Research On Positioning Technology Of Vision And Inertial Navigation Information Fusion For Coal Mine Mobile Robot

The positioning technology of mobile robots in coal mines is a key technical issue that directly affects the performance of coal mine robots.Due to the absence of GPS and BDS positioning signals in coal mines,complex work scenarios and serious shielding of signals by coal walls,it is difficult to apply ground mobile robot positioning technology in underground mines.Aiming at the positioning problem of mobile robots in coal mines,a positioning method based on information fusion of monocular vision and strapdown inertial navigation is proposed.In this paper,the calculation principle of the monocular camera and the basic positioning principle of the strapdown inertial navigation system are studied and analyzed,the positioning error model of the monocular camera and strapdown inertial navigation is established,the parameter calibration experiment of monocular camera and strapdown inertial navigation is completed,The internal and external parameter matrices of the monocular camera and strapdown inertial navigation are obtained,the tightly coupled method is used to initialize the combination of two sensors,which lays the foundation for further fusion.Aiming at the problem of the mismatch between the monocular camera and the strapdown inertial navigation data,the inertial navigation data pre-integration method is adopted.After the inertial navigation data is pre-calculated,the inertial navigation data between two key frames is integrated.Finally,the data is transfered to the visual coordinate system by coordinate system rotation,the preprocessing and format matching of sensor data are completed.Aiming at the problem of visual positioning under low light conditions in coal mines,an improved ORB feature extraction method is proposed to increase the number of feature points extracted,and an adaptive threshold method is added to enable monocular vision to be performed in areas with weak illumination and large illumination changes.For accurate positioning,the RANSAC feature matching algorithm is selected to filter out mismatched feature points.Through the experiment of feature extraction and matching in low-light areas,compared with traditional visual positioning algorithms,it is verified that the improved method can have higher positioning accuracy in low-light areas.Aiming at the optimization problem of the combined positioning of strapdown inertial navigation and monocular camera,the tightly coupled data fusion method based on graph optimization is improved,and a sliding window is added as a constraint,and the range of each data fusion is limited to a sliding window to reduce Optimized the calculation amount of positioning and increased the real-time performance of positioning.Loop detection is added to the back-end optimization,so that the robot is in cyclic motion,reducing the accumulated error and improve positioning accuracy during cyclic motion,.Through simulation experiments,it is proved that the method has higher positioning accuracy and better performance.Finally,the public data set is used to test the combined positioning method proposed,and it was compared with the traditional combined positioning method to test the performance of the combined positioning method.Our team’s underground coal mine mobile robot platform was used to test the combined positioning method proposed in this article,and the algorithm is compared with ORB-SLAM2 to detect the positioning accuracy of this method.The results show that the combined positioning method which was proposed in this paper has good performance and positioning accuracy in simulating underground coal mine environment.