Low-altitude large-speed small-radius maneuvering control method for unmanned aerial vehicle

Abstract

The invention discloses a low-altitude large-speed small-radius maneuvering control method for an unmanned aerial vehicle. The method comprises the steps of data acquisition and preprocessing, establishment of an accurate unmanned aerial vehicle model, design of a model prediction controller, sensor fusion, real-time control and the like. Flight state data is collected and preprocessed through various sensors, dynamics and kinematics models considering aerodynamic nonlinearity are established, a model prediction controller is designed by using a recurrent neural network and a particle swarm optimization algorithm, sensor fusion is performed by using an extended Kalman filtering algorithm, and a model prediction model is established. Accurate control of the attitude and track of the unmanned aerial vehicle is realized. Meanwhile, the functions of environment perception and obstacle avoidance, fault diagnosis and fault-tolerant control, real-time controller parameter adjustment, target optimization and the like are added, the maneuvering performance, adaptability and safety of the unmanned aerial vehicle are improved, and the unmanned aerial vehicle can be widely applied to the fields of military reconnaissance, logistics distribution, film and television shooting and the like.

Description

Technical Field

[0001] The present invention relates to the technical field of unmanned aerial vehicle (UAV) control, and in particular to a method for controlling UAV low-altitude, high-speed, and small-radius maneuvers. Background Art

[0002] With the continuous development of drone technology, its application areas are becoming increasingly broad. In many practical applications, such as military reconnaissance, which requires rapid avoidance of enemy radar detection, logistics distribution, and flexible navigating through complex urban environments, as well as capturing specific perspectives in film and television shooting, drones are required to be able to maneuver at low altitudes, high speeds, and small radiuses.

[0003] However, existing drone control methods have numerous shortcomings in achieving low-altitude, high-speed, and small-radius maneuvers. Traditional drone models are often oversimplified, failing to fully account for factors such as complex low-altitude airflow and aerodynamic nonlinearities. This leads to significant deviations between the models and actual flight conditions, compromising control accuracy. Furthermore, some commonly used control algorithms, such as PID control, have poor adaptability and struggle to cope with rapidly changing flight conditions and complex external interference.

[0004] Furthermore, issues with sensor accuracy and reliability also hinder UAV maneuverability. In low-altitude environments, GPS signals are susceptible to obstruction and interference, and IMUs are affected by vibration and temperature, resulting in measurement errors. These factors can lead to inaccurate flight status information, which in turn affects control decisions. Furthermore, issues such as communication latency, bandwidth limitations, and actuator response speed make existing UAV control methods difficult to meet the requirements of low-altitude, high-speed, and small-radius maneuvers. Therefore, developing an efficient and accurate UAV control method for low-altitude, high-speed, and small-radius maneuvers is of great practical significance. Summary of the Invention

[0005] The present invention provides a method for controlling the low-altitude, high-speed, and small-radius maneuvers of an unmanned aerial vehicle (UAV), aiming to solve the problems raised by the above-mentioned background technology.

[0006] The present invention is implemented as follows: a method for controlling low-altitude, high-speed, and small-radius maneuvers of a UAV, comprising the following steps:

[0007] Data collection and preprocessing: Utilizing multiple sensors onboard the drone, including but not limited to the inertial measurement unit (IMU), global positioning system (GPS), barometer, and visual sensor, real-time data collection is performed on the drone's flight status, including position, velocity, acceleration, attitude angle, and altitude. The collected data is preprocessed, including noise removal, data normalization, and data filtering, to improve data accuracy and reliability.

[0008] Establish an accurate drone model: Based on the drone's physical properties and aerodynamic principles, establish the drone's dynamic and kinematic models. The dynamic model considers the drone's mass, moment of inertia, aerodynamic forces, and torque to describe the drone's state changes. The kinematic model is used to describe the drone's motion trajectory and attitude changes in space.

[0009] Design a model predictive controller: Based on the UAV model, design a predictive model to predict the UAV's state at multiple future time steps. Use an optimization algorithm to find the optimal control input sequence based on the current UAV state and the target state, combined with constraints including maximum speed, maximum acceleration, minimum turning radius, and motor speed limits. Compare the measured UAV state with the predicted state, calculate the state error, and calibrate the predictive model and control inputs.

