Flight control methods, devices, equipment, storage media, and computer program products
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- CHINA MOBILE GROUP DESIGN INST
- Filing Date
- 2026-03-18
- Publication Date
- 2026-06-30
Smart Images
Figure CN122308465A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of intelligent flight control technology for unmanned aerial vehicles (UAVs), and more particularly to a flight control method, device, equipment, storage medium, and computer program product. Background Technology
[0002] Multi-rotor drones, due to their simple structure and high maneuverability, have been widely used in aerial photography, logistics transportation, emergency rescue, and other fields. During flight, the rotor, as a critical actuator, directly affects the drone's flight safety and mission execution capability. If a single rotor or multi-rotor malfunction occurs and is not detected and effectively controlled in time, the drone will face the risk of attitude instability, power imbalance, or even crash.
[0003] While existing technologies have laid a foundation for research on UAV flight control, shortcomings remain in real-time detection and adaptive control after rotor failure. Traditional control methods often rely on preset models or offline parameter adjustments, making it difficult to quickly reconstruct control strategies in the event of sudden failures. Furthermore, existing solutions often lack effective redundant power distribution mechanisms after a failure, failing to fully utilize the remaining rotor thrust resources to maintain flight stability.
[0004] Therefore, ensuring that drones can maintain stable flight and successfully complete their missions even after some rotors fail, without the need for additional hardware, has become a pressing technical problem that needs to be solved. Summary of the Invention
[0005] This application provides a flight control method to solve the problem that the prior art cannot ensure that a multi-rotor UAV can maintain stable flight and successfully complete its mission after some rotors fail without adding additional hardware.
[0006] This application also provides a flight control device to solve the problem that the prior art cannot ensure that a multi-rotor UAV can maintain stable flight and successfully complete its mission after some rotors fail without adding additional hardware.
[0007] This application also provides a flight control device to solve the problem that the prior art cannot ensure that a multi-rotor UAV can maintain stable flight and successfully complete its mission after some rotors fail without adding additional hardware.
[0008] This application also provides a computer-readable storage medium to address the problem that, without adding additional hardware, existing technologies cannot ensure that multi-rotor drones can maintain stable flight and successfully complete their missions after some rotors fail.
[0009] A computer program product designed to address the problem that existing technologies cannot ensure stable flight and successful mission completion of multi-rotor drones after partial rotor failure without the addition of extra hardware.
[0010] The embodiments of this application adopt the following technical solutions: A flight control method includes: when a rotor failure is determined based on collected operating data of each rotor of an aircraft, acquiring thrust data of the remaining normal rotors; constructing a flight optimization model based on the thrust data, wherein the flight optimization model aims to minimize energy consumption and thrust deviation, and satisfies torque balance constraints and motor thrust limiting constraints; determining sliding mode control parameters based on a pre-trained deep reinforcement learning network and the flight optimization model, and determining a sliding mode control law based on the sliding mode control parameters, wherein the sliding surface in the sliding mode control parameters is constructed based on tracking error; generating control commands based on the sliding mode control law, and adjusting the operating parameters of the normal rotors based on the control commands.
[0011] A flight control device includes: a fault detection unit, configured to acquire thrust data of the remaining normal rotors when a rotor fault is determined to occur in the aircraft based on collected operating data of each rotor; an optimization model construction unit, configured to construct a flight optimization model based on the thrust data, wherein the flight optimization model aims to minimize energy consumption and thrust deviation, and satisfies torque balance constraints and motor thrust limiting constraints; a parameter generation unit, configured to determine sliding mode control parameters based on a pre-trained deep reinforcement learning network and the flight optimization model, and determine a sliding mode control law based on the sliding mode control parameters, wherein the sliding surface in the sliding mode control parameters is constructed based on tracking error; and a flight control unit, configured to generate control commands based on the sliding mode control law, and adjust the operating parameters of the normal rotors based on the control commands.
[0012] A flight control device, comprising: The system includes a processor and a memory configured to store computer-executable instructions, which, when executed, cause the processor to perform the following operations: when a rotor failure is determined based on collected operating data of each rotor of the aircraft, the processor acquires thrust data of the remaining normal rotors; based on the thrust data, the processor constructs a flight optimization model, wherein the flight optimization model aims to minimize energy consumption and thrust deviation, and satisfies torque balance constraints and motor thrust limiting constraints; based on a pre-trained deep reinforcement learning network and the flight optimization model, the processor determines sliding mode control parameters and a sliding mode control law, wherein the sliding surface in the sliding mode control parameters is constructed based on tracking error; the processor generates control commands based on the sliding mode control law, and adjusts the operating parameters of the normal rotors based on the control commands.
[0013] A computer-readable storage medium storing one or more programs, which, when executed by an electronic device including multiple applications, cause the electronic device to perform the following operations: when a rotor failure is determined to have occurred in the aircraft based on collected operating data of each rotor, the electronic device acquires thrust data of the remaining normal rotors; constructs a flight optimization model based on the thrust data, wherein the flight optimization model aims to minimize energy consumption and thrust deviation, and satisfies torque balance constraints and motor thrust limiting constraints; determines sliding mode control parameters based on a pre-trained deep reinforcement learning network and the flight optimization model, and determines a sliding mode control law based on the sliding mode control parameters, wherein the sliding surface in the sliding mode control parameters is constructed based on tracking error; generates control commands based on the sliding mode control law, and adjusts the operating parameters of the normal rotors based on the control commands.
