A Novel Multi-Sensor Fusion SLAM Framework for Anti-Interference

Nowadays, people have high requirements for the robustness of autonomous driving. As a key technology in the field of autonomous driving, SLAM requires a lightweight and explainable algorithm framework when dealing with perception degradation scenes. At present, most ‎multi-sensor fusion SLAM algorithms use methods such as training neural networks or adding new odometry constraints to achieve anti-‎interference, but these methods do not meet the requirements. We use the data communication technology of the ROS (Robot Operating ‎System) platform to integrate the CLAHE (Contrast Limited Adaptive Histogram Equalization) algorithm, the LiDAR Noise Filtering algorithm, and RTAB-Map (a multi-sensor fusion algorithm based on graph optimization). We then conducted experiments in three designed ‎perception degradation scenes and used the MME indicator to quantify the experiments. The results showed that the MME of our framework in the perception-degraded environment was reduced by an average of 0.107, proving that the framework we proposed performs bet-‎ter than RTAB-Map in dealing with some perception degradation scenes.