[0010] Sensor fusion and real-time control: A sensor fusion algorithm is used to fuse data from multiple sensors to provide accurate status feedback to the controller. Based on the optimal control input calculated by the model prediction controller, control instructions are sent to the UAV's actuators, including motors and servos, to control the UAV's attitude and trajectory. During flight, the controller's parameters and optimization targets are adjusted in real time based on the actual flight status and environmental changes.

[0011] Preferably, in the data collection and preprocessing steps, the Kalman filter algorithm is used to remove noise, the minimum-maximum normalization method is used for data normalization, and the sliding average filter method is used for data filtering.

[0012] Preferably, in the step of establishing an accurate UAV model, when establishing the dynamic model, the nonlinear changes of aerodynamic parameters under different flight attitudes and speeds are taken into account, and the model is calibrated and optimized through wind tunnel experiments and flight test data.

[0013] Preferably, in the step of designing a model predictive controller, the predictive model uses a recursive neural network (RNN) for training and prediction, and the optimization algorithm uses a particle swarm optimization algorithm to solve the optimal control input sequence.

[0014] Preferably, in the sensor fusion and real-time control step, the sensor fusion algorithm adopts an extended Kalman filter (EKF) algorithm to fuse the data of the IMU, GPS and vision sensor.

[0015] Preferably, it also includes environmental perception and obstacle avoidance steps: using lidar, millimeter-wave radar and visual sensors to perceive the environmental information around the drone in real time, and identify the position, size and shape of obstacles; based on the environmental information and the flight status of the drone, a path planning algorithm is used to generate an obstacle avoidance path, and the flight trajectory of the drone is adjusted to avoid collision with obstacles.

[0016] Preferably, the path planning algorithm adopts A* algorithm or Dijkstra algorithm, combined with the dynamic constraints and maneuverability of the UAV, to generate a smooth and feasible obstacle avoidance path.

[0017] Preferably, it also includes fault diagnosis and fault-tolerant control steps: establishing a fault diagnosis model, monitoring the working status of the drone's sensors and actuators in real time, and determining whether a fault has occurred by analyzing abnormal conditions in sensor data and control instructions; when a fault is detected, adopting a fault-tolerant control strategy according to the type and severity of the fault, adjusting the control algorithm and control parameters to ensure the safe flight of the drone.

[0018] Preferably, the fault diagnosis model uses a support vector machine (SVM) or a neural network for training and fault classification, and the fault-tolerant control strategy includes redundant backup, reconstruction of control law and reduction of flight performance requirements.

[0019] Preferably, in the step of adjusting the controller parameters and optimizing the targets in real time, a reinforcement learning algorithm is used to dynamically adjust the controller parameters and optimization targets according to the flight status and mission objectives of the UAV to improve the adaptability and robustness of the control method.

[0020] Due to the adoption of the above scheme, the beneficial effects of the present invention are: improving control accuracy: by establishing a precise UAV model, fully considering the aerodynamic characteristics and nonlinear factors in the complex environment at low altitude, and combining with the model predictive control algorithm, the state of the UAV can be predicted more accurately, thereby achieving more precise control.

[0021] Enhanced adaptability and robustness: Sensor fusion technology leverages data from multiple sensors to improve the accuracy and reliability of state feedback. Furthermore, real-time adjustments to controller parameters and optimization targets during flight enable better adaptation to varying flight conditions and environmental changes, enhancing the robustness of the control approach.

[0022] Achieve safe flight: Added environmental perception and obstacle avoidance steps as well as fault diagnosis and fault-tolerant control steps, which can perceive the surrounding environment in real time, avoid collision with obstacles, and take fault-tolerant measures in time when a fault occurs to ensure the safe flight of the drone.

[0023] Improved maneuverability: Through optimized control algorithms and real-time adjustment strategies, the drone can achieve high-speed and small-radius maneuverable flight at low altitudes, meeting the needs of complex application scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Figure 1 Schematic diagram of the system flow of the present invention. DETAILED DESCRIPTION

[0025] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0026] like Figure 1 A method for controlling a UAV's low-altitude, high-speed, and small-radius maneuvering is shown, comprising the following steps:

[0027] Data collection and preprocessing: Utilizing multiple sensors onboard the drone, including but not limited to the inertial measurement unit (IMU), global positioning system (GPS), barometer, and visual sensor, real-time data collection is performed on the drone's flight status, including position, velocity, acceleration, attitude angle, and altitude. The collected data is preprocessed, including noise removal, data normalization, and data filtering, to improve data accuracy and reliability.

