Admittance-Based Orientation Control for Perception-Driven Robotic Inspection Assistance

Abstract

Precision and adaptive alignment of sensing modalities with complex surface geometries isdesired in robotic inspection while maintaining robustness to perception uncertainty and operator input. This work presents an admittance-based orientation control framework for perception-driven robotic inspection. A unified control structure has been designed to enable perception-derived orientation errors and teleoperation inputs to be mapped into equivalent force and torque commands with the end-effector of the manipulator modeled as a virtual mass–damper system. Surface normals are estimated from depth point clouds using robust methods, and a critically damped PD controller generates bounded virtual torques subject to saturation constraints. These are integrated through an admittance model to produce smooth task-space velocity commands for execution on a collaborative manipulator. Demonstration of stable and responsive orientation alignment, justification of the choice of filters is presented through experimental results, with close agreement between simulation and hardware, validating the effectiveness of the proposed approach for compliant, perception-aware robotic inspection.