Research On SLAM Based On RGB-D Sensor For Mobile Robots

When a mobile robot navigates in an unknown environment,it has to build a consistent map of this environment according to its current pose.At the same time,the mobile robot should determine its self-location within this map.The problem mentioned above is named as SLAM(Simultaneous Localization and Mapping).In this paper,we will use RGB-D sensor in the robot SLAM,to improve the loop closure detection,so the robustness of robot navigation system can be improved.Firstly,the state-of-art approaches of robot SLAM based on visual sensor are summarized.And to meet the need of long-term auto-navigation in large area,RatSLAM,as a method inspired by models of mapping and navigation in the rodent hippocampus,is chosen as the SLAM system to study.RatSLAM,as a vision-based system,uses a single camera as its sensor,which causes the low accuracy of the loop closure algorithm,especially in bad illumination situations.And the random sliding window algorithm is proposed to realize loop closure detection based on RGB-D information,so the influence of image noise can be reduced,and the accuracy of loop closure detection can be improved.Finally,a series of indoor experiments are conducted based on the Nubot mobile robot designed by our lab.The results show that the accuracy rate and success rate of the RatSLAM are improved after using RGB-D information.