[0028] Establish an accurate drone model: Based on the drone's physical properties and aerodynamic principles, establish the drone's dynamic and kinematic models. The dynamic model considers the drone's mass, moment of inertia, aerodynamic forces, and torque to describe the drone's state changes. The kinematic model is used to describe the drone's motion trajectory and attitude changes in space.

[0029] Design a model predictive controller: Based on the UAV model, design a predictive model to predict the UAV's state at multiple future time steps. Use an optimization algorithm to find the optimal control input sequence based on the current UAV state and the target state, combined with constraints including maximum speed, maximum acceleration, minimum turning radius, and motor speed limits. Compare the measured UAV state with the predicted state, calculate the state error, and calibrate the predictive model and control inputs.

[0030] Sensor fusion and real-time control: A sensor fusion algorithm is used to fuse data from multiple sensors to provide accurate status feedback to the controller. Based on the optimal control input calculated by the model prediction controller, control instructions are sent to the UAV's actuators, including motors and servos, to control the UAV's attitude and trajectory. During flight, the controller's parameters and optimization targets are adjusted in real time based on the actual flight status and environmental changes.

[0031] In the data collection and preprocessing steps, the Kalman filter algorithm is used to remove noise, the minimum-maximum normalization method is used for data normalization, and the sliding average filter method is used for data filtering.

[0032] In the step of establishing an accurate UAV model, when establishing the dynamic model, the nonlinear changes of aerodynamic parameters under different flight attitudes and speeds are taken into account, and the model is calibrated and optimized through wind tunnel experiments and flight test data.

[0033] In the step of designing a model predictive controller, the predictive model uses a recurrent neural network (RNN) for training and prediction, and the optimization algorithm uses a particle swarm optimization algorithm to solve the optimal control input sequence.

[0034] In the sensor fusion and real-time control step, the sensor fusion algorithm adopts the extended Kalman filter (EKF) algorithm to fuse the data of the IMU, GPS and vision sensor.

[0035] It also includes environmental perception and obstacle avoidance steps: using lidar, millimeter-wave radar and visual sensors to perceive the environmental information around the drone in real time, and identify the position, size and shape of obstacles; based on the environmental information and the flight status of the drone, a path planning algorithm is used to generate an obstacle avoidance path, and the flight trajectory of the drone is adjusted to avoid collisions with obstacles.

[0036] The path planning algorithm uses the A* algorithm or the Dijkstra algorithm, combined with the dynamic constraints and maneuverability of the UAV, to generate a smooth and feasible obstacle avoidance path.

[0037] It also includes fault diagnosis and fault-tolerant control steps: establishing a fault diagnosis model, monitoring the working status of the drone's sensors and actuators in real time, and determining whether a fault has occurred by analyzing abnormalities in sensor data and control instructions; when a fault is detected, a fault-tolerant control strategy is adopted according to the type and severity of the fault, and the control algorithm and control parameters are adjusted to ensure the safe flight of the drone.

[0038] The fault diagnosis model uses a support vector machine (SVM) or a neural network for training and fault classification, and the fault-tolerant control strategy includes redundant backup, reconstructing the control law, and reducing flight performance requirements.

[0039] In the step of adjusting the controller parameters and optimizing the target in real time, a reinforcement learning algorithm is used to dynamically adjust the controller parameters and optimization targets according to the flight status and mission objectives of the UAV to improve the adaptability and robustness of the control method.

