HIGH-PRECISION ROBOTIC ASSEMBLY UNDER VARIABLE ILLUMINATION: A ROBUST MECHATRONIC ARCHITECTURE FOR VISUAL SERVOING

Abstract

Introduction: Visual servoing is a key enabler for high-precision robotic assembly; however, illumination variability (dim scenes, backlighting, flicker, and specular highlights) can destabilize feature extraction and degrade closed-loop accuracy. Methods: We propose a lighting-robust visual servoing pipeline that combines (i) photometric normalization, (ii) illumination-aware feature extraction and a signal-quality index q(t), and (iii) confidence-weighted control with safety limits and a robustness monitor. Because full hardware trials were not provided as input, the empirical evaluation in this draft is conducted via a controlled simulation stress test that models lighting-induced measurement noise and dropouts across five illumination profiles. Results: In the simulation study (200 trials per condition), the proposed method improves convergence behavior and success rate in nominal, dim, and backlit conditions, while highlighting residual failure modes under fast flicker and specular events. Across conditions, the results indicate that combining photometric stabilization with confidence-aware control reduces sensitivity to illumination-driven measurement corruption, but requires additional handling for high-frequency lighting and specular transients. Conclusions: The study offers a mechatronic methodology for integrating illumination-aware perception, monitoring, and control into vision-guided assembly.