A Constrained Pilot Model Based on Prediction and Modeling Method

Abstract

The present invention provides a prediction-based constrained pilot model and modeling method, comprising a ship-machine ring module, a fuzzy perception module, a pilot decision module and a pilot execution module connected in sequence, wherein the pilot execution module is connected to the ship-machine ring module; the fuzzy perception module comprises a pilot altitude deviation perception submodule, a pilot lateral deviation perception submodule and a pilot flight status perception submodule, and each module works in coordination with each other. The fuzzy perception module is responsible for receiving and processing optical guidance information, centering deviation information, distance to the ship and LSO voice command information from the ship-machine ring module, simulating the process of the pilot estimating the deviation and the expected control target based on the information; the pilot decision module outputs the expected operation instruction based on the information solved by the fuzzy perception module, comprehensively predicts the flight trend and considers the pilot's control constraints, and generates the final actual operation instruction after passing through the pilot execution module, thereby effectively improving the safety and stability of the flight.

Description

Technical Field

[0001] The present invention relates to the field of flight mechanics and flight control technology, and in particular to a prediction-based constrained pilot model and a modeling method. Background Art

[0002] The essence of the carrier-based aircraft pilot model is a practical model designed to meet the need for describing human control under environmental constraints in aircraft dynamics modeling. However, as flight missions become more complex, in order to meet research needs, the pilot model has undergone many variations from a modeling perspective and algorithm application. However, its core of matching mission requirements has never changed. An effective pilot model always needs to describe control characteristics while matching aircraft characteristics, environmental constraints, and mission requirements.

[0003] During the approach and landing process, carrier-based aircraft pilots must control three key variables. They must navigate the aircraft within a defined speed and attitude range within a constantly moving arresting area the size of a basketball court, subject to disturbances such as the ship's wake and atmospheric turbulence. This task is challenging and carries a significant workload. Rigorously trained pilots demonstrate regular behavior when completing specific tasks, making it possible to develop models of pilot control behavior. Pilot perception, decision-making, and execution have long been research focuses in flight dynamics, flight control, and flight safety. Developing models that can simulate and predict pilot behavior in complex environments is crucial for flight safety analysis and flight strategy development. Currently, a variety of pilot models are available for carrier-based aircraft. These models can be broadly categorized into three types: quasi-linear models based on classical control theory, optimal models based on modern control theory, and fuzzy control models based on fuzzy theory.

[0004] Quasi-linear models based on classical control theory include the McRuer model and the structural pilot model. The McRuer model, based on analysis of numerous pilot-in-the-loop simulation results, studies the pilot's control behavior and is a parameter adjustment model. Hess first proposed the structural pilot model for compensatory tracking tasks in 1979. Compared to the McRuer model, it embodies more realistic signal processing. This model consists of two components: the central nervous system and the neuromuscular system, highlighting the pilot's dual functions of signal processing and control decision-making. The core parameter of the structural pilot model is the internal feedback signal. By varying its form, the pilot model can be adapted to different controlled object characteristics. Hess used this model to analyze the human-in-the-loop pitch angle control characteristics of different aircraft types during carrier-based aircraft approach and landing. However, this model only models the inner-loop control loop and lacks an outer-loop representation, making it unsuitable for direct application to the complete carrier-based aircraft human-machine system. Furthermore, due to the inherent characteristics of the model, it cannot represent coupled operations. Existing researchers have addressed the application issues of structural pilot models in carrier-based aircraft approach and landing missions to some extent by describing the pilot's outer-loop control behavior as a relay-like bang-bang control. However, this approach still employs decoupling. Furthermore, because the outer loop is bang-bang-like, the model's convergence is poor in complex environments. Furthermore, in addition to the McRuer and structural pilot models, there are other variations of quasi-linear models for modeling pilot control behavior, but none are as representative as these two. Although quasi-linear models have been very successful in the field of pilot-in-place (PIO) control, they are difficult to apply to multi-loop control tasks. As the NASA report points out, for simple tasks, the linear portion of the model accounts for the majority of the controller's output power. However, as task complexity increases, quasi-linear models become insufficient. When multi-channel control is involved, even if the construction method of a single-channel model can be extended to multiple channels, it is difficult to resolve the model structure problem and requires decoupling of the channels. These factors limit the application of quasi-linear models in complex operating conditions, making them unsuitable for describing the control behavior of carrier-based aircraft pilots during approach and landing in complex environments.

[0005] Optimal pilot models based on modern control theory offer significant advantages in addressing multi-loop control problems. Because the model describes pilot control behavior from an overall performance perspective based on optimal assumptions, strict decoupling is not required. The key difference between optimal pilot models and structural models is that they are not based on frequency-domain identification criteria, but rather on an intuitive assumption: that human pilot control behavior is optimal to a certain extent. The rationale for this assumption has been previously studied. Pilots typically expect to maintain a phase margin of at least 60° between the human-machine system and the aircraft, which aligns with some theoretical principles in optimal control theory. Based on this assumption, it becomes feasible to discuss pilot control behavior from the perspective of overall performance optimization. By simply designing a suitable model structure, the optimal model can be gradually extended to a wider range of flight missions, such as the LQR pilot model and the MOCM-AE pilot model, demonstrating the rationale of extrapolating models based on optimal assumptions. Existing researchers have applied the MOCM-AE model to carrier-based aircraft landing missions, taking into account the effects of low-altitude turbulence and wake turbulence, demonstrating the applicability of pilot models based on modern control theory to carrier-based aircraft landing missions. However, this model still has certain shortcomings. First, it does not consider the pilot's control input constraints. During approach and landing, to prevent overshoot of the carrier-based aircraft, the LSO usually constrains the pilot's control rate. Due to the characteristics of optimal control, control constraints can significantly affect the model's performance. Second, the model only considers the pilot's adaptive estimation of noise and does not incorporate the pilot's trend prediction behavior into the model. Therefore, it cannot describe the carrier-based aircraft pilot's control behavior based on the overall trend.

[0006] Due to human uncertainty and ambiguity, some research has established pilot models based on fuzzy control theory. For example, a variable-strategy pilot model designed by existing researchers has an outer loop structure that approximates a fuzzy mapping consisting of two sets. Another existing researcher has designed a discrete fuzzy pilot model suitable for carrier-based aircraft landing control. This model decomposes the pilot's control behavior into a series of discrete square waves and designs the pilot's response pattern based on fuzzy rules