[0040] In this embodiment, data collection and preprocessing

[0041] Sensor selection and installation: Various sensors are installed on the drone to obtain comprehensive flight status data. An inertial measurement unit (IMU) is typically mounted at the drone's center of gravity to measure the drone's acceleration and angular velocity. High-precision MEMS (micro-electromechanical system) IMUs are ideal, offering advantages such as small size, light weight, and low cost. A Global Positioning System (GPS) antenna is mounted on the top of the drone to ensure good signal reception. A multi-frequency, multi-mode GPS module can be selected to improve positioning accuracy. A barometer is installed inside the drone to measure atmospheric pressure and calculate the drone's altitude. Visual sensors, such as cameras, are mounted on the front of the drone or other suitable locations to capture images of the surrounding environment. Data acquisition: Various sensors collect data in real time at a specific sampling rate. For example, the IMU's sampling rate can be set to 100Hz-200Hz to ensure it can capture rapid changes in the drone's movements. The GPS sampling rate is generally 1Hz-5Hz. The barometer's sampling rate can be set to 10Hz-50Hz depending on actual needs. The visual sensor's frame rate can be set to 10fps-60fps depending on the specific application scenario. Data Preprocessing - Noise Removal: Kalman filtering is used to remove noise from the collected data. Kalman filtering is an optimal estimation algorithm that can optimally estimate data based on the system's dynamic model and the statistical characteristics of measurement noise. For example, using acceleration data collected by an IMU, by establishing a dynamic acceleration model and a measurement noise model, the Kalman filter can effectively remove noise interference and obtain more accurate acceleration values.

[0042] Data normalization: Use the minimum-maximum normalization method to normalize the data to the [0,1] interval. The calculation formula of this method is: Where x is the original data, x min and x max are the minimum and maximum values ​​of the data, respectively, x norm The data is normalized. Data normalization can eliminate the dimensional differences between different sensor data and improve the efficiency and accuracy of subsequent processing.

[0043] Data filtering: The data is further filtered using a sliding average filter. This is a simple and effective filtering method that smooths data by calculating the average value of the data within a specific time window. For example, for GPS location data, setting a sliding window of 5 and calculating the average value of the data within the window as the filtered data at the current moment can reduce data fluctuations and improve data stability.

[0044] Establish an accurate UAV model and dynamic model. Based on Newton's second law and Euler equation, establish the UAV dynamic model. Consider the UAV's mass m, moment of inertia I, and aerodynamic force F. a and moment M a , the dynamic equation of the UAV can be expressed as:

[0045]

[0046] Among them, ∑F is the resultant force acting on the UAV, i is the acceleration of the UAV, ∑M is the resultant torque acting on the UAV, is the angular acceleration of the UAV.

[0047] Aerodynamic forces and moments are key components of UAV dynamics models. They are closely related to factors such as the UAV's flight attitude, speed, and airflow. When developing the model, the nonlinear variations of aerodynamic parameters at different flight attitudes and speeds are fully considered. The model is calibrated and optimized using wind tunnel experiments and flight test data to improve its accuracy. For example, in wind tunnel experiments, the lift, drag, and moment coefficients of the UAV are measured at different angles of attack, sideslip angles, and speeds. This data is then fitted into the dynamics model.

[0048] Design a model predictive controller and predictive model. Based on the established drone model, a predictive model is designed to predict the drone's state multiple time steps into the future. A recurrent neural network (RNN) is used for training and prediction. RNNs have memory capabilities and can process sequential data, making them suitable for predicting the drone's dynamic state. The drone's current state and control inputs are used as inputs to the RNN. By training the network, it learns the changing patterns of the drone's state and can predict its state multiple time steps into the future.

[0049] An optimization algorithm solves the problem by finding the optimal control input sequence based on the current drone state and target state, combined with constraints including maximum speed, maximum acceleration, minimum turning radius, and motor speed limits. Particle swarm optimization (PSO) is used to solve this problem. PSO, an optimization algorithm based on swarm intelligence, simulates the foraging behavior of flocks of birds or fish to search for the optimal solution within the solution space. Each particle represents a possible control input sequence, and the particle's position and velocity are continuously updated to move it closer to the optimal solution.

[0050] Feedback correction compares the actual measured drone state with the predicted state and calculates the state error. This error information is used to correct the prediction model and control inputs. For example, when the deviation between the actual and predicted states exceeds a certain threshold, the RNN parameters are adjusted to improve the accuracy of the prediction model. Simultaneously, the control input sequence is modified based on the error, allowing the drone to return to the desired trajectory more quickly.

[0051] Sensor fusion and real-time control. The sensor fusion algorithm uses the Extended Kalman Filter (EKF) algorithm to fuse data from the IMU, GPS, and vision sensors. The EKF is a nonlinear filtering algorithm that linearizes nonlinear systems and uses the principles of Kalman filtering to perform state estimation. The EKF uses IMU acceleration and angular velocity data, GPS position and velocity data, and vision sensor image feature data as inputs. By continuously updating the state estimate, more accurate drone status information is obtained.