[0014] A computer program product includes a computer program that, when executed by a processor, performs the following: when a rotor failure is determined based on collected operating data of each rotor of an aircraft, the thrust data of the remaining normal rotors is obtained; based on the thrust data, a flight optimization model is constructed, wherein the flight optimization model aims to minimize energy consumption and thrust deviation, and satisfies torque balance constraints and motor thrust limiting constraints; based on a pre-trained deep reinforcement learning network and the flight optimization model, sliding mode control parameters are determined, and a sliding mode control law is determined based on the sliding mode control parameters, wherein the sliding surface in the sliding mode control parameters is constructed based on tracking error; control commands are generated based on the sliding mode control law, and the operating parameters of the normal rotors are adjusted based on the control commands.
[0015] The above-described technical solutions adopted in the embodiments of this application can achieve the following beneficial effects: Using the flight control method provided in this application, when it is determined that the aircraft has a rotor failure based on the collected operating data of each rotor, the thrust data of the remaining normal rotors can be obtained first. Based on the thrust data, a flight optimization model is constructed. Based on the pre-trained deep reinforcement learning network and the flight optimization model, sliding mode control parameters and sliding mode control laws are determined. Then, control commands can be generated based on the sliding mode control laws, and the operating parameters of the normal rotors can be adjusted according to the control commands. The flight control method provided in this application has two advantages. First, by collecting rotor operation data in real time, it can quickly identify rotor faults during flight and dynamically reconstruct the control model based on the thrust data of the remaining normal rotors. Compared with existing solutions that rely on preset routes or historical data, the method provided in this application can immediately initiate fault-tolerant control after a fault occurs, avoiding the risk of flight attitude instability or loss of control due to response delays, and significantly improving the survivability of the UAV under sudden faults. Second, the flight optimization model constructed in this application aims to minimize energy consumption and thrust deviation, and solves the optimal thrust distribution scheme under torque balance constraints and motor thrust limiting constraints. By minimizing the thrust deviation, it ensures that the resultant force and resultant torque generated by the remaining rotors can accurately compensate for the power lost by the faulty rotor, maintaining the attitude stability of the UAV. Furthermore, by minimizing energy consumption, secondary failures caused by overload of the remaining rotor can be avoided, extending the UAV's loiter time and ensuring mission completion or safe return. Finally, by introducing a strategy combining deep reinforcement learning networks with sliding mode control, the switching gain adjustment value of sliding mode control can be adaptively output according to the real-time state of the aircraft and the optimization model. This allows the control parameters to be dynamically optimized according to changes in fault type, fault severity, and flight environment. At the same time, precise control commands are generated through sliding mode control laws. When the UAV experiences asymmetric faults or is disturbed by environmental factors such as strong winds, sliding mode control can force the system state to move along the sliding surface, ensuring that attitude angle and position tracking errors converge quickly to zero. Compared with traditional PID control, which is prone to oscillation or overshoot after a fault, the flight control method provided in this application has higher control accuracy and response speed. Attached Figure Description
[0016] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings: Figure 1 This is a schematic flowchart illustrating a flight control method provided in an embodiment of this application. Figure 2 This is a schematic diagram of the specific structure of a flight control device provided in an embodiment of this application; Figure 3This is a schematic diagram of the specific structure of a flight control device provided in an embodiment of this application. Detailed Implementation
[0017] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions of this application will be clearly and completely described below in conjunction with specific embodiments and corresponding drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.
[0018] This application provides a flight control method to solve the problem that existing technologies cannot ensure that multi-rotor drones can maintain stable flight and successfully complete their missions after some rotors fail, without adding additional hardware.
[0019] To facilitate understanding of the flight control method provided in the embodiments of this application, the technical terms involved in the embodiments of this application will be explained below: 1. IMU: Inertial Measurement Unit, used to measure the three-axis attitude angles and acceleration of an aircraft.
[0020] 2. FPGA: Field-Programmable Gate Array, is a programmable logic device used to implement high-speed parallel computing.
[0021] 3. ARM: A 32-bit reduced instruction set processor architecture. In this application embodiment, it mainly refers to a microprocessor based on the ARM architecture, used for system management and data processing.
[0022] 4. DDPG: Deep Deterministic Policy Gradient, is a deep reinforcement learning algorithm applicable to continuous action spaces.
[0023] 5. PID: Proportional-Integral-Derivative (PID) controller.
[0024] 6. MPC: Model Predictive Control.
[0025] 7. GPS: Global Positioning System.
[0026] 8. ENU: East-North-Up coordinate system, a commonly used inertial coordinate system.
[0027] The execution subject of the flight control method provided in this application embodiment may be, but is not limited to, at least one of a flight control server, a flight management server, and an exception handling server; in addition, the execution subject of the method may also be the system or application (APP) itself running on these servers.
[0028] For ease of description, the following description uses the flight control system as the executing entity to illustrate the implementation of this method. It should be understood that using the flight control system as the executing entity is merely an illustrative example and should not be construed as a limitation of the method.
