Recent Advances in Multi-Sensor Fusion SLAM for Mobile Robots in Complex Environments

Abstract

Regarding single-sensor defects in complex environments, propose various mainstream multi-sensor fusion methods, to enhance the robot’s positioning capabilities in complex environments. This article studies the sensor fusion technology in recent years, summarizes the algorithms and methods for fusing data from different sensors, and analyzes the optimization approaches taken to cope with various environments. First, compare the respective advantages and disadvantages of visual sensors, LiDAR, and inertial measurement units. Proposed two sensor fusion methods, Visual sensors and inertial measurement units combined with LiDAR Then, VINS-Mono and MSCKF algorithms were introduced to improve the degree of data fusion, and based on multiple sets of algorithm- based optimization methods, the optimization directions in recent years were summarized. It then elaborates on the synchronization method between visual sensors and LiDAR, which is achieved through data-level and feature- level fusion to enhance the system’s robustness. Subsequent research on the fusion methods of new types of sensors can refer to this approach.