Event-triggered adaptive guidance method for pressing section of aircraft
By employing an event-triggered adaptive guidance method in the aircraft's down-thrust phase, and utilizing proportional guidance and adaptive dynamic programming to optimize the guidance controller parameters, the problems of decreased guidance accuracy and neural network overfitting caused by nonlinear effects were solved, thereby improving guidance accuracy and combat effectiveness.
Patent Information
- Application Number
- CN202510980675.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-07-16
- Publication Date
- 2025-10-31
AI Technical Summary
Existing guidance methods for the down-phase of aircraft suffer from decreased guidance accuracy when faced with nonlinear effects, and adaptive dynamic programming methods are prone to overfitting and cannot adapt to rapidly changing battlefield environments.
An event-triggered adaptive guidance method is adopted to calculate the aircraft velocity change information in the ballistic coordinate system. The proportional guidance method and adaptive dynamic programming are used to optimize the guidance controller parameters. The line-of-sight angle change rate is used as an input term, and a neural network is used for strategy evaluation and improvement to avoid overfitting of the neural network.
It improves guidance accuracy and aircraft combat effectiveness, enhances adaptability to highly dynamic environments, avoids overfitting problems in neural networks, and improves the stability of the controller and the effectiveness of parameters.
Smart Images
Figure CN120871918A_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of flight guidance technology, specifically relating to an event-triggered adaptive guidance method for the down-pressure phase of an aircraft. Background Technology
[0002] Attacks during the descent phase of an aircraft can improve target accuracy. High-altitude gliding avoids air defense interception, reduces the impact of terrain changes, and makes it easier to obtain target information. If intercepted, secondary damage is less likely. Overall, this improves strike accuracy and combat effectiveness, demonstrating excellent application prospects. Currently, guidance and control engineering primarily employs improved methods based on proportional guidance or sliding diaphragm control. These methods offer stable control structures, are easy to implement, and provide high guidance accuracy. According to the literature, existing research mainly involves adding angle of attack or time constraints during the descent phase, or performing trajectory planning to optimize trajectory tracking control. Many studies also integrate guidance technology with inertial navigation systems to obtain more accurate target information globally, thereby improving guidance accuracy.
[0003] Current down-phase guidance and control methods mainly suffer from the following problems: Down-phase interception is a crucial aspect of air defense operations. When an aircraft is in the down-phase, its maneuverability is limited, making it most vulnerable to interception and requiring higher precision guidance strategies; the environment changes rapidly during down-phase flight, and changes in altitude and speed affect the aircraft's aerodynamic parameters. Fixed guidance law parameters can lead to decreased guidance accuracy and reduced combat effectiveness; most studies do not consider the nonlinear characteristics of the aircraft. During flight, nonlinear effects caused by aerodynamic heating and high-frequency vibrations can affect the stability of the controller; adaptive dynamic programming methods use neural networks for optimal parameter fitting, which is highly dependent on training data. If the training data changes too little, overfitting can easily occur, making it unable to adapt to new battlefield environments. Periodic triggering may result in alternating periods of good and poor parameters. Summary of the Invention
[0004] The technical problem solved by this invention is to address the issue of decreased guidance accuracy caused by nonlinear effects during the down-phase flight of an aircraft. This invention provides an event-triggered adaptive guidance method for the down-phase flight of an aircraft. The method calculates the aircraft's velocity change information in a ballistic coordinate system and solves the guidance law in a relative motion coordinate system between the aircraft and the target. The guidance controller employs a proportional guidance method, with the proportional coefficient optimized through adaptive dynamic programming. The line-of-sight angle change rate is used as the input, and the missile's motion state is used as the output. The controller parameters are adaptively optimized to solve the problems of decreased guidance accuracy and reduced combat effectiveness during the down-phase flight.
[0005] To address the aforementioned technical problems, this invention discloses an event-triggered adaptive guidance method for the down-short phase of an aircraft. The specific steps of the method are as follows:
[0006] S1. When the aircraft enters the target radius of 100 kilometers, the aircraft ends the gliding phase and enters the terminal guidance down-pressure phase.
[0007] S2. The line-of-sight angular velocity information is obtained through the seeker. The guidance controller adopts the proportional guidance method to calculate the velocity change information of the aircraft in the ballistic coordinate system and solve the guidance law in the relative motion coordinate system of the aircraft and the target.