[0029] This invention provides a flight control system applicable to multi-rotor unmanned aerial vehicles (UAVs), particularly suitable for stable flight control after a single-rotor or multi-rotor malfunction. A schematic diagram illustrating the specific implementation flow of the flight control method provided in this application based on this flight control system is shown below. Figure 1 As shown, the main steps include the following: Step 11: Collect real-time operating data of each rotor of the aircraft. When it is determined that the aircraft has a rotor failure based on the collected operating data, obtain the thrust data of the remaining normal rotors. It is important to note that, in order to accurately describe the dynamic behavior of a multi-rotor UAV in terms of position, attitude, force, and torque balance, a dynamic model incorporating these relationships should be established before collecting operational data. This model forms the basis for subsequent fault detection, optimization modeling, and control law design.
[0030] In the embodiments of this application, the established dynamic model may include, but is not limited to, the following: 1. Position dynamics equations: Assume that the UAV's center of mass has the position coordinates P = [x, y, z] in an inertial coordinate system (e.g., an ENU coordinate system) and its velocity V = [v]. x v y v z ] T The acceleration is a = [a x a y a z ] T According to Newton's second law, the corresponding position dynamics equation is constructed as shown in the following formula [1]: [1] Where m represents the mass of the drone, F represents the total lift vector of the drone, and g represents the gravitational acceleration vector.
[0031] 2. Attitude dynamics equations: In this embodiment of the application, the attitude angle of the UAV can be defined as follows: ,in For roll angle, The pitch angle, This is the yaw angle. The attitude angular velocity is... The constructed attitude dynamics equations are shown in the following formula [2]: [2] Where J represents the inertia matrix of the UAV, τ represents the torque of the UAV, and e3 is the unit vector.
[0032] 3. Force and torque equilibrium equations: Assume that the UAV has n rotors, each rotor generates a thrust of Ti and a torque of τi. Then the total lift and torque can be expressed by the following formula [3]: [3] Where ri is the relative position vector of the rotor.
[0033] 4. Rotor dynamics model The lift Fi and counter-torque Mi of each rotor are proportional to the square of the rotational speed ωi: The lift can then be expressed by the following formula [4]: [4] The reverse torque can be expressed by the following formula [5]: [5] Among them, C T and C M This is an empirical coefficient. In the embodiments of this application, the total lift T = ∑F can be generated by adjusting the rotational speeds of the four rotors. i and triaxial torque τ x ,τ y ,τ z This enables attitude control of the drone.
[0034] 5. Coordinate system and attitude representation: The coordinate system can include an inertial coordinate system (ENU) and a body coordinate system (X, Y, Z), and coordinate transformation can be achieved through rotation matrices or quaternions.
[0035] Meanwhile, to avoid the gimbal lock problem of Euler angles, this invention preferably uses quaternions q = [q0, q1, q2, q3] to represent the attitude. Its kinematic equation is shown in the following formula [6]: [6] in, This represents quaternion multiplication.
[0036] 6. Aerodynamics and Disturbance Model: In this embodiment of the application, in order to more accurately describe the actual flight environment, an air resistance model can be introduced, as shown in the following formula [7]: [7] Where v is velocity, k d This is the drag coefficient.
[0037] In addition, considering the rotor wake interference and dynamic inflow effect, blade element theory or slipstream theory can be used for modeling, which will not be elaborated here.
[0038] 7. Coupling relationship between position and attitude: In this embodiment, position control can be achieved by adjusting the total lift T and the horizontal thrust component, with the horizontal thrust component generated by attitude tilt. Attitude control can be achieved by the torque τ generated by the rotor speed difference. x ,τ y ,τ z Adjust pitch, roll, and yaw angles. Attitude changes affect the direction of horizontal thrust, which in turn alters positional motion; conversely, position errors also affect attitude adjustment through the control law, forming a strong coupling relationship.
[0039] 8. Equilibrium of force and torque: Taking a quadcopter drone as an example, in hovering mode, the total lift T = mg, and the rotors rotate at equal speeds to counteract gravity and torque. During maneuvering flight, a resultant torque is generated through asymmetric rotor speeds, for example: Pitch-up motion: ; Yaw motion: .
[0040] 9. Nonlinearity and dynamic response: It should be noted that, due to the strong coupling and nonlinearity of the UAV system, and the first-order hysteresis characteristic (time constant) of the motor, The rotational inertia J of the machine affects the speed and stability of the system, so these characteristics need to be considered when designing the controller.
[0041] After the above dynamic model is constructed, rotor monitoring and flight control of the aircraft can be performed. Specifically, during flight, the working status of each rotor can be detected in real time through multi-sensor fusion technology. The sensors used in this application embodiment may include, but are not limited to, inertial measurement units (IMU), motor speed sensors, and current sensors.
[0042] In one embodiment, the flight control system can specifically perform rotor fault detection by the following method: collecting the motor speed and motor current of each rotor of the aircraft; determining the speed difference between the motor speed of each rotor and the preset normal speed, and the current difference between the motor current of each rotor and the preset normal current; when the speed difference is greater than a preset speed threshold and the current difference is greater than a preset current threshold, the rotor is determined to have failed.
[0043] Specifically, the flight control system can collect the motor speed of each rotor through various sensors. and motor current I i And calculate the speed deviation and current deviation of each rotor according to the following formulas [8] and [9] respectively: [8] [9] Where, ω nominal and I nominal These are the normal values for motor speed and current, respectively.
