Calibration-Free Vision-Guided Localization Method for Refueling Robots Using Adaptive Kalman Filter

To address the localization challenges arising from arbitrary stopping poses of refueling vehicles, this paper introduces a calibration-free vision-guided method based on adaptive Kalman filter. First, it establishes a vision-guided localization framework using a calibrated camera perspective model. Then, an adaptive Kalman filtering algorithm is incorporated to eliminate calibration dependencies, enhancing the system's flexibility and robustness. After validating the method through MATLAB simulation, we construct an experimental refueling robot platform. Experimental results demonstrate the localization error of end-effector less than 2 mm, meeting the precision requirements for automatic refueling processes. This research provides a novel approach to precise guidance for random vehicle poses.