Wheeled Robot SLAM in Degenerate Environments under IoT and SDN: Multi-Sensor Fusion Based on Multi-Feature Loop Closure Optimization

This paper proposes a visual–LiDAR–inertial fusion system for wheeled robot navigation in IoT and SDN environments, particularly under GPS-denied conditions. Leveraging real-time data exchange, edge computing, and distributed architecture, the system enables collaboration between local and global SLAM. A matching-weighted fusion method is introduced in the VIO module to adaptively adjust visual–inertial weights via the IoT platform, improving robustness under motion. For global SLAM, a multi-feature loop closure module supports cross-device processing and enhances performance on sparse long-range point clouds. Experiments on the VIRAL dataset and real-world tests demonstrate strong performance, achieving an average APE of 0.081 m, RPE of 0.126 m, and a 3.94 ms reduction in per-frame processing time.