Research On Autonomous Navigation And Flight Control Methods Of Quadrotor UAV In Unknown Environment

The autonomous flight capability of the small unmanned aerial vehicles(UAVs)such as quadrotors is the prerequisite for complex missions and an important development direction in related fields.In the unknown environment,autonomous flight mission of UAV relies on the combination of high-precision flight control method,stable motion estimation,stable and reliable map building method,accurate motion decision and motion planning.These problems are related to each other and form a complete autonomous flight system.In this paper,starting from the task of the quadrotor UAV exploring the unknown environment,the research status of related fields is surveyed.The paper focuses on the mathematical model,the visual-inertial integrated navigation method,the active disturbance rejection attitude control,simultaneous localization and mapping algorithm,motion decision and motion planning method of the quadrotor UAV.The quadrotor aircraft can be regarded as a rigid body driven by 4 sets of actuators consisting of motors and propellers.A mathematical model including the actuator model and the rigid body dynamic equation is established in this paper.According to the aerodynamic model of the propeller,the influence of wind on the tension and torque of the propeller is analyzed.The parameters of the aerodynamic model are calculated using the measured physical dimension,tension and torque curves of the propellers.This paper establishes a visual-inertial integrated navigation system,which combines the camera and IMU data in a tightly coupled form,and directly uses the original intensity value of the camera output as the measurements.The deviation of the intensity between current image and the keyframe is used as the residual to compensate for the drift of the inertial navigation error.A visual-inertial navigation system is built with the extended information filter.Based on that,the cubature rules are employed for dealing with the nonlinearity of the system model.By analyzing the nonlinearity and state dimension of the navigation system,the 3rd-order spherical-radial cubature rule is applied in the time update equation,and the 5th-order spherical simplex-radial cubature rule is applied in the measurement update equation for the numerical stability and filtering accuracy.Considering the non-Gaussian distribution characteristics of the noise of the camera's raw measurements,the above-mentioned mixed-degree cubature information filter is improved by H_?filter theory to obtain robustness toward non-Gaussian measurement noises.Finally,a robust nonlinear visual-inertial navigation method with linear time complexity is obtained.Aiming at the problem that the small quadrotor is susceptible to disturbances such as dynamic wind,mass eccentricity and actuator failure,the active disturbance rejection attitude control method is studied.The feedback linearization technique is used to regularize the system model,which avoids the parameter coupling problem between the extended state observer and the nonlinear feedback control law in the classical active disturbance rejection control.A modified Rodrigues paramter-based active disturbance refection controller is designed on that basis.The enhanced switched-type extended state observer and multivariable nonlinear feedback control law are established by applying high-order sliding mode theory,and the convergence speed is improved.The method has the advantages of fixed time convergence,easy setting and strong real-time performance.Through simulation and experimental comparison,the robustness of the fixed-time convergence active disturbance rejection attitude control method to unknown external disturbances is proved.Considering the coupling characteristics of the problem of absolute pose estimation of unmanned aerial vehicles in unknown environment with the mapping problem,the particle filter based simultaneous localization and mapping algorithm is studied.According to the requirements of the environment description of exploration task for the unknown environment,the map description method based on the octree structure is used to study the simultaneous localization and mapping algorithm in the three-dimensional environment.The autonomous exploration problem of the unmanned aerial vehicle in the unknown environment requies solving the motion decision problem of selecting the optimal motion target and the motion planning problem of driving the UAV towards the target.In order to establish an integrated solution for the above two problems,this paper uses the virtual Brownian motion-particles to extract the frontier,and uses the parallel A*algorithm to evluate the global optimal path and the corresponding motion target in the octree map.In order to meet the needs of map establishment and location in an unknown environment,an information-based exploration strategy is introduced to obtain an approximate information gain through closed-loop control and truncating measurement rays for the disadvantage that information gain is difficult to calculate online.A new cost function is defined based on the approximate information gain to improve the original A*algorithm,so that the UAV moves toward the boundary while optimizing the information gain during the motion.The method can realize the autonomous navigation capability of the UAV in an unknown environment.Based on this,the path tracking method is further studied based on the backstepping method and the differential flatness characteristics of the quadrotor.In summary,this paper takes the promotion of the autonomy of the quadrotor UAV as the fundamental purpose,and exploring the unknown environment as the task object.The study focuses on the active disturbance rejection attitude control method,the visual inertial navigation method,the motion decision and planning of the unknown environment exploration.The research results provide some new methods and new ideas for the autonomous navigation of UAVs.