Research On The UAV Localization Based On The Low-cost And Micro Sensors

Unmanned Aerial Vehicle(UAV)has become enormously popular in military and civilian areas for its low-cost,no casualties,powerful viability and good flexibility.Navigation and localization have become one of the most important performances for UAVs.For large UAVs,there are many types of navigation technologies,such as Global Navigation Satellite System/Inertial Navigation System(GNSS/INS),National via Signals of Opportunity(NAVSOP)/INS.For micro-UAVs,however,as the limitation of their payloads and prices,the common uses of navigation models are small and cheap GPS models,whose signals are unstable when the UAVs are used in standalone soldier tasks,rescuer tasks,environments discovery and express delivery.For these application areas,navigating based on GPS is not enough.In order to improve the adaptability of UAVs,autonomous navigation technology based on MEMS sensors has become one of the most problems needs to be solved.In this thesis,based on Robot Operating System(ROS),micro-UAV autonomous localization framework,MEMS sensors precision analysis,multi-sensor fusion algorithms and landmark assisted refinement approaches were researched.Simulation and flying experiments were conducted.The main achievements and progress are listed as follows:(1)Based on Robot Operating System,a low-cost,easy-get and integration autonomous localization framework was proposed.According to the special task domains of microUAVs and the localization technologies overview,an autonomous localization framework was explored based on MEMS sensors(accelerometer,gyroscope,pressure sensor and magnetometer).The software architecture was designed based on ROS to improve the localization accuracy and adaptability in multi-task environments.(2)The MEMS sensors characteristics in different motion conditions were analysed,and a series of accurate estimation methods were designed.The feasibility of MEMS accelerometer,gyroscope,pressure sensor and orientation sensor were explored and boundary conditions of using the MEMS sensors for localization were researched,which can estimate the MEMS sensors characteristics in different altitude and acceleration statuses and improve the validity and practicability of sensors measurements.(3)Considering MEMS sensors characteristics and motion of the UAVs,multi-sensor integration algorithms were investigated to extend the localization domain from indoor spaces to outdoor spaces,and improve the adaptability in weak illumination environments.A maples visual odometry(MLVO)method was proposed based on camera,pressure and orientation sensor for UAV uniform linear,up and down movements.Based on the errorstate extended Kalman Filter,a multi-sensor fusion visual odometry(MFVO)algorithm was presented by integrating the MLVO,acceleration and gyroscope measurements.The two methods supplement each other to improve the localization accuracy.(4)To solve the problem of drift errors in autonomous localization,a landmarkassisted visual odometry(LAVO)was proposed by integrating web-based map and artificial masks.The obvious buildings were selected as beacons and their absolute positions were used as prior,the UAV location probability map was built based on a probability filter framework.As the increase of the observability information in on-board images,the location evidences were accumulated to eliminate the drift errors.(5)The simulation environment was built based on the Gazebo simulation,and a Samsung Galaxy S3 smartphone was selected to test the accurate performance of the algorithms.In Gazebo simulation environment,different flying experiments were designed and MLVO,MFVO and LAVO performances were analysed.In real environments,the thesis selected MOVEMASTER RM501 robot arm,AR-Drone,XAircraft650 drone as the main test platforms,and a Samsung Galaxy S3 as the on-board sensor.Several phone application programs were developed to access the built-in sensors.Outdoor flying experiments in the real worlds were conducted and compared with TUM_PTAM[1].According to four flying experiments in indoor,corn fields,garden and city environments,the adaptability of the proposed algorithms were tested.As the results shown,the localization system can estimate the position accurately in indoor(<0.5m)and in weak illumination outdoor environments(<10m)by exploring the phone MEMS sensors.All the above achievements improve current researches on UAV localization technology in a certain sense,and provide a new solution for solving long term UAV flight.On the other hand,the codes and date of the algorithms will be open source,in order to provide more research resources for others.