Research On Multi-sensor Combined Positioning Technology Based On SLAM

Autonomous Navigation of robot in unknown environment is a key link to realize robot intelligence.Using sensor information to locate robot in unknown environment and establish map are the key technologies to realize robot autonomous navigation.In order to adapt to a variety of application scenarios,robots usually carry a variety of sensors,such as cameras,lidar and inertial measurement unit(IMU).How to effectively combine multiple sensors to achieve high-precision positioning has become the research difficulty of robot positioning and navigation.In order to effectively combine a variety of sensors,a combined positioning system of camera,inertial navigation and wheel tachometer based on extended Kalman filter is designed.Firstly,the front-end visual odometer is studied,the time and effect of different feature points are analyzed,and the Harris corner with the best real-time and stability is selected.Aiming at the time-consuming problem of calculating descriptor in feature point method,sparse optical flow method is selected for feature tracking to effectively improve the real-time performance of visual odometer.Finally,the pose of IMU and wheel tachometer is used as the filter prediction input,and the pose of camera is used as the observation input to realize stable realtime pose output.In the loose coupling framework using filter,the utilization rate of sensor data is low and the positioning accuracy is not improved.The nonlinear optimization method based on sliding window is further used to fuse the original data of camera and inertial navigation.According to the IMU pre integration theory,the noise covariance recurrence matrix and residual function are established,the re projection error is used for cluster adjustment,and the visual residual function is established.The error functions of the two kinds of sensors are jointly optimized in the sliding window to obtain the global optimal pose.In addition,this thesis introduces the elimination and uniform distribution of feature points and noise to ensure the stable operation of visual odometer.It also ensures the perfection of the system from the aspects of ROS system program design and the calibration of camera and inertial navigation.Finally,the experiment of vision inertial navigation wheel tachometer positioning system is completed by using ROS(Robot Operating System)mobile car in indoor environment.The experimental results show that the positioning effect of the system is good in complex environment.The visual inertial odometer is tested in the Eu Roc public data set and indoor environment.The error between the evaluation and the true value shows that the integrated positioning system has higher stability and better robustness.