Research On Collaborative Mapping Technology Of Unmanned System

This topic is based on the intelligent unmanned system,focusing on the construction of multi-robot collaborative maps.The use of robot-mounted visual sensors to quickly build local maps in unknown environments,and the realization of multi-robot map fusion to meet the accuracy and complexity of complex environments.Requirements for realtime,reliability and environmental adaptability.In this paper,a map fusion algorithm based on a dictionary model is designed to complete the multi-robot collaborative mapping task in a complex environment.The unmanned platform is used as the front end to collect map information,and the central control unit is used as the back end to complete the map construction.The tag algorithm is used to optimize the pose and map construction,and the map matching and fusion based on the overlapping area are combined to merge multiple sub-end maps into a global map.The relevant research methods of collaborative mapping in unmanned systems in this paper are as follows:First,the overall framework of the multi-machine system is designed,and a centralized robot system is adopted to divide the system into a sub-end unmanned platform and a server end of the central control unit.The sub-end uses a monocular camera as a sensor to implement a visual odometer.Based on this,the local map part is designed,and a deletion mechanism and a cache mechanism are designed to ensure the real-time map construction of the sub-end.Secondly,a tag-based UAV pose calculation and optimization algorithm is designed.Apriltag is used to detect and understand the original camera pose,and the pose relationship between the camera and the drone and the world coordinate system is used to obtain the drone pose.And the optimization design of map offset is adopted to optimize the position and attitude of the UAV.Thirdly,a map fusion algorithm based on a dictionary model is designed,orb features are used to extract feature points,a dictionary library of pictures is constructed,similarity calculation is used to select matching key frames,and then Sim3 transformation under RANSAC iteration is designed to calculate the matched The posture transformation relationship between the maps,and finally the fusion of map key frames and map points to form a global map.Finally,the overall design of the collaborative system is implemented under the Ubuntu system,and actual,data set and simulation tests and verifications are performed on the system's sub-ends and the system as a whole.Running the system in a laboratory environment,the overall system can complete the sub-end mapping,posture optimization and map fusion functions,realize the multi-robot system mapping function,verify the feasibility and accuracy of the algorithm.