Research On Technology Of Navigation And Visual Odometer For Mobile Robot Under Indoor Dynamic Environment

Mobile robots are widely used in storage,restaurant,monitoring and other indoor sences.Autonomous navigation is the foundation for completing other complex tasks.Autonomous navigation mainly includes simultaneous localization and mapping(SLAM)and path planning.Over past few years,most autonomous navigation systems based on visual SLAM algorithm assume the environment is static,ignoring the existence of dynamic object will lead to problems such as increased tracking error,decreased accuracy of location,and non-shortest length of the path.Therefore,the thesis of the autonomous navigation system suitable for indoor dynamic environment has very important theoretical significance and application value.First of all,in this thesis,the mobile robot autonomous navigation system designed which is suitable for indoor dynamic environment are introduced.The commonly used sensors and related models involved in it are discussed,and the problems faced by visual SLAM and path planning in the indoor dynamic environment are analysed.Then,a linear extended visual odemetry method to solve the problems that the presence of dynamic targets leads to the increase of tracking error and the decrease of positioning accuracy of navigation system are proposed.The dynamic feature points on the dynamic object are identified and eliminated through the optical flow values and predict optical flow values calculation and depth differences between two adjacent frames,and the static linear expansion is found according to the collinear relationship of remaining static feature pointst.At the same time,the camera pose is estimated according to the static point feature and static line feature.The experimental results show that the static line expansion visual odemetry proposed can effectively reduce the influence of dynamic objects,at the same time,avoid the influence of dynamic objects and the increase of tracking error caused by too little static feature points,the accuracy and robustness of the autonomous navigation system under the dynamic environment are improved.Next,a hybrid route planning method which combines global path planning with local path planning to solve the problem that the single path planning algorithm can not adapt to the complicated and changeful environment when mobile robot works in dynamic environment are proposed.In the part of global path planning,the repair operator and the deletion operator are introduced into the conventional genetic operator to repair the path intersecting with obstacles and delete unnecessary nodes.At the same time,path coherence and security constraints were added into the fitness function to solve the incoherent planning path in the process of path planning and the global path too close to the obstacle.In the part of local path planning,the scrolling window seeks for local target point through heuristic search to complete the obstacle avoidance.After the obstacle avoidance completed,a preview mechanism is introduced to determine the heading angle of the global path and return to the global path to continue driving.Experimental results show that the algorithm proposed in this thesis has certain advantages in path length,ath smoothness and path safety under dynamic environment.Finally,the software and hardware platform is built according to the overall scheme of the autonomous navigation system designed in this paper,and the corresponding function package is configured in the ROS module with the algorithm.In the real indoor dynamic and static environment,map building and path planning experiments are carried out.The experimental results show that the mobile robot autonomous navigation system designed in this topic is effective.