[0008] S3. The overload command is transmitted to the aircraft. The aircraft's attitude is adjusted and guided by the servo motor. It is determined whether the target has entered the damage radius. If yes, the workflow ends; otherwise, proceed to the next step.
[0009] S4. The event triggering mechanism is judged in the onboard computer to avoid overfitting of the adaptive dynamic programming method. If the number of triggers in the past 3 seconds exceeds 50% of the total number of judgments, the process continues. If the event is not triggered, return to step S2 and repeat the flight process.
[0010] S5. Conduct strategy evaluation and strategy improvement based on the value function constructed from the aircraft's line-of-sight angular velocity and state information;
[0011] S6. Evaluate the parameters of the neural network and the execution neural network, and calculate the approximation error. If the approximation error is less than 0.01, then jump to step S8. If not, continue until the change in the approximation error is less than 0.01 for two consecutive times.
[0012] S7. Based on the approximation error, update the parameters of the evaluation neural network, evaluate the network parameter update rule, update the parameters of the control neural network, and execute the network parameter update rule. If the number of iterations of the evaluation network exceeds the given number, continue; otherwise, jump to step S4.
[0013] S8. Replace the guidance controller parameters with the updated control parameters. If the flight termination condition is not reached, proceed to step S2. If the termination condition is reached, the aircraft ends its workflow.
[0014] Furthermore, the method for calculating the line-of-sight angular velocity information is as follows:
[0015]
[0016] Where R represents the relative distance between the aircraft and the target, Let R be the differential, V represent the relative velocity, θ represent the aircraft trajectory angle, and q represent the line-of-sight angular velocity. The derivative of q;
[0017] An excitation function is added to the line-of-sight angular velocity to prevent the adaptive dynamic programming method from stopping updates when the rate of change of the line-of-sight angular velocity becomes static. The line-of-sight angular velocity with the added excitation function is:
[0018]
[0019] Where e represents the activation function;
[0020] The method for solving the guidance law is as follows:
[0021]
[0022] Where n represents overload and K represents proportional guidance coefficient. The line-of-sight angular velocity change rate is transmitted to the guidance controller, and the proportional guidance law is solved by the onboard computer.
[0023] Furthermore, the aircraft attitude adjustment includes transmitting overload commands to the autopilot, which then outputs control to the servos to adjust the missile's attitude, calculates the aerodynamic forces under the current attitude, obtains the aerodynamic parameters using a lookup table method, and transmits the overload commands to the aircraft for guidance.
[0024]
[0025] Where the subscripts x, y, z represent the three-axis components in the ballistic coordinate system, C represents the aerodynamic coefficient, Q represents the dynamic pressure, S represents the characteristic area, m represents the mass of the aircraft, and g represents the gravitational acceleration.
[0026] Furthermore, the triggering condition determined by the event triggering mechanism is:
[0027]
[0028] Furthermore, the strategy evaluation method is as follows: input the line-of-sight angular velocity information and the aircraft flight state information of the previous fixed cycle into the onboard computer, use an adaptive dynamic programming method to train the guidance controller parameters, construct an evaluation function at the cost of the aircraft's line-of-sight angular velocity information and state information during the process, construct an evaluation neural network, use the constructed evaluation neural network to fit the evaluation function, and use strategy evaluation to estimate the value function of the current state.
[0029] The strategy improvement method is as follows: based on the strategy evaluation, the strategy is improved to generate a new strategy. The trained evaluation neural network is used as input to calculate the strategy in this state. An execution neural network is constructed, and the execution neural network is used to fit the aircraft model to obtain the execution neural network parameters of the fitted aircraft. The execution neural network parameters are then used to calculate the strategy result.
[0030] Furthermore, the evaluation function is:
[0031] Where V(x(·)) represents the evaluation function, x represents the state, Q and R are positive definite matrices, and u represents the control quantity.
[0032] Furthermore, the construction of the evaluation neural network includes:
[0033] The neural network is configured with four layers and trained using the backpropagation (BP) algorithm. Based on the gradient descent strategy, the learning rate is set to 0.01. The input to the training set is line-of-sight angular velocity information, and the output is an evaluation metric. It has two hidden layers, each containing six hidden neurons.
[0034] The constructed evaluation neural network fits the evaluation function, and the input of the neural network is the basis function.
[0035]
[0036] in, Let φ1(x) represent the neural network fitting parameters, and let φ1(x) represent the basis functions.
[0037] The strategy evaluation method is as follows: estimating the value function of the current state.