[0052] Real-time control: Based on the optimal control inputs calculated by the model predictive controller, control commands are sent to the drone's actuators, including the motors and servos. The motors adjust their speed according to the control commands, generating lift and thrust; the servos adjust the control surface angles according to the control commands, changing the drone's attitude and flight direction. During flight, the drone's status and environmental changes are monitored in real time, and controller parameters and optimization targets are adjusted accordingly. For example, when encountering strong wind interference, the control gain is increased to improve the drone's anti-interference ability. When approaching obstacles, the target trajectory is adjusted to perform obstacle avoidance maneuvers.

[0053] Environmental perception and obstacle avoidance. Environmental perception utilizes lidar, millimeter-wave radar, and vision sensors to perceive the drone's surroundings in real time. Lidar emits a laser beam and measures the time it takes for the reflected light to acquire the distance and location of obstacles, offering high precision and resolution. Millimeter-wave radar performs well in adverse weather conditions and can monitor the movement of targets around the drone in real time. Vision sensors use image processing and computer vision algorithms to identify the location, size, and shape of obstacles.

[0054] Obstacle avoidance path planning uses a path planning algorithm to generate an obstacle avoidance path based on environmental information and the drone's flight status. Either the A-algorithm or the Dijkstra algorithm can be used, combining the drone's dynamic constraints and maneuverability to generate a smooth, feasible obstacle avoidance path. For example, the A-algorithm uses a heuristic search to find the shortest path from the current location to the target location within a map. During the search, constraints such as the drone's turning radius and maximum speed are considered to ensure the generated path is feasible.

[0055] Fault diagnosis and fault-tolerant control involve building a fault diagnosis model to monitor the operating status of the drone's sensors and actuators in real time. Support vector machines (SVMs) or neural networks are used for training and fault classification. For example, sensor fault diagnosis involves collecting sensor data under normal and fault conditions and using them as training samples to train the SVM or neural network model. While the drone is in flight, the real-time sensor data is fed into the trained model to determine whether a fault has occurred and what type of fault it is.

[0056] When a fault is detected, a fault-tolerant control strategy is implemented based on the fault type and severity. This includes redundant backups, reconfiguring control laws, and reducing flight performance requirements. For example, if a sensor fails, a backup sensor is activated. If an actuator fails, the control law is reconfigured and the control inputs of other actuators are adjusted to ensure the drone's basic flight capabilities. If the fault is severe and normal flight cannot be fully restored, flight performance requirements are reduced, such as reducing speed and maneuvering range, to ensure the drone can land safely.

[0057] A reinforcement learning algorithm is used to dynamically adjust controller parameters and optimization objectives based on the drone's flight status and mission objectives. Reinforcement learning is a method that continuously learns optimal strategies through interaction between an intelligent agent and its environment. The drone is considered the intelligent agent, the flight environment is the environment, control inputs are considered actions, and mission completion is considered rewards. Through continuous interaction with the environment, the intelligent agent learns the optimal control strategy, dynamically adjusting the controller parameters and optimization objectives to improve the adaptability and robustness of the control method.

