A Lightweight Calibration and Compensation Framework for High-Precision Mobile Robotic System Under Heavy Load

Abstract

Abstract This article presents a novel mobile robotic system (MRS) with a Stewart platform (SP), aimed at high-precision neutron diffraction applications that involve heavy loads and limited sampling resources. Existing kinematic calibration methods predominantly address geometric errors and fail to account for nongeometric errors introduced by robot relocation or payload-induced deformations. Moreover, current nongeometric error compensation strategies typically require extensive data collection and retraining after each movement, rendering them impractical for mobile robot applications. These challenges are further exacerbated in parallel robots due to their complex kinematics, which makes accurate error compensation particularly difficult. To address these gaps, we propose a data-efficient, lightweight nongeometric error compensation method based on Gaussian process regression (GPR) that requires only 30 samples and a single postmovement reference configuration. Additionally, this article presents a comprehensive self-calibration framework for mobile parallel robotic systems. It includes kinematic calibration using the product of exponentials (POEs) method before movement, a rapid automatic localization method postmovement, and nongeometric calibration to mitigate accuracy degradation at the new location. Experimental results demonstrate a significant improvement in accuracy, with the average position error of the diffractometer reduced by over 80% and joint coordinate errors decreased by no less than 90%.