Design And Experiment Of UAV Obstacle Sensing System Based On Multiple Sensing Data

Small-size unmanned aerial vehicles have light weight,stable,reliable,low-altitude low-speed flight characteristics.There are many applications in the agricultural fertilizer pesticide spraying,remote sensing and other fields.Obstacle avoidance is an important means to achieve UAV safety flight.Enhance of perception of obstacles,is conducive to improving the safety of UAVs.This paper mainly aims to improve the ability of UAV obstacle perception through multi-source data fusion.It is used for obstacle scanning perception of UAV in low visibility.Ultrasonic and laser sensors are used as unmanned aerial vehicles obstacles to the means of obstruction.For the ultrasonic sensor,the pretreatment test was carried out to calibrate the ultrasonic sensor.The ultrasonic sensor was mainly tested for the measurement distance,divergence angle,effective detection angle,and the basic parameters of the ultrasonic sensors obtained.Based on the ultrasonic sensor,the obstacle sensing system was designed.With ultrasonic sensor application problems encountered in the small UAV flight,in different throttles of UAV,the vibration of the space and spectrum distribution,the influence of UAV in different throttles and pitch angles on relative error of ultrasonic sensor are tested.The effect of different scanning methods and data acquisition methods on the validity of the data is tested by the pretreatment test,the influence of the measurement stroke and the different acquisition frequency on the measurement error.The method of data fusion is used to improve the performance of UAV obstacle detection system.In the spectrum distribution,the vibration of small UAV is mainly in the three frequency bands,0-1.5KHz,8-9KHz,16-18 KHz.As the frequency increases,the amplitude of vibration decreases.In the spatial distribution,along the arm extension direction,with the distance from the UAV center distance increasing,the vibration appears strengthened and weakeneddistricts.As the distance increases,the vibration amplitude appears sliding down trend;With the throttle increases,ultrasonic sensor ranging error increases;As the pitch angle increases,the distance error of the ultrasonic sensor rises relatively.When the sensor sensing direction is parallel to the ground,there existsthe smallest relative error.In the case of perpendicular to the ground,the relative error is the largest.In the obstacle sensing experiment with ultrasonic sensor,the relative error of obstacle depth sensing is less than 4%.The test of the laser sensor measurement shows that estimationis linear to the real distance.The acquisition frequency of 5Hz is the most stable between different acquisition frequencies.The designed obstacle sensing system is different for pitch scanning and vertical scanning.In the case of pitching angles,the scanning trajectory will produce distortionvarying degrees,and the vertical scanning is,theoretically,not distorted.But the vibration caused by gimbalhaseffectson vertical scan data during the rotation.In different scanning modes,synchronous scanning is better than asynchronous scanning.In different acquisition frequency of asynchronous scanning case,the higher the acquisition frequency,the outline of the obstacle details are more obvious.The Kalman filter algorithm is used to fuse the ultrasonic and the laser data.The fusion mainly is applied to use ultrasonic data to optimize the laser data on the XY axis.The accuracy of the three-dimensional point cloud is enhanced.