[0038]
[0039] V (i) (0)=0
[0040] Where f represents the state transition matrix and g represents the control matrix.
[0041] Furthermore, the method for improving the strategy is as follows:
[0042] Based on the strategy evaluation, the strategy is improved to generate a new strategy.
[0043]
[0044] The trained evaluation neural network is used as input to calculate the policy in this state.
[0045]
[0046] A three-layer neural network was constructed and trained using the backpropagation (BP) algorithm. Based on a gradient descent strategy, the training set input consisted of line-of-sight angular velocity information, and the output was the aircraft state information [P, V]. A [θ], the aircraft model is fitted with an execution neural network to obtain the parameters of the execution neural network for the fitted aircraft. The results of the strategy are calculated using the execution neural network parameters.
[0047]
[0048] in, This represents the parameters fitted to the neural network.
[0049] Furthermore, the approximation error e is calculated as follows:
[0050] Furthermore, the evaluation network parameter update rule is as follows:
[0051]
[0052] Where α1 is the learning rate;
[0053] Update the parameters of the control neural network and execute the network parameter update rules:
[0054]
[0055] Where F1 and F2 are positive definite matrices, the updated control parameters Replace the guidance controller parameter K.
[0056] The present invention has the following advantages:
[0057] (1) The present invention provides an event-triggered adaptive guidance method for the down-pressure phase of an aircraft. The flight environment is highly dynamic and the nonlinear terms of the aircraft are greatly affected. This technology dynamically optimizes the controller parameters during flight, which improves the guidance accuracy and enhances the aircraft's capabilities compared to the traditional fixed-parameter guidance controller.
[0058] (2) An event-triggered adaptive guidance method for the down-pressure section of an aircraft according to the present invention has a periodic triggering mechanism. The use of neural networks is prone to overfitting problems. Event triggering selects a suitable training dataset, which avoids the overfitting problem of neural networks and provides controller parameter update instructions, thereby improving the adaptability to highly dynamic environments.
[0059] (3) In the event-triggered adaptive guidance method for the down-short section of an aircraft, under the condition of sudden change in guidance state, the quantitative error event triggering is prone to misjudgment and cannot adapt to the current air combat environment. Based on the Lyapunov event triggering mechanism, the appropriate time to trigger training is judged according to the stability principle, thereby improving the effectiveness of training parameters. Attached Figure Description
[0060] To more clearly illustrate the embodiments of the present invention
[0061] The following will briefly introduce the accompanying drawings used in the description of the embodiments or existing technologies, or the technical solutions in the prior art. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. It should be noted that the drawings are not necessarily drawn to scale for clarity and ease of explanation.
[0062] Figure 1 This is a schematic diagram of the event-triggered adaptive guidance method for the down-phase of an aircraft according to the present invention.
[0063] Figure 2 This is a schematic diagram of a combat scenario for the event-triggered adaptive guidance method for the down-phase of an aircraft according to the present invention.
[0064] Figure 3 This is a schematic diagram of the adaptive dynamic programming method for the event-triggered adaptive guidance method of an aircraft in its down-phase phase, as described in this invention.
[0065] Figure 4 The diagram shows the guidance control loop of the event-triggered adaptive guidance method for the down-slope phase of an aircraft according to the present invention.
[0066] Specific implementation methods
[0067] The invention will now be further described with reference to the accompanying drawings.
[0068] The event-triggered adaptive guidance method for the down-phase flight of an aircraft of the present invention includes: a relative motion model of the aircraft and the target, a guidance controller design, aerodynamic calculation, an event triggering mechanism, and adaptive dynamic programming parameter optimization. In order to address the problem of decreased guidance accuracy caused by nonlinear effects during the down-phase flight of the aircraft, a proportional guidance method is adopted to adaptively optimize the controller parameters in order to eliminate the overfitting problem of the neural network in the adaptive dynamic programming.
[0069] like Figure 1 A schematic diagram of the event-triggered adaptive guidance method for the down-phase of an aircraft according to the present invention and Figure 2 The event-triggered adaptive guidance method for aircraft in the down-phase phase of this invention is illustrated in the operational scenario diagram. The process of the adaptive guidance method for aircraft in the down-phase phase mainly includes: down-phase activation, obtaining line-of-sight angular velocity information, guidance law generation, guidance controller, flight status update, target entry damage radius judgment, event triggering mechanism judgment, strategy evaluation, strategy improvement, and approximation error judgment.
