From Actuation to Perception: Sensorizing a Compliant Gripper for Perceiving Deformation in Contact-Rich Behaviors
Abstract
Contact feedback is critical for robotic manipulation in contact-rich tasks across industrial and everyday applications, enabling safe interaction between the robot and its environment. Traditional approaches rely on expensive and fragile force-torque or tactile sensors mounted on grippers, limiting their adoption. To address these challenges, we introduce the "From Actuation to Perception" paradigm: a cost-efficient and durable vision-based deformation sensing method integrated into a compliant robotic gripper. Inspired by how humans perceive relative force but not precise absolute force in contact-rich tasks, the paradigm works when external forces deform the gripper’s soft structure, and an embedded camera captures these visual deformations via marker tracking. A geometric model then converts the tracked deformations into real-time proprioceptive force sensing clues for the robot. The gripper excels not only in robust grasping of diverse objects but also in versatile contact-rich behaviors. Extensive experiments demonstrate its stability, accuracy, and feasibility, highlighting substantial potential for advancements in real-world robotic applications.