Research On Multi-uavs Autonomous Cooperation SLAM Based On Adaptive Data Associatioin

With the continuous development of UAV technology,its characteristics of low cost,high flexibility and less space constraints provide more possibilities for solving various practical problems.So UAV is widely used in environmental detection,agricultural plant protection and other fields.In this dissertation,an unknown environment exploration situation is considered.In the case of global positioning system limited,Simultaneous Localization And Mapping(SLAM)algorithm can provide key technical support for UAV autonomous flight and navigation.Due to its limitation of performance and exploration range.In large-scale and complex environment,single UAV cannot complete high-precision detection mission quickly and reliably.Therefore,this dissertation mainly considers a multi-UAVs cooperative SLAM system,and proposes an adaptive data association algorithm framework for the problem of information interaction between multi-UAVs.Finally,an active SLAM algorithm for UAVs is proposed.Aiming at the problems of SLAM,this dissertation proposes a multi-UAVs cooperative SLAM framework.Each UAV is equipped with depth camera,micro-computer and communication module.It can execute SLAM algorithm and interact with the surrounding UAV nodes in the flight process.It can solve the problem of localization and sub map construction in unknown environment.So this dissertation proposes a low bandwidth requirement of multi-UAVs data exchange.And it can also calculate the relative pose of similar scene by using bag of words,random sample consensus(RANSAC)and iterative closest point(ICP)algorithm.This part provides basic data set for subsequent data association.Based on the multi-UAVs cooperative SLAM framework,an adaptive data association algorithm is proposed to solve the problem of perceived aliasing in data association.According to the data of SLAM and similar scenes,a new data structure called relative initial pose is proposed.Based on the relative initial pose set,this dissertation processes the adaptive number clustering.The cluster with the highest probability is selected according to the clustering selection algorithm.If it passes the threshold verification,they are considered as the same scene,otherwise are different scene.Finally,a global map composed of multiple sub maps is obtained.It proves the effectiveness of the algorithm in solving the problem of perceptual aliasing.In the traditional SLAM algorithm research,a kind of passive SLAM scenario is usually considered.This is inconsistent with the requirements of autonomous environment detection in the actual scene.In order to solve this problem,this dissertation attempts to realize the autonomous exploration of multiple UAVs based on the cooperative SLAM framework.First,the three-dimensional environment space is rasterized by octree map.A multi-UAVs framework with active guidance data association is proposed.Moreover,an improved active exploration coverage method is proposed to make the process closer to reality.Finally,through the experimental verification,this method achieves the balance of exploration and coverage in the autonomous SLAM algorithm.