Force-Based Online Estimation and Adaptive Alignment for Robotic Peg-in-Hole Insertion

Abstract

In recent years, numerous methods for robotic peg-in-hole insertion have been developed to automate the assembly process. This paper introduces a novel approach that integrates a real-time control and estimation framework to dynamically estimate the hole's location. A contact model is developed to relate the forces and relative position, and a particle filter is used for real-time estimation. This study employs a Model Predictive Control (MPC)-based approach to guide the robot for informed probing points, further enhanced with two step-size adjustment techniques: fixed and adaptive, to correct the misalignment. Experimental validation using a UR robot and a peg-hole test bed demonstrates the effectiveness of the proposed method, with successful insertion rates achieved under various misalignments and initial conditions. This robust framework offers a significant advancement in high-precision assembly by enabling informed, data-driven path planning in industrial automation.