Large-Scale Navigation Method for Autonomous Mobile Robot Based on Fusion of GPS and Lidar SLAM

Abstract

A function-well robot should accommodate the indoor and outdoor environment, which requires the perception system to be more robust, and the localization should keep good accuracy. In this paper, we focus on the localization problems. we try to propose a fusion method to combine the RTK-GPS with the lidar based SLAM method to design a localization mechanism to make a four-wheel-independent-drive mobile robot run free to the environment. To the SLAM(Simultaneous Localization And Mapping) method, we develop a method based on Cartographer. However, Cartographer was designed to locate in door, not for the outside environment. A main constraint is the loop closure, which is difficult to complete in the open area because of the large-scale environment. Therefore, we propose a dynamic sub-maps switching method. To avoid the large memory occupancy, our approach first divides the whole Cartographer map into several small piece of maps, and then assemble them into a global map when loading. To the whole localization system, a complementary filter was designed to combine the GPS, Cartographer, and the Dead-Reckoning. We evaluate our approach on our self-designed wheeled robot. The results show that the fusion approach can accommodate the indoor, outsides, tunnel and other scenarios with a high accuracy.