Eye-in-hand vision based PID control of a Dobot Magician with online homography calibration for precision manipulation
TL;DR
This study evaluates a vision-guided robotic pick-and-place system integrating an eye-in-hand camera with a proportional integral derivative (PID) controller and an adaptive online homography calibration algorithm, suggesting the static homography model already provides high baseline accuracy.
Abstract
This study evaluates a vision-guided robotic pick-and-place system integrating an eye-in-hand camera with a proportional integral derivative (PID) controller and an adaptive online homography calibration algorithm. To mitigate environmental disturbances, the framework utilizes hue, saturation, and value (HSV)-based object detection and a RANSAC-refined homography method that continuously updates the mapping between image pixels and the robot workspace. The system’s precision was tested under varying lighting conditions by comparing a non-PID baseline against the PID-integrated framework using mean execution time, standard deviation (SD), and t-tests. Results indicate that while both configurations maintain stability, their performance differences are not statistically significant, suggesting the static homography model already provides high baseline accuracy. These findings underscore the importance of rigorous system identification, recommending that future research transition toward advanced control strategies and spatial trajectory tracking metrics to further enhance robotic manipulation precision.