An Indoor-Outdoor Continuous Localization Method Based on Multi-Sensor Fusion

As the basis of mapping and navigation of mobile robots, localization technology plays a vital role in the field of robotics. In this paper, a fault-tolerant indoor-outdoor multi-sensor adaptive fusion localization method is presented. This method aims to solve the problem of a single sensor losing signal when a mobile robot is located in indoor-outdoor environments. Firstly, ground point cloud segmentation is applied to the NDT (Normal Distributions Transform) algorithm, which improves the registration speed by nearly 36 % compared to the original method. Then, based on the improved point cloud registration localization method, a multi-sensor adaptive switching fusion algorithm is carried out by combining ESKF (Error State Kalman Filter) and the motion constraints of the robot kinematics model. This algorithm effectively isolates the failed sensors and improves the accuracy and robustness of continuous localization in indoor-outdoor environments. Finally, in order to verify the effectiveness of the proposed method, a long-term localization test is conducted in indoor-outdoor continuous environments. The experimental results show that the proposed method can effectively improve the speed of LiDAR (Light Detection and Ranging) point cloud registration, and ensure localization accuracy and robustness when a single sensor fails.