Depth Information-Based Autonomous Aerial Refueling Relative Pose Estimation And Control Technology Research

The relative pose estimation and control is one of the key points in autonomous aerial refueling(AAR). The depth information-based autonomous aerial refueling relative pose estimation and control technology is proposed. In spite that many research on the key point of AAR are made and project are carried out, there are some defects and weaknesses in some degree, such as overdependeny on GPS, 2D visual sensor being subject to the interfere of weather and so on. The 3D imaging sensor(Flash LIDAR) based on time-of-flight technology can reduce dependence on GPS and other 2D visual sensor and provide a dissimilar redundancy for AAR. According to the condition of AAR research and development in our country, the study of autonomous aerial refueling relative pose estimation and control technology based on depth information is carried out by combining the national key projects of “863”:“development of the high payload UAV for smart geophysical prospecting”. The study involve the feasibility, date process method and error analysis of relative pose estimation, the experimental design about the point clouds of simulation model. On this basis, the special flight control issues related the robustness of relative pose estimation in AAR are also studied, which are how to plan and follow the reference trajectory in disturbance. In addition, the study try to combine the experience of the polit and control technique in relative pose close control loop. The main works of the thesis are as follows:1. The composition form and current research status of relative pose estimation and control in AAR are introduced based on a lot of publications, conference papers and flight test. According to its defects and weaknesses, main works of the thesis are put forward.2. In order to make the massive point clouds of tanker processed in real time, the horizontal tail and vertical tail of the tanker is selected as the feature plane, and feature extraction algorithm for the tanker’s point cloud subset is worked out. Lagrange multiplier method is employed to calculate the unit normal vector of the best fitting feature plane. The unit normal vectors of the successive point cloud frames are available data for visual odometry to estimate relative attitude change between oil tanker and refueled aircraft. According to the flight formation characteristics in AAR, the two feature plane can derivation a line, which can obtain a feature point group by calculate the distance between the line and feature point. As a result, the relative pose can be estimated. The pose of drogue can be fitted by the point cloud of drogue’s outline extracted by level set method.3. The sensor and aircraft’s influence on measure accuracy is analyzed. The plane filter and edge filter are designed to eliminate measurement errors. The tanker’s trajectory prediction model is established and Kalman filter to eliminate errors caused by movement of aircraft.4. Relative pose estimation and control are complementary to each other. In order to improve the robustness, it is necessary to study how to plan a reference trajectory to approach to the tanker in the interference such as wake and gust. The internal model control(IMC) based predictive functional control(PFC) is adopted for the sake of using pilot’s experience. The relative height, relative lateral position and relative velocity internal model controller are established to calculate the basis function of PFC by inverse solution of dynamic model and kinematics model. The first order filter matrix parameters of internal mode controller are set to reduce or eliminate the effect of system coupling by using Smith-Mc Millan form.5. The ultimate goal in AAR is following the reference trajectory and achieving autonomous aerial refueling mission. The auto disturbance rejection control(ADRC) is used to design the flight control law. It is used to deal with the method of observation and compensation for the nonlinear coupling and uncertainty.The innovation of the thesis is as follows:1. According to current research status, defects and development trends of AAR, the autonomous aerial refueling relative pose estimation and control method based on depth information is put forward. The method can provide dissimilar redundancy and reliable data source for relative pose estimation.2. According to tanker model characteristics and AAR requirements, the relative pose can be estimated by visual odometry, which use the unit normal vectors of the horizontal tail,vertical tail and feature point group as the estimation data source.3. The internal model control(IMC) based predictive functional control(PFC) is adopted for the sake of using pilot’s experience. The relative height, relative lateral position and relative velocity internal model controller are established to calculate the basis function of PFC by inverse solution of dynamic model and kinematics model. The first order filter matrix parameters of internal mode controller are set to reduce or eliminate the effect of system coupling by using Smith-Mc Millan form. 4. The auto disturbance rejection control(ADRC) is used to design the flight control law. It is used to deal with the method of observation and compensation for the nonlinear coupling and uncertainty.