[0070] like Figure 3The principle diagram of the adaptive dynamic programming method for the event-triggered adaptive guidance method of the aircraft down-short section of the present invention is shown. The guidance controller parameter optimization is achieved by fitting an intelligent agent through an evaluation network and an execution network, and the optimal control output is sent to the aircraft. The error is fed back to the network, so that the network parameters are iteratively updated until the optimal strategy is obtained.
[0071] In this embodiment, the initial position is set to (0, 49000, 200) m, the velocity is set to 170 m / s, the trajectory angle is set to (0°, 0°), the proportional guidance coefficient is 5, and the aerodynamic coefficient is set to (-0.149, 0.268, -0.268).
[0072] Control excitation approximation error set to
[0073] e=0.1sin2πt+sin4πt+sin5πt
[0074] The flight mission will proceed from the starting position to the target point, which is located at (25000, 0, 5000) m.
[0075] The evaluation function is set to
[0076] R = 5
[0077] In the neural network, the basis function is set to φ1(x) = 1, x, the learning rate is set to 0.01, and there is only one hidden layer containing 6 hidden neurons.
[0078] like Figure 4 The guidance control loop diagram of the event-triggered adaptive guidance method for the down-phase of an aircraft in this invention is shown. When executing a flight mission, the seeker obtains target information and the aircraft's own state information to obtain line-of-sight angle information. The onboard computer calculates the controller instructions and transmits the instructions to the controller to provide overload to the aircraft, thereby realizing the change of the aircraft's state.
[0079] During flight, the onboard computer determines the event triggering conditions. When the triggering conditions are met, it triggers controller parameter training and updates the controller parameters.
[0080] The event-triggered adaptive guidance method for the down-phase of an aircraft provided by the invention dynamically optimizes controller parameters and selects appropriate training datasets during flight to avoid overfitting of the neural network. Compared with the traditional fixed-parameter guidance method, it effectively improves guidance accuracy, aircraft capability, and adaptability to highly dynamic environments, and greatly enhances the effectiveness of training parameters.
[0081] While the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the invention. Any person skilled in the art can make possible variations and modifications to the technical solutions of the present invention using the disclosed methods and techniques without departing from the spirit and scope of the invention. Therefore, any simple modifications, equivalent changes, and alterations made to the above embodiments based on the technical essence of the present invention, without departing from the scope of the invention, are all within the protection scope of the present invention. Content not described in detail in this specification is common knowledge to those skilled in the art.
Claims
1. An event-triggered adaptive guidance method for the down-phase of an aircraft, characterized in that, Includes the following steps: S1. Enter the terminal guidance down-pressure phase working state; S2. Obtain line-of-sight angular velocity information, and the guidance controller uses the proportional guidance method to solve the guidance law in the relative motion coordinate system of the aircraft and the target. S3. Use the aircraft's attitude adjustment to guide and determine whether the target has entered the damage radius. If yes, end the workflow; otherwise, proceed to the next step. S4. Determine the event triggering mechanism to avoid overfitting of the adaptive dynamic programming method. If the number of triggers in the past 3 seconds exceeds 50% of the total number of judgments, then trigger the continuation process. If the event is not triggered, return to step S2. S5. Conduct strategy evaluation and strategy improvement based on the value function constructed from the aircraft's line-of-sight angular velocity and state information; S6. Evaluate the parameters of the neural network and the execution neural network, and calculate the approximation error. If the approximation error is less than 0.01, then jump to step S8. If not, continue until the change in the approximation error is less than 0.01 for two consecutive times. S7. Based on the approximation error, update the parameters of the evaluation neural network, implement the parameter update rule for the evaluation neural network, update the parameters of the control neural network, and execute the neural network parameter update rule. If the number of iterations of the evaluation neural network exceeds the given number, continue; otherwise, jump to step S4. S8. Replace the guidance controller parameters with the updated control parameters. If the flight termination condition is not reached, proceed to step S2. If the termination condition is reached, the aircraft ends its workflow.
2. The event-triggered adaptive guidance method for the down-phase of an aircraft as described in claim 1, characterized in that, The method for calculating the line-of-sight angular velocity information in step S2 is as follows: Where R represents the relative distance between the aircraft and the target, Let R be the differential, V represent the relative velocity, θ represent the aircraft trajectory angle, and q represent the line-of-sight angular velocity. The derivative of q; An excitation function is added to the line-of-sight angular velocity to prevent the adaptive dynamic programming method from stopping updates when the rate of change of the line-of-sight angular velocity becomes static. The line-of-sight angular velocity with the added excitation function is: Where e represents the activation function; The method for solving the guidance law mentioned in step S2 is as follows: Where n represents overload and K represents proportional guidance coefficient. The line-of-sight angular velocity change rate is transmitted to the guidance controller, and the proportional guidance law is solved by the onboard computer.