[0058] The above description of the embodiments is intended to facilitate the understanding and use of the present invention by those skilled in the art. It is obvious that those skilled in the art can easily make various modifications to these embodiments and apply the general principles described herein to other embodiments without having to go through creative work. Therefore, the present invention is not limited to the above embodiments. Improvements and modifications made by those skilled in the art based on the principles of the present invention without departing from the scope of the present invention should be within the scope of protection of the present invention. The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for controlling low-altitude, high-speed, and small-radius maneuvers for a drone, characterized in that: The following steps are involved: Data collection and preprocessing: Utilizing multiple sensors onboard the drone, including but not limited to the inertial measurement unit (IMU), global positioning system (GPS), barometer, and visual sensor, real-time data collection is performed on the drone's flight status, including position, velocity, acceleration, attitude angle, and altitude. The collected data is preprocessed, including noise removal, data normalization, and data filtering, to improve data accuracy and reliability. Establish an accurate drone model: Based on the drone's physical properties and aerodynamic principles, establish the drone's dynamic and kinematic models. The dynamic model considers the drone's mass, moment of inertia, aerodynamic forces, and torque to describe the drone's state changes. The kinematic model is used to describe the drone's motion trajectory and attitude changes in space. Design a model predictive controller: Based on the UAV model, design a predictive model to predict the UAV's state at multiple future time steps. Use an optimization algorithm to find the optimal control input sequence based on the current UAV state and the target state, combined with constraints including maximum speed, maximum acceleration, minimum turning radius, and motor speed limits. Compare the measured UAV state with the predicted state, calculate the state error, and calibrate the predictive model and control inputs. Sensor fusion and real-time control: A sensor fusion algorithm is used to fuse data from multiple sensors to provide accurate status feedback to the controller. Based on the optimal control input calculated by the model prediction controller, control instructions are sent to the UAV's actuators, including motors and servos, to control the UAV's attitude and trajectory. During flight, the controller's parameters and optimization targets are adjusted in real time based on the actual flight status and environmental changes.

2. The method for controlling low-altitude, high-speed, and small-radius maneuvers of a UAV according to claim 1, characterized in that: In the data collection and preprocessing steps, the Kalman filter algorithm is used to remove noise, the minimum-maximum normalization method is used for data normalization, and the sliding average filter method is used for data filtering.

3. The method for controlling low-altitude, high-speed, and small-radius maneuvers of a UAV according to claim 1, characterized in that: In the step of establishing an accurate UAV model, when establishing the dynamic model, the nonlinear changes of aerodynamic parameters under different flight attitudes and speeds are taken into account, and the model is calibrated and optimized through wind tunnel experiments and flight test data.

4. The method for controlling low-altitude, high-speed, and small-radius maneuvers of a UAV according to claim 1, characterized in that: In the step of designing a model predictive controller, the predictive model uses a recurrent neural network (RNN) for training and prediction, and the optimization algorithm uses a particle swarm optimization algorithm to solve the optimal control input sequence.

5. The method for controlling low-altitude, high-speed, and small-radius maneuvers of a UAV according to claim 1, characterized in that: In the sensor fusion and real-time control step, the sensor fusion algorithm adopts the extended Kalman filter (EKF) algorithm to fuse the data of the IMU, GPS and vision sensor.

6. The method for controlling low-altitude, high-speed, and small-radius maneuvers of a UAV according to claim 1, characterized in that: It also includes environmental perception and obstacle avoidance steps: using lidar, millimeter-wave radar and visual sensors to perceive the environmental information around the drone in real time, and identify the position, size and shape of obstacles; based on the environmental information and the flight status of the drone, a path planning algorithm is used to generate an obstacle avoidance path, and the flight trajectory of the drone is adjusted to avoid collisions with obstacles.

7. The method for controlling low-altitude, high-speed, and small-radius maneuvers of a UAV according to claim 6, characterized in that: The path planning algorithm uses the A* algorithm or the Dijkstra algorithm, combined with the dynamic constraints and maneuverability of the UAV, to generate a smooth and feasible obstacle avoidance path.

8. The method for controlling low-altitude, high-speed, and small-radius maneuvers of a UAV according to claim 1, characterized in that: It also includes fault diagnosis and fault-tolerant control steps: establishing a fault diagnosis model, monitoring the working status of the drone's sensors and actuators in real time, and determining whether a fault has occurred by analyzing abnormalities in sensor data and control instructions; when a fault is detected, a fault-tolerant control strategy is adopted according to the type and severity of the fault, and the control algorithm and control parameters are adjusted to ensure the safe flight of the drone.

9. The method for controlling low-altitude, high-speed, and small-radius maneuvers of a UAV according to claim 8, characterized in that: The fault diagnosis model uses a support vector machine (SVM) or a neural network for training and fault classification, and the fault-tolerant control strategy includes redundant backup, reconstructing the control law, and reducing flight performance requirements.

10. The method for controlling low-altitude, high-speed, and small-radius maneuvers of a UAV according to claim 1, characterized in that: In the step of adjusting the controller parameters and optimizing the target in real time, a reinforcement learning algorithm is used to dynamically adjust the controller parameters and optimization targets according to the flight status and mission objectives of the UAV to improve the adaptability and robustness of the control method.

Images