[0044] Then, when the following condition
[10] is met, the i-th rotor is determined to have malfunctioned:
[10] Where k1 and k2 represent threshold parameters, for example, , .
[0045] Once a rotor malfunction is detected through the above steps, the thrust data of the remaining normally functioning rotors is immediately obtained.
[0046] Step 12: Based on the thrust data of the remaining normally operating rotors obtained by performing Step 11, construct a flight optimization model; Specifically, in the embodiments of this application, the flight optimization model can be constructed with minimizing energy consumption and thrust deviation as the optimization objective, as shown in the following formula
[11] :
[11] Among them, T r This represents the thrust vector of the remaining rotor. Let λ be the average thrust of the remaining rotor, and λ1 and λ2 be weighting parameters. Constraints include the force and torque balance equations and the upper and lower limits of the motor thrust.
[0047] In one implementation, the constraints may include: 1. Force and torque balance constraints: The total lift and torque generated by the remaining rotor must be equal to the desired value in order to maintain the attitude and position of the UAV. The desired value is determined by the current flight mission and the reference trajectory. The specific constraint equations are shown in the following formula
[12] :
[12] Where, r rj Let be the position vector of the remaining rotor blades in the airframe coordinate system. This is the corresponding reverse torque.
[0048] The upper and lower limits of motor thrust are constrained by the physical limitations of the motor for each rotor.
[0049] The optimization model can be solved using quadratic programming or numerical optimization methods to obtain the initial thrust command for each remaining rotor, which serves as the basis for subsequent control.
[0050] Step 13: Determine the sliding mode control parameters based on the pre-trained deep reinforcement learning network and flight optimization model, and determine the sliding mode control law based on the sliding mode control parameters; Specifically, based on the initial thrust command determined in step 12, a deep reinforcement learning algorithm can be used to dynamically optimize the sliding mode control parameters to improve the system's robustness and dynamic performance under fault conditions. In one implementation, the Deep Deterministic Policy Gradient (DDPG) algorithm can be used.
[0051] Specifically, to achieve dynamic optimization of the switching gain, a DDPG network can be constructed as follows: Wherein, the state vector s in the state space t Including the drone's position error e and velocity error And the sliding surface value s, i.e. s t = [e, , s].
[0052] Action space: Action a t The adjustment value for switching gain, i.e., a t =Δη.
[0053] Reward function: Taking into account both tracking error and control signal jitter, the reward function is designed as follows
[13] :
[13] Where α and β are weighting coefficients, and Δu is the change in the control signal.
[0054] In this embodiment, sliding mode control can be performed using sliding surfaces and sliding mode control laws. Assuming that the dynamic model of the multi-rotor UAV in this embodiment is as shown in the following formula
[14] :
[14] Where x is the state vector, u is the control input, and d is the external disturbance.
[0055] In this embodiment of the application, the sliding surface s can be represented by the following formula
[15] :
[15] Where e is the tracking error and λ is the design parameter.
[0056] In this embodiment of the application, the sliding mode control law can be expressed by the following formula
[16] :
[16] Where, x d Let η be the desired state and η be the switching gain.
[0057] In actual flight control, the pre-trained strategy network can output the switching gain adjustment value Δη in real time according to the current state, thereby obtaining the current switching gain η, and then calculate the control input u according to the sliding mode control law, which is the final thrust command of each remaining normal operating rotor.
[0058] Step 14: Generate control commands based on the sliding mode control law obtained by executing step 13, and adjust the operating parameters of the normal rotor according to the control commands.
[0059] Specifically, the control input u obtained by executing step 13 can be converted into rotational speed commands for each rotor. The motor speed is then adjusted via the motor drive module, thereby adjusting the UAV's attitude and position. The control command update frequency is typically above 100Hz to ensure real-time performance.
[0060] Additionally, it should be noted that, to accommodate different fault severity levels, this application embodiment also includes a handling mode switching strategy for different faults. In this application embodiment, the flight control system can switch handling modes based on the number of remaining operational rotors. Specifically, the switching can be performed using the following methods: Sub-step 1401: Determine the number m of the remaining normal rotors; Sub-step 1402: When m is greater than or equal to the preset minimum redundancy threshold m red (For example, the minimum redundancy threshold m of a quadcopter) red When the value is 4), the normal flight mode is maintained, and traditional PID control is used.
[0061] Sub-step 1403, when m is less than the minimum redundancy threshold m red But greater than or equal to the preset safety threshold m safe (For example, the safety threshold m of a quadcopter) safe If it can be 2 or 3), then switch to fault-tolerant mode, start redundant power distribution and adaptive sliding mode control, that is, execute the above steps 12 to 14.
[0062] Sub-step 1404, when m is less than the safety threshold m safe In such cases, switch to emergency landing mode and execute a controlled landing procedure, such as slowly descending the altitude and finding a safe area to make an emergency landing.
[0063] Furthermore, to further improve the control effect, in this embodiment of the application, the flight control system can also adaptively adjust the control parameters according to the severity of the fault.
[0064] In one implementation, the severity S of the fault can be calculated by weighting multiple fault characteristic parameters based on the following formula
[17] :
[17] Among them, w i It is the weight of the i-th fault characteristic parameter, and satisfies .
