Conform to grip: Joint reaction force-driven adaptive variable admittance control of biped climbing robots

Current biped climbing robots encounter persistent challenges in aligning their grippers with structural elements like poles, regardless of teleoperation or perception-based control implementations. To overcome this limitation, we present a joint reaction force-driven adaptive variable admittance control framework that enables autonomous compliant alignment with enhanced precision. The proposed method utilizes unintentional gripper-pole contact-induced joint reaction forces to drive alignment through a variable damping admittance controller. Damping parameters are adaptively regulated through proportional-derivative control law based on real-time joint reaction force errors. By establishing zero reference reaction force, the system concurrently accomplishes dual objectives: gripper pose alignment and reaction force minimization. Experimental validation confirms that our framework significantly enhances alignment efficiency and gripping reliability without requiring explicit gripper-pole pose detection. This methodology proves particularly effective for robotic systems transitioning between open-chain and closed-chain configurations while resolving inherent joint conflicts.