Design Of Autonomous Unmanned Aerial Vehicle With Multi-sensor Fusion

In recent years,the technology of unmanned aerial vehicle(UAV)has developed rapidly.With simple structure and the ability of vertical take-off and hovering,it has promising potential of application for industry,agriculture,military,etc.As an aerial robot,high precision,intelligence and capability of autonomous flight should be required to meet the requirement of the application in various fields.Therefore,accurate state estimation and environmental perception of the UAV should be studied.First of all,the paper starts with the state estimation of the UAVs.Conventional UAVs use Global Positioning System(GPS)for localization,which makes them impossible to accurately estimate the position in a GPS-denied or poorly-lit environment,while UAV's ability of stable hovering is based on its accurate estimation of the position.Based on different sensors,including inertial measurement unit(IMU),GPS,barometer,laser ranging sensor,laser scanner,optical flow and stereo camera,this paper first analyzed the signal characteristics of different sensors,and then optimized the positioning algorithm for laser scanner with fusion of IMU and the velocity information.And then a multi-sensor fusion framework based on Extended Kalman Filter(EKF)was designed to accurately estimate the position and velocity of the UAV in the vetical and horizontal directions.Secondly,in order to meet the demand of the application in regional operation flight,this paper designed a regional trajectory planning method based on spiral spline.First a method was put forward for generating control points in a given region,and then the control points were fitting with NURBS curve.Based on the kinematic constraints of the UAV,the trajectory was optimized by establishing an optimization problem with the shortest flight time.And trajectory interpolation was used to generate the final flight trajectory.An optimal trajectory was finally generated.Such method can effectively improve the efficiency of regional operation flight tasks.Finally,to avoid the obstacles automatically during the process of autonomous flight,this paper came out an improved artificial potential field method for obstacle avoiding.And a rasterization method was designed to extract the obstacle information based on the scan data,with the real-time obstacle data from sensors,the UAV can realize the obstacle avoidance flight in real time under the unknown environment.And a simulation platform was established in Gazebo,and the obstacle avoidance algorithm was effectively executed in the simulation environment.Finally,experiments were carried out and the algorithm was effectively executed in the real time flight.