[0065] In one implementation, the flight control system can also adaptively adjust control parameters according to the severity of the fault, including: 1. Adjustment of control parameters: Taking the classic PID (proportional-integral-derivative) controller as an example, the formula for calculating its control output u(t) is as follows
[18] :
[18] Where e(t) is the error signal (such as attitude error, position error, etc.), K p It is the proportionality coefficient, K i It is the integral coefficient, K d These are the differential coefficients.
[0066] After a fault occurs, these coefficients are adjusted according to the severity S of the fault. For example, a simple linear adjustment method can be expressed as the following formulas
[19] ~
[21] :
[19]
[20] [twenty one] in, , and This is an adjustment factor, which can be set according to the actual situation.
[0067] 2. Adjustment of motor thrust distribution: Specifically, multi-rotor UAVs achieve attitude and position control by adjusting the thrust of each motor. In the event of a malfunction, the motor thrust needs to be reallocated. Assume that under normal circumstances, the thrust of the j-th motor is T. j After the failure, the new thrust T jnew Adjustments can be made based on the severity and type of the fault.
[0068] For example, when one motor partially fails, the failure can be compensated for by increasing the thrust of other motors. Let the faulty motor be numbered k, and its fault severity be S. k Then the thrust adjustment formula for other motors can be as follows
[22] : [twenty two] Where m is the total number of motors, It is the thrust compensation coefficient.
[0069] 3. State estimation adjustment: In case of sensor failure, adjust the measurement noise covariance R or process noise covariance Q in the Kalman filter. The prediction and update steps of the Kalman filter are shown in the following formula: The prediction steps include the following formulas
[23] ~
[24] : [twenty three] [twenty four] The update steps include the following formulas
[25] ~
[29] :
[25]
[26]
[27]
[28]
[29] in, Here, P is the state estimate, F is the estimation error covariance, B is the control input matrix, u is the control input, Q is the process noise covariance, z is the measurement value, H is the measurement matrix, R is the measurement noise covariance, y is the residual, S is the innovation covariance, and K is the Kalman gain.
[0070] When a sensor fails, the measurement noise covariance R can be adjusted according to the severity of the failure, as shown in the following formula
[30] :
[30] in, This is for adjusting the coefficient.
[0071] Furthermore, it should be noted that, to meet real-time requirements, the flight control system provided in this application embodiment can be deployed on a heterogeneous hardware platform composed of an ARM processor and an FPGA. Wherein: ARM processor: Responsible for tasks such as system management, data acquisition, fault detection, optimization model building, and communication with ground stations.
[0072] Specifically, in this embodiment, the ARM can collect sensor data such as IMU, rotation speed, and current, and execute a fault detection algorithm; when a fault is detected, it constructs a flight optimization model and solves for the initial thrust command; it sends the UAV status information and the initial thrust command to the FPGA; it receives the final control command returned by the FPGA and outputs it through the motor drive module.
[0073] In this embodiment, the FPGA is responsible for high-speed parallel computing, particularly the inference of the DDPG network and the calculation of the sliding mode control law. The FPGA can internally implement the forward propagation of the policy network and value network in parallel, quickly outputting switching gain adjustment values and control inputs. Data exchange between the FPGA and the ARM is achieved through a high-speed communication interface (such as SPI, UART, or Ethernet).
[0074] It should also be noted that, to improve system reliability, a dual-loop hardware redundancy architecture can be constructed in this embodiment, where two independent control loops operate simultaneously. Each loop contains an independent ARM and FPGA, but only the output of the main loop is used to control the drone. When the main loop fails, the system automatically switches to the backup loop.
[0075] Furthermore, for multi-drone formation flight scenarios, this application embodiment also provides a distributed cooperative control scheme. Each drone exchanges state information with its neighbors through a communication module, and a consensus algorithm is used to achieve formation state synchronization. The specific steps are as follows: Sub-step a: Establish a communication topology graph G, which consists of a node set V and an edge set E. V is the set of UAV nodes, and E is the set of edges, representing communication connections.
[0076] Sub-step b involves updating the state of each UAV using the first-order consensus protocol shown in formula
[31] :
[31] Where, N i It is the set of neighbors of drone i, a ij It is an adjacency matrix element, when At that time, a ij =1, otherwise, a ij =0.
[0077] Using a second-order consensus protocol, let x i For position state, v i For the velocity state, the state update of each UAV i is as shown in the following formula
[32] :
[32] Each drone obtains its own status through sensors (such as GPS and visual sensors). i It exchanges status information with neighboring drones via a communication module. It updates its own status according to a consensus protocol, iterating until the position error is less than 0.2m. Simultaneously, it periodically monitors communication quality and distance between drones. For example, if two drones are out of communication range, their connection is broken; if a new drone enters the communication range, a new connection is established.
[0078] When a topology change is detected, update the adjacency matrix A=[a ij ], and recalculate the neighbor information in the consensus protocol.
