Research On Collaborative Navigation Technology Of Miniature Air-ground Unmanned System

With the improvement of military and civil needs,a single robot has been unable to complete tasks requiring too much complexity and accuracy due to the limited working range,single sensor type,incomplete functions and other factors.Therefore,multi-robot collaboration technology arises at the historical moment.Based on the background of military operations and civil rescue,this paper studies a collaborative navigation system for a small unmanned air-ground platform.By combining the advantages of large-scale environment perception of UAV and local precise positioning of UGV,the UAV can observe and identify targets in advance,and then the UGV can navigate to the target autonomously.The main work of this paper is as follows:(1)In this paper,the architecture of the collaborative navigation system for air-ground platform is designed.In terms of hardware architecture,the four-rotor wing is selected as the UAV in this paper,and its body structure and flight control module are designed.Wheeled robot is selected as the UGV in this paper,and the body structure and layout of the car are designed;At the same time,the design of ground monitoring station and communication system is completed,and the hardware selection of each sensor of the system is carried out.In terms of software architecture,the development environment of Linux operating system and ROS development platform is adopted.Combined with the system functions,the image acquisition module,image processing module and autonomous navigation module are designed respectively,and the message transmission interface of each module is designed.(2)In this paper,a lightweight and real-time global map construction method based on image processing is proposed.The method takes advantage of the UAV’s global perspective and integrates various image processing methods to realize the construction of global raster map.Among them,KSW entropy threshold segmentation method is used to complete the detection of the obstacle area,the method of pixel gray scale accumulation is designed to complete the global map rasterization,and the method of square variance template matching and auxiliary marker is used to complete the scale calculation.The whole global raster map is generated in real time by image processing,which solves the problem of large computation amount building global map based on octree model and the problem of high time cost of building global map based on Bayesian filtering,and finally provides navigation map for collaborative navigation.(3)In this paper,we design a global path planning algorithm in the collaborative and air-ground environment.A method of visual sign detection and odometer information fusion is designed to estimate the real-time position and pose of ground vehicles,which provides a starting point for path planning.The feature matching method is used to locate the target in real time,which provides the terminal point for path planning.In this paper,an improved A*path planning algorithm is designed.By adding the offset on the heuristic function,it solves the problem of extending useless nodes in the search process for the traditional A* path planning algorithm and improves the efficiency of path planning.The global path planning algorithm can be applied to dynamic scenes and moving targets,and it provides an optimal path for collaborative navigation.(4)Finally,this paper designs the system performance indicators according to the functional requirements of the system,and built a simulation platform based on Gazebo,respectively in the static and dynamic scenarios to verify the system indicators,all performance meet the requirements.Then the physical experiment test is carried out,the hardware platform is built,the experiment scene and scheme are designed,and the navigation experiment under the real scene is completed.The final system image acquisition rate is 25 Hz,the global map refresh rate is 2Hz,the ground co-positioning accuracy is 4.6d %,the maximum linear velocity of the unmanned vehicle is 1.25m/s,and the maximum angular velocity is 1 rad/s.The navigation accuracy is 0.18 m,the success rate of navigation is 93.3%,all performance meet the indicators,the system research and the system design are completed.