Research On SLAM For Mobile Robot Base On Kinect

Simultaneous Localization and Mapping(SLAM)is an important technology for robot to localize and cognitive environment in an unknown environment.Kinect in vSLAM is widely used for its rich information,accurate description of the environment and low price.The main contents of this paper includes building mobile robot platform,feature points extraction and matching with Kinect depth error constraint,loop closure detection and pose estimation based on the bag of words method,pose recovery method in slam and some related experimental research.Firstly,we build the mobile platform of robot,including mobile platform,hardware circuit and software control.The bottom of the system is based on ARM processor,which is responsible for processing the data from sensors and communication with the host computer;The mobile robot platform is equipped with intelligent speech control to realize speech dialogue and motion control.In the host computer system,ROS is built to realize the share of messages in each node.Secondly,in order to solve the problem that the depth error obtained by Kinect camera with the distance increases,the method of pose estimation based on Kinect depth error constraint is studied in SLAM system.And we increase the number of interior points in the extracted the plane features by using the depth error function and recalculate depth values for plane point according to the plane parameters;We set the weight factor in the posture estimation function according to the depth error function and reduce the weight of remote feature points in the posture estimation.Then,the SLAM loop closure detection will be more invalid and get false loop in similar environment.In this paper,the number of points in the motion estimation and the motion threshold are added to the word bag model to verification candidate loop closure frame.Besides,the chi square check is added to the pose optimization to determine whether the loop constraint is consistent with the whole motion.And also the odometry is used to recovery the positon of robot when it lost its position during moving.Finally,a number of experiments were carried out to verify the design of the mobile platform as well as the SLAM algorithm proposed in this paper.The results of experiment show that the method can improve the accuracy of pose estimation based on Kinect depth information in SLAM process and the accuracy of pose optimization in loop closure detection with similar environment?...