[0079] Using the flight control method provided in this application, when it is determined that the aircraft has a rotor failure based on the collected operating data of each rotor, the thrust data of the remaining normal rotors can be obtained first. Based on the thrust data, a flight optimization model is constructed. Based on the pre-trained deep reinforcement learning network and the flight optimization model, sliding mode control parameters and sliding mode control laws are determined. Then, control commands can be generated based on the sliding mode control laws, and the operating parameters of the normal rotors can be adjusted according to the control commands. The flight control method provided in this application has two advantages. First, by collecting rotor operation data in real time, it can quickly identify rotor faults during flight and dynamically reconstruct the control model based on the thrust data of the remaining normal rotors. Compared with existing solutions that rely on preset routes or historical data, the method provided in this application can immediately initiate fault-tolerant control after a fault occurs, avoiding the risk of flight attitude instability or loss of control due to response delays, and significantly improving the survivability of the UAV under sudden faults. Second, the flight optimization model constructed in this application aims to minimize energy consumption and thrust deviation, and solves the optimal thrust distribution scheme under torque balance constraints and motor thrust limiting constraints. By minimizing the thrust deviation, it ensures that the resultant force and resultant torque generated by the remaining rotors can accurately compensate for the power lost by the faulty rotor, maintaining the attitude stability of the UAV. Furthermore, by minimizing energy consumption, secondary failures caused by overload of the remaining rotor can be avoided, extending the UAV's loiter time and ensuring mission completion or safe return. Finally, by introducing a strategy combining deep reinforcement learning networks with sliding mode control, the switching gain adjustment value of sliding mode control can be adaptively output according to the real-time state of the aircraft and the optimization model. This allows the control parameters to be dynamically optimized according to changes in fault type, fault severity, and flight environment. At the same time, precise control commands are generated through sliding mode control laws. When the UAV experiences asymmetric faults or is disturbed by environmental factors such as strong winds, sliding mode control can force the system state to move along the sliding surface, ensuring that attitude angle and position tracking errors converge quickly to zero. Compared with traditional PID control, which is prone to oscillation or overshoot after a fault, the flight control method provided in this application has higher control accuracy and response speed.
[0080] In one embodiment, this application also provides a flight control device to address the problem that existing technologies cannot ensure stable flight and successful mission completion of multi-rotor UAVs after partial rotor failure without adding additional hardware. A schematic diagram of the specific structure of the flight control device is shown below. Figure 2 As shown, it includes: a fault detection unit 21, an optimization model construction unit 22, a parameter generation unit 23, and a flight control unit 24.
[0081] The fault detection unit 21 is used to obtain the thrust data of the remaining normal rotors when it is determined that the aircraft has a rotor fault based on the collected operating data of each rotor of the aircraft. The optimization model construction unit 22 is used to construct a flight optimization model based on the thrust data, wherein the flight optimization model takes minimizing energy consumption and thrust deviation as the optimization objective and satisfies torque balance constraints and motor thrust limiting constraints as the constraint conditions; The parameter generation unit 23 is used to determine sliding mode control parameters based on the pre-trained deep reinforcement learning network and the flight optimization model, and to determine the sliding mode control law based on the sliding mode control parameters, wherein the sliding surface in the sliding mode control parameters is constructed based on the tracking error; The flight control unit 24 is used to generate control commands according to the sliding mode control law and adjust the operating parameters of the normal rotor according to the control commands.
[0082] In one embodiment, the fault detection unit 21 is specifically used to: collect the motor speed and motor current of each rotor of the aircraft; determine the speed difference between the motor speed of each rotor and the preset normal speed, and the current difference between the motor current of each rotor and the preset normal current; when the speed difference is greater than a preset speed threshold and the current difference is greater than a preset current threshold, determine that the rotor has malfunctioned.
[0083] In one implementation, the deep reinforcement learning network is constructed based on the deep deterministic policy gradient algorithm. The deep reinforcement learning network includes a policy subnetwork and a value subnetwork, wherein: the policy subnetwork is used to determine the adjustment value of the sliding mode control parameters based on the current flight data of the aircraft; the value subnetwork is used to update the parameters of the policy subnetwork based on the current flight data of the aircraft and the received flight control commands.
[0084] In one embodiment, a mode switching unit is further included, specifically configured to: determine the current number of remaining normal rotors; maintain normal flight mode when the number of remaining normal rotors is greater than or equal to a preset minimum redundancy threshold; switch to fault-tolerant mode and execute a redundant power distribution procedure when the number of remaining normal rotors is less than the minimum redundancy threshold but greater than or equal to a preset safety threshold; and switch to emergency landing mode and execute a landing procedure when the number of remaining normal rotors is less than the safety threshold.
[0085] In one embodiment, the system further includes a mechanical model building unit, specifically used to: build a dynamic model of the aircraft, wherein the dynamic model includes at least one of the following: position dynamics equations, attitude dynamics equations, force and torque balance equations, rotor dynamics model, aerodynamics and disturbance model, coordinate system and attitude representation, coupling relationship between position and attitude, and balance relationship between force and torque.
