EFM2: An Enhanced Sensor Fusion and Path-Tracking Control Framework for Omnidirectional Automated Guided Vehicles

Omni-directional robot platforms offer enhanced flexibility, maneuverability, and adaptability, making them suitable for diverse applications. This study introduces and evaluates a novel sensor fusion algorithm, Extended Fusion Method 2 (EFM2), designed to integrate inputs from a Magnetometer, Camera (with SURF/SIFT object detection), and Lidar (SLAM-based) to achieve fast, accurate, and reliable navigation. The EFM2 algorithm was assessed through both simulation and real-world experiments across multiple path geometries, including continuous and cut paths, as well as industrial-like rectangular tracks. To provide a comprehensive performance evaluation, a Weighted Overall Performance Index (WOPI) was developed, combining conventional metrics velocity, accuracy, stability, and task success into a single holistic measure. Monte Carlo simulations were employed to account for variability in sensor behavior and path conditions, enabling robust ranking of sensor configurations. Results demonstrate that the optimal configuration Magnet at speed 10, Camera at speed 5 with SURF, and Lidar at speed 10 achieves the highest performance in both simulated and real environments. The study further shows that sensor fusion consistently outperforms single-sensor setups, with EFM2 providing superior balance between speed, accuracy, and reliability. These findings validate the effectiveness of the WOPI framework and EFM2 algorithm for advanced, performance-aware navigation in dynamic robotic applications, while highlighting directions for future improvements, including learning-based weight optimization and integration of additional sensing modalities.