Research On 3D SLAM Based On Distributed Computing

Robot SLAM usually uses laser radar,sonar and other sensors,the amount of information to obtain less and a single form.It is difficult to adapt to complex environments.There are computational complexity and fault-tolerant performance Poor problem due to most of the current SLAM technology using centralized architecture.At the same time,the traditional robot development model is pre-programmed for each specific task.The economic cost is high and the robot can not share the resources which hinders the development and application of the intelligent robot.The combination of robot and cloud computing that the cloud server-side as a development platform,the robot can be obtained from the cloud server platform,knowledge,ability to expand the scope of the same time greatly reduce development costs.In this thesis,3D SLAM based on distributed computing is studied and realize the 3D composition of mobile robot in the indoor environment by relying on the cloud computing platform,which has positive scientific significance and application value.Aiming at the 3D SLAM algorithm in indoor environment,this thesis uses the feature method to build the algorithm flow.In order to improve the efficiency of the algorithm,meanwhileand,to reduce the transmission pressure,a key frame strategy is adopted to solve the closed-loop detection problem by using the bag model.In order to improve the efficiency of drawing and solve the shortcomings of SLAM centralized architecture,this thesis proposes a SLAM architecture based on distributed file system(HDFS).In the aspect of mobile robot navigation,this thesis establishes the dynamic model of mobile robot,and studies Monte Carlo localization algorithm,A*path planning algorithm,dynamic window algorithm of local obstacle avoidance and so on.In order to avoid the 3D obstacle,a 2.5D map model is put forward.Through the improved A*algorithm,an optimal path can be found quickly and applied to the actual indoor robot navigation.In order to verify the effectiveness of this thesis,an experimental system based on Kinect,ROS and iRobot Create robot was consti acted.In order to validate the 3D model of mobile robots,3D automatic 3D SLAM and 3D distributed architecture are proposed.The experimental results show that the three-dimensional mapping,autonomous navigation and distributed computing of the indoor mobile robot are effective and can be used to construct the 3D map of the environment,which can fully show the advantages of 3D SLAM in distributed computing.