The Graph-Based Simultaneous Localization And Mapping Of Unmanned Ground Vehicles

Being able to build a map of surroundings and to simultaneously locate using this map is an indispensable ability for UGV navigating in unknown environments. This so-called Simultaneous Localization and Mapping (SLAM) problem has became a prevalent research topic of mobile robotics for the last two decades and effective approaches for solving this problem have been proposed. These researches have pushed forward the developing of SLAM.Although a GPS is fixed in the UGV platform, the shadows of buildings and trees result in the poor performance of GPS from which we can’t acquire valid localization. In general, a partial pose estimation method is needed to solve this problem.In this paper, both dead reckoning based on photoelectric coder and visual odometer based on stereo camera are proposed to hit the mark that can estimate the pose of UGV in the case of GPS-Independent.In this paper we formulate SLAM problem by using a Graph structure whose nodes stand for poses of the UGV at disparate points in different time and edges indicate constraints between poses. Specifically, the former are obtained from odometer module and the latter are acquired from observing system of stereo vision at the same time. With the exploring of UGV in unfamiliar environment the graph is constructed step by step. Once such a graph is constructed, the optimal pose estimation that makes global error minimize can be acquired by using the popular Gauss-Newton Algorithm to find the configuration of poses that is mostly consistent with the observing system. Thus, the result can offer accurate localization and direct UGV’s autonomous navigation.To verify the proposed approach’s effectiveness and practicability, a sequence of experiments based on the UGV test platform have been conducted in the campus of DLUT. Numerous results and related data analysis show that odometer module can offer available UGV’s pose prediction and stereo vision observing system can afford exact UGV’s pose observation. On this base, Graph-based SLAM algorithm accomplishes optimal pose estimation which can offer accurate localization and direct UGV’s autonomous navigation effectively.