Intensity-SLAM: Intensity Assisted Localization and Mapping for Large Scale Environment

Simultaneous Localization And Mapping (SLAM) is a task to estimate the robot location and reconstruct environment based on observation from sensors such as LIght Detection Ranging (LiDAR) camera. It widely used in robotic applications autonomous driving drone delivery. Traditional LiDAR-based SLAM algorithms mainly leverage geometric features scene context, while intensity information LiDAR ignored. Some recent deep-learning-based consider train pose estimation network an end-to-end manner. However, they require significant data collection effort their generalizability environments other than trained one remains unclear. In this paper we introduce system. propose novel full framework that leverages both geometry features. The proposed involves intensity-based front-end odometry back-end optimization. Thorough experiments are performed including outdoor indoor warehouse manipulation. results show method outperforms existing geometric-only methods.