[0086] Using the flight control device provided in this application embodiment, when it is determined that the aircraft has a rotor failure based on the collected operating data of each rotor, the thrust data of the remaining normal rotors can be obtained first. Based on the thrust data, a flight optimization model is constructed. Based on the pre-trained deep reinforcement learning network and the flight optimization model, sliding mode control parameters and sliding mode control law are determined. Then, control commands can be generated based on the sliding mode control law, and the operating parameters of the normal rotors can be adjusted according to the control commands. The flight control method provided in this application has two advantages. First, by collecting rotor operation data in real time, it can quickly identify rotor faults during flight and dynamically reconstruct the control model based on the thrust data of the remaining normal rotors. Compared with existing solutions that rely on preset routes or historical data, the method provided in this application can immediately initiate fault-tolerant control after a fault occurs, avoiding the risk of flight attitude instability or loss of control due to response delays, and significantly improving the survivability of the UAV under sudden faults. Second, the flight optimization model constructed in this application aims to minimize energy consumption and thrust deviation, and solves the optimal thrust distribution scheme under torque balance constraints and motor thrust limiting constraints. By minimizing the thrust deviation, it ensures that the resultant force and resultant torque generated by the remaining rotors can accurately compensate for the power lost by the faulty rotor, maintaining the attitude stability of the UAV. Furthermore, by minimizing energy consumption, secondary failures caused by overload of the remaining rotor can be avoided, extending the UAV's loiter time and ensuring mission completion or safe return. Finally, by introducing a strategy combining deep reinforcement learning networks with sliding mode control, the switching gain adjustment value of sliding mode control can be adaptively output according to the real-time state of the aircraft and the optimization model. This allows the control parameters to be dynamically optimized according to changes in fault type, fault severity, and flight environment. At the same time, precise control commands are generated through sliding mode control laws. When the UAV experiences asymmetric faults or is disturbed by environmental factors such as strong winds, sliding mode control can force the system state to move along the sliding surface, ensuring that attitude angle and position tracking errors converge quickly to zero. Compared with traditional PID control, which is prone to oscillation or overshoot after a fault, the flight control method provided in this application has higher control accuracy and response speed.
[0087] Figure 3 This is a schematic diagram of the structure of an electronic device according to an embodiment of this application. Please refer to it. Figure 3At the hardware level, the electronic device includes a processor, and optionally also includes an internal bus, a network interface, and memory. The memory may include main memory, such as high-speed random-access memory (RAM), or non-volatile memory, such as at least one disk drive. Of course, the electronic device may also include other hardware required for other business operations.
[0088] The processor, network interface, and memory can be interconnected via an internal bus, which can be an ISA (Industry Standard Architecture) bus, a PCI (Peripheral Component Interconnect) bus, or an EISA (Extended Industry Standard Architecture) bus, etc. This bus can be divided into address bus, data bus, control bus, etc. For ease of representation, Figure 3 The symbol is represented by a single double-headed arrow, but this does not mean that there is only one bus or one type of bus.
[0089] Memory is used to store programs. Specifically, programs may include program code, which includes computer operation instructions. Memory may include main memory and non-volatile memory, and provides instructions and data to the processor.
[0090] The processor reads the corresponding computer program from non-volatile memory into main memory and then runs it, forming the flight control system at the logical level. The processor executes the program stored in memory and specifically performs the following operations: When a rotor failure is determined based on the collected operating data of each rotor of the aircraft, the thrust data of the remaining normal rotors is obtained. Based on this thrust data, a flight optimization model is constructed, where the optimization objective is to minimize energy consumption and thrust deviation, and the constraints are torque balance and motor thrust limiting. Sliding mode control parameters are determined based on a pre-trained deep reinforcement learning network and the flight optimization model, and a sliding mode control law is determined based on these parameters, where the sliding surface in the sliding mode control parameters is constructed based on the tracking error. Control commands are generated based on the sliding mode control law, and the operating parameters of the normal rotors are adjusted according to these control commands.
[0091] The above is as stated in this application. Figure 3The method executed by the flight control electronic device disclosed in the illustrated embodiment can be applied to a processor or implemented by a processor. The processor may be an integrated circuit chip with signal processing capabilities. During implementation, each step of the above method can be completed by integrated logic circuits in the processor's hardware or by instructions in software form. The processor can be a general-purpose processor, including a Central Processing Unit (CPU), a Network Processor (NP), etc.; it can also be a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), a Field-Programmable Gate Array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. It can implement or execute the methods, steps, and logic block diagrams disclosed in the embodiments of this application. The general-purpose processor can be a microprocessor or any conventional processor. The steps of the method disclosed in the embodiments of this application can be directly embodied in the execution of a hardware decoding processor, or executed by a combination of hardware and software modules in the decoding processor. The software module can reside in a mature storage medium in the field, such as random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, or registers. This storage medium is located in memory, and the processor reads information from the memory and, in conjunction with its hardware, completes the steps of the above method.
[0092] Of course, in addition to software implementation, the electronic device of this application does not exclude other implementation methods, such as logic devices or a combination of hardware and software, etc. In other words, the execution subject of the following processing flow is not limited to each logic unit, but can also be hardware or logic devices.
[0093] This application also proposes a computer-readable storage medium that stores one or more programs, the programs including instructions that, when executed by a portable electronic device including multiple applications, enable the portable electronic device to perform... Figure 1 The flight control method of the illustrated embodiment is specifically used to perform the following operations: When a rotor failure is determined based on the collected operating data of each rotor of the aircraft, the thrust data of the remaining normal rotors is obtained. Based on this thrust data, a flight optimization model is constructed, where the optimization objective is to minimize energy consumption and thrust deviation, and the constraints are torque balance and motor thrust limiting. Sliding mode control parameters are determined based on a pre-trained deep reinforcement learning network and the flight optimization model, and a sliding mode control law is determined based on these parameters, where the sliding surface in the sliding mode control parameters is constructed based on the tracking error. Control commands are generated based on the sliding mode control law, and the operating parameters of the normal rotors are adjusted according to these control commands.
[0094] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.
[0095] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.
[0096] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.
[0097] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.
[0098] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.
[0099] Memory may include non-persistent storage in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of computer-readable media.