3. The event-triggered adaptive guidance method for the down-phase of an aircraft as described in claim 1, characterized in that: Step S3, which involves adjusting the aircraft's attitude, transmits the overload command to the autopilot. The autopilot then outputs control to the servo motors to adjust the missile's attitude and calculates the aerodynamic forces under the current attitude. The aerodynamic parameters are obtained by looking up a table and transmitted to the aircraft for guidance via the overload command. Where the subscripts x, y, z represent the three-axis components in the ballistic coordinate system, C represents the aerodynamic coefficient, Q represents the dynamic pressure, S represents the characteristic area, m represents the mass of the aircraft, and g represents the gravitational acceleration.
4. The event-triggered adaptive guidance method for the down-phase of an aircraft as described in claim 1, characterized in that, The triggering condition determined by the event triggering mechanism in step S4 is:
5. The event-triggered adaptive guidance method for the down-phase of an aircraft as described in claim 1, characterized in that, The strategy evaluation method described in step S5 is as follows: input the line-of-sight angular velocity information and the flight state information of the previous fixed cycle into the onboard computer, use the adaptive dynamic programming method to train the guidance controller parameters, construct an evaluation function at the cost of the flight line-of-sight angular velocity information and state information during the process, construct an evaluation neural network, use the constructed evaluation neural network to fit the evaluation function, and use strategy evaluation to evaluate the value function of the current state. The strategy improvement method described in step S5 is as follows: Based on the strategy evaluation, the strategy is improved to generate a new strategy. The trained evaluation neural network is used as input to calculate the strategy in this state. An execution neural network is constructed, and the execution neural network is used to fit the aircraft model to obtain the execution neural network parameters of the fitted aircraft. The execution neural network parameters are then used to calculate the strategy result.
6. The event-triggered adaptive guidance method for the down-phase of an aircraft as described in claim 1, characterized in that, The evaluation function of the evaluation neural network is: Where V(x(·)) represents the evaluation function, x represents the state, Q and R are positive definite matrices, and u represents the control quantity.
7. The event-triggered adaptive guidance method for the down-phase of an aircraft as described in claim 1, characterized in that, The construction of the evaluation neural network includes: The evaluation neural network is set to four layers, trained using the BP algorithm, based on the gradient descent strategy, with a learning rate of 0.
01. The input to the training set is line-of-sight angular velocity information, and the output is the evaluation index. There are two hidden layers, each containing 6 hidden neurons. The constructed evaluation neural network fits the evaluation function as follows: In this neural network, the input is the basis function. Let φ1(x) represent the neural network fitting parameters, and let φ1(x) represent the basis functions. The strategy evaluation method is as follows: estimating the value function of the current state. V (i) (0)＝0 Where f represents the state transition matrix and g represents the control matrix.
8. The event-triggered adaptive guidance method for the down-phase of an aircraft as described in claim 1, characterized in that, The method for improving the strategy is as follows: Based on the strategy evaluation, the strategy is improved to generate a new strategy: The trained evaluation neural network is used as input to calculate the policy in this state. An execution neural network is constructed, consisting of three layers, and trained using the backpropagation (BP) algorithm based on a gradient descent strategy. The input to the training set is line-of-sight angular velocity information, and the output is the aircraft state information [P, V]. A [θ], use an execution neural network to fit the aircraft model, obtain the execution neural network parameters of the fitted aircraft, and use the execution neural network parameters to calculate the strategy results: in, This indicates that the neural network is being fitted with parameters.
9. The event-triggered adaptive guidance method for the down-phase of an aircraft as described in claim 1, characterized in that, The approximation error e is calculated as follows:
10. The event-triggered adaptive guidance method for the down-phase of an aircraft as described in claim 1, characterized in that, The evaluation neural network parameter update rule described in step S7 is as follows: Where α1 is the learning rate; Update the parameters of the control neural network and execute the parameter update rules of the control neural network: Where F1 and F2 are positive definite matrices, which will update the control neural network parameters. Replace the guidance controller parameter K.