Enhancing the Robustness of Mobile Robot Navigation: Fusion Strategies Based on Wheeled Odometry and Visual Estimation

In this paper, we propose a novel sensor fusion strategy that combines AprilTag-based visual localization and wheeled odometry using a simplified continuous Kalman filter for high-precision robot positioning in dynamic environments. Traditional localization methods, relying on either vision or odometry alone, face challenges such as visual sensor limitations in varying lighting conditions and odometric drift over time. By integrating the strengths of both sensor types, our method corrects odometric drift and compensates for visual errors induced by environmental factors. The Kalman filter continuously refines the robot’s position and orientation, ensuring stable and accurate navigation. Experimental validation on a Swerve Drive robot platform demonstrates that the proposed fusion approach significantly improves localization accuracy and robustness compared to using visual or odometric data independently. The results highlight the potential of sensor fusion techniques to enhance autonomous navigation in complex and unpredictable environments, with applications ranging from autonomous vehicles to industrial robots and search-and-rescue missions.