| Without environmental priors,by collecting the information obtained by the sensors mounted on the robot,the robot’s pose is calculated and the map is created in real time.This process is SLAM(Simultaneous Localization and Mapping).Pure visual SLAM cannot meet the increasingly complex localization needs.The introduction of IMU(Inertial Measurement Unit)can make up for the lack of low-frequency of the camera,and effectively solve the robot’s pose estimation under fast motion.However IMU will cause the accumulative drift problem after running for a long time.In order to solve these problems such as the camera’s susceptibility to occlusion,the camera moving too fast,the image blur,the low accuracy of the algorithm under the pure rotation of the robot,and the accumulative drift of the IMU,a SLAM algorithm fused with the Multi-Camera and IMU is researched to realize the localization problem of outdoor mobile robots.The research content and innovations of this article mainly include the following aspects:(1)An outdoor mobile robot platform for localization was built,and the work included hardware selection and software architecture design.The system design of multi-sensor fusion is completed,including the establishment of each hardware,interface operation,data transmission and time stamp unification.In this system,the four cameras and IMU are responsible for localization,and the laser radar is used to obtain the real trajectory(groundtruth)of the robot movement.(2)In order to solve the problem that time of each sensor is not synchronous caused by different frequency between cameras and IMU and different exposure time between cameras,an FPGA hardware triggering method was used,and an exposure compensation module based on exposure compensation was proposed innovatively.Experimental results show that these methods can achieve the time synchronization between the four cameras and the IMU,and the synchronization accuracy is about 1ms.(3)The external parameter calibration of the four cameras and IMU is completed.One is the external reference calibration of binocular camera and IMU,and the optimal result is selected after 8 groups of experiments are carried out on the handheld calibration platform.The second is the external parameter calibration of the four cameras,and the optimal result is selected after two calibration experiments for each two adjacent cameras.The accuracy of the calibration results is verified by the visual reprojection errors in the calibration process.(4)Based on the extension of VINS framework,a SLAM algorithm for the fusion of Multi-Camera and IMU is proposed.This algorithm supports the four cameras and IMU fusion system built in this paper.The front end adopts the method of combining dual eye flow tracking and two single eye flow tracking.The system uses a static initialization method of IMU.The visual residuals at the tight coupling back end are defined as the form of single point residuals plus double point residuals,and the external parameters of multiple cameras are added for online calibration.Experiments on the Multi-Camera and IMU localization platform show the effect of initialization,verify the effectiveness of the proposed algorithm,and give the calibration results of external parameters of multiple cameras in a single localization process.(5)The outdoor mobile robot platform was used to carry out relevant experiments.In order to test the localization performance of the proposed algorithm,seven comparative experiments,such as linear,circular and large loop,are carried out to verify that the localization accuracy of the proposed algorithm is better than that of VINS algorithm.In addition,the localization performance of the proposed algorithm is compared with Open VINS,an existing multi-camera and IMU fusion algorithm.The experiment also shows that the localization performance of the proposed algorithm is better.Aiming at the camera occlusion problem,the camera occlusion experiment was carried out.Through four different data sets,such as the straight line segment and the cross segment,it was verified that the arrangement of multiple cameras could make up for the decreased localization accuracy caused by camera occlusion and improve the robustness of the system. |
