Statistical Correction Framework for Vision-Based Reference Points in Robotic Panel Handling

Abstract
This paper proposes a statistical correction framework to address uncertainty in vision-based reference point detection for robotic panel handling systems. By modeling measurement errors as bounded uniform distributions and applying Monte Carlo simulation, the method estimates corrected reference positions to improve positioning accuracy. Experimental results demonstrate up to 86.2% improvement in accuracy and a 68% reduction in defect rates. The framework supports real-time correction and enables predictive maintenance with anomaly detection accuracy of 94–97%, making it suitable for precision manufacturing applications. Keywords - robotic vision, statistical correction, Monte Carlo simulation, precision manufacturing, uncertainty quantification, predictive maintenance