and external contacts characteristic of soft robots.

Purpose of the Study:

• To develop a holistic approach for perceiving the pose of soft robots during free bending and under external contact perturbations.
• To enable precise control of soft robotic systems by accurately estimating their deformed state.
• To provide a generalizable method for contact identification and pose estimation in soft actuators.

Main Methods:

• Utilized multiple soft strain sensors non-collocated with the point of contact to measure deformation.
• Developed a data-driven model to characterize free-bending deformation modes (front, lateral, twist).
• Implemented a perturbation perception method by comparing sensor deviations from an unperturbed state to estimate pose.

Main Results:

• Successfully characterized sensor responses to various deformation types on a 2-degree-of-freedom soft finger.
• Demonstrated the ability to perceive perturbed poses on both single-finger and two-finger soft robotic grippers.
• Validated the effectiveness of the holistic contact identification and pose perception method.

Conclusions:

• The proposed holistic approach enables robust pose perception in soft robots, even with external perturbations and contact.
• This method is generalizable and crucial for advancing the control capabilities of compliant robotic systems.
• Improved pose perception directly addresses a key limitation in the practical application of soft robots.