Research On SLAM Of Autonomous Mobile Robot System In Unstructured Environment

Localization and map construction become the focus of robot system on SLAM research.At present,many scholars have studied the algorithms of localization and map construction at home and abroad,most of them leverage sensors such as odometer,laser range finder and visual camera.However,due to the noise effect of the sensor,unstructured environment with uncertainty and the defect of algorithms,the mobile robot cannot be located and build map accurately.The automatic mobile robot control system for unstructured environment is designed to improve the accuracy of robot self-localization and mapping.Meanwhile,based on SLAM,the algorithm which can solve the problem that error accumulation is too serious to make the map matched and localization is not accurate on indoor 3D map building is investigated in this paper.Firstly,an improved ORB feature extraction algorithm based on quadtree encoding is proposed in this thesis.The image pyramid is built to make the scale invariance.Then the feature points are extracted on each image pyramid and quadtree is introduced to homogenize the feature points,which can solve the problem that the detected feature points are too dense to show the picture information completely in unstructured environment.Experimental results show the effectiveness and accuracy of the proposed method.Secondly,based on EPnP,a camera pose estimation optimization algorithm is proposed to make full use of spatial points and reduce the effect of noise on camera pose estimation process.The camera pose is estimated using EPnP.Then the camera pose and positions of spatial points are regarded as optimization variables to adjust based on least squares method,which can improve the accuracy of camera pose in unstructured environment.Thirdly,particle filter algorithm is used on laser radar and vision fusion to solve the problem that error accumulation is too serious to make the map matched and localization is not accurate on indoor 3D map building,which can improve the effectiveness in SLAM process.Finally,an experimental robot and the SLAM system platform for unstructured environment are designed and achieved.The robot is used as an experimental tool to extract the feature points indoor environment,and to compare the proposed algorithm in this paper and others.Experimental results show the effectiveness of the improved ORB feature extraction algorithm.Meanwhile,the pose estimation experiment is in progress on simulated environment.Experimental results show the accuracy and robustness of the proposed pose estimation optimization algorithm.The algorithms proposed in this paper can be used as effective methods in the field of computer vision and the research on SLAM,which are significant for the research on cooperative control and autonomous delivery.