Research On Autonomous Obstacle Avoidance Methods For UAVs Based On Multi-sensors Fusion

The autonomous obstacle avoidance technology of unmanned aerial vehicle(UAV),that is the "sensing and avoidance",is one of the hot research topics in the field of UAV for the whole world in recent years.During the development of UAV,both the 3D environment perception,multi-sensor information fusion,the three dimensional environment construction and the three dimensional path planning,UAV controlling and so on several aspects are researched respectively by the author of this paper.After perceiving and detecting the environment information through the binocular vision technology combined with 2D laser radar scanning technology,which including the following process such as the binocular image calibration,matching and reconstruction technology for 3D environment space,the autonomous flight of obstacle avoidance of UAV could been realized.The method which is proposed that the evolutionary computation and the ant colony optimization are respectively applied in the threedimensional obstacle avoidance path planning for the UAV especially suitable for model in view of the city community building environment.The good results are shown by computer simulation.The four rotor UAV prototyping system for obstacle avoidance has been designed for airborne online real-time data procession acquired by binocular vision and the lidar,by which a large number of targeted research has been completed about the obstacle avoidance in the city of building environment space for UAV.In this paper,the development of the obstacle avoidance for four rotor UAV due to the combined with binocular CCD,the 2D laser radar,ultrasound,airborne IMU,GPS and so on sensory detection,navigation and positioning sensors,thus which can satisfy the different scales environment space such as indoor and outdoor,city and in the wild for UAV's autonomous obstacle avoidance flight,the intelligent UAV as practical platform is applied in the following field such as military counterterrorism,street fighting,intelligence detection in the battlefield,as well as civil parcel delivery,electric inspection,natural disaster patrol and environment remote sensing etc..In this paper,it is needed to obtain three-dimensional coordinates of obstacles for the optimal flight path planning which the UAV avoids obstacle in measurement environment by a binocular CCD.Through imaging geometric model and binocular vision measurement principle to calculate,and through Open CV tools the software was secondarily developed,which realizes processing,calibration,matching,measuring of binocular images after acquired using the network CCD and reconstruction on the environment after obtained internal and external parameters of binocular CCD,at the same time the measurement error is analyzed.A kind of 2D laser radar calibration method is presented,and the error analysis and contrast was finished by mixing calibration with CCD about the two kind of measuring methods.In this thesis,it is needed to obtain three-dimensional coordinates of obstacles for the optimal flight path planning which the UAV avoids obstacle in measurement environment by a binocular CCD.In this paper the matching accuracy of matching feature points in two matching image is guaranteed through polar line correction,dense matching method of the binocular CCD.Based on triangulation method threedimensional coordinates of obstacles are calculated in the three-dimensional space environment.By the method of information fusion,the depth information of the radar fuses to the images acquired by binocular CCD,and disparity estimation for the radar is obtained through fusion.At present,because the path planning problem for the UAV in three-dimensional space is still in the researching stage,especially which is by intelligent method to calculate the three-dimensional planning path for UAV.In this paper,the method is proposed for the UAV's obstacle avoidance and threedimensional path planning especially suitable for model in view of the city community building environment based on the evolutionary computation and the ant colony optimization.The algorithm process and satisfied computational simulation results are given through calculating the math model which was built based on the constraint condition of the real city community courtyard building,and evaluating cost for avoiding dangerous obstacles which is under the condition of the approximately real building and ignoring the terrain elevation.The different simulation results of the same algorithm is compared by changing the parameter of environment model.Through the simulation results,under the condition of the same model,especially in the city building barrier environment of discretely distributed danger,evolutionary algorithm has more advantages,and the ant colony optimization algorithm is more adapted to the continuous variable height conditions such as wild mountain terrain environment.Based on the analysis for four rotor dynamics modeling and the motion analysis,the attitude and position controller were designed successively and the controlling of the UAV was realized by PID controller.The four rotor UAV prototyping system platform for obstacle avoidance has been designed for airborne online real-time data procession acquired by binocular vision and the lidar,and a great deal of flight experiment was carried out.Results show that the proposed theory algorithm and design method of UAV autonomous obstacle avoidance are effective and feasible,which provides strong reference for the UAV applied research under the condition buildings in the city environment space.