[0100] Computer-readable media includes both permanent and non-permanent, removable and non-removable media that can store information using any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic magnetic disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include transient computer-readable media, such as modulated data signals and carrier waves.
[0101] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.
[0102] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.
[0103] The above description is merely an embodiment of this application and is not intended to limit the scope of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of the claims of this application.
Claims
1. A flight control method, characterized in that, include: When a rotor failure is determined based on the collected operating data of each rotor of the aircraft, the thrust data of the remaining normal rotors is obtained. Based on the thrust data, a flight optimization model is constructed, wherein the flight optimization model takes minimizing energy consumption and thrust deviation as the optimization objective, and satisfies torque balance constraints and motor thrust limiting constraints as the constraints. Based on the pre-trained deep reinforcement learning network and the flight optimization model, sliding mode control parameters are determined, and a sliding mode control law is determined based on the sliding mode control parameters, wherein the sliding surface in the sliding mode control parameters is constructed based on the tracking error; Control commands are generated based on the sliding mode control law, and the operating parameters of the normal rotor are adjusted according to the control commands.
2. The method according to claim 1, characterized in that, The determination of a rotor malfunction based on collected operational data from each rotor of the aircraft specifically includes: The motor speed and motor current of each rotor of the aircraft were collected; The speed difference between the motor speed of each rotor and the preset normal speed, and the current difference between the motor current of each rotor and the preset normal current are determined respectively. When the speed difference is greater than a preset speed threshold and the current difference is greater than a preset current threshold, it is determined that the rotor has malfunctioned.
3. The method according to claim 1, characterized in that, The deep reinforcement learning network is constructed based on the deep deterministic policy gradient algorithm, and the deep reinforcement learning network includes a policy subnetwork and a value subnetwork, wherein: The strategy subnetwork is used to determine the adjustment value of the sliding mode control parameters based on the current flight data of the aircraft; The value subnetwork is used to update the parameters of the strategy subnetwork based on the current flight data of the aircraft and the received flight control commands.
4. The method according to claim 1, characterized in that, Also includes: Determine the number of remaining normal rotor blades; When the number of remaining normal rotors is greater than or equal to the preset minimum redundancy threshold, maintain normal flight mode; When the number of remaining normal rotors is less than the minimum redundancy threshold and greater than or equal to the preset safety threshold, switch to fault-tolerant mode and execute the redundant power distribution procedure. When the number of remaining normal rotors is less than the safety threshold, switch to emergency landing mode and execute landing procedure.
5. The method according to claim 1, characterized in that, The method is applied to a flight control system that includes an ARM processor and a field-programmable gate array (FPGA), wherein: The ARM processor is used to collect operating data, determine rotor faults, build flight optimization models, and send the control commands to the motor drive module of the aircraft. The FPGA is used to interact with the ARM processor and determine sliding mode control parameters and sliding mode control laws based on the pre-trained deep reinforcement learning network and the flight optimization model.
6. The method according to claim 1, characterized in that, Before collecting the operating data of each rotor, the following steps are also included: Construct a dynamic model of the aircraft, wherein the dynamic model includes at least one of the following: position dynamics equation, attitude dynamics equation, force and torque balance equation, rotor dynamics model, aerodynamics and disturbance model, coordinate system and attitude representation, coupling relationship between position and attitude, and balance relationship between force and torque.
7. A flight control device, characterized in that, include: The fault detection unit is used to obtain the thrust data of the remaining normal rotors when it is determined that the aircraft has a rotor fault based on the collected operating data of each rotor of the aircraft. An optimization model building unit is used to build a flight optimization model based on the thrust data, wherein the flight optimization model takes minimizing energy consumption and thrust deviation as the optimization objective and satisfies torque balance constraints and motor thrust limiting constraints as the constraint conditions. The parameter generation unit is used to determine sliding mode control parameters based on the pre-trained deep reinforcement learning network and the flight optimization model, and to determine the sliding mode control law based on the sliding mode control parameters, wherein the sliding surface in the sliding mode control parameters is constructed based on the tracking error; The flight control unit is used to generate control commands based on the sliding mode control law, and to adjust the operating parameters of the normal rotor based on the control commands.
8. A flight control device, comprising: processor; as well as A memory configured to store computer-executable instructions, which, when executed, cause the processor to perform the following operations: When a rotor failure is determined based on the collected operating data of each rotor of the aircraft, the thrust data of the remaining normal rotors is obtained. Based on the thrust data, a flight optimization model is constructed, wherein the flight optimization model takes minimizing energy consumption and thrust deviation as the optimization objective, and satisfies torque balance constraints and motor thrust limiting constraints as the constraints. Based on the pre-trained deep reinforcement learning network and the flight optimization model, sliding mode control parameters are determined, and a sliding mode control law is determined based on the sliding mode control parameters, wherein the sliding surface in the sliding mode control parameters is constructed based on the tracking error; Control commands are generated based on the sliding mode control law, and the operating parameters of the normal rotor are adjusted according to the control commands.
9. A computer-readable storage medium storing one or more programs, which, when executed by an electronic device including a plurality of applications, cause the electronic device to perform the flight control method as described in any one of claims 1-6.
10. A computer program product, characterized in that, It includes a computer program that, when executed by a processor, implements the flight control method as described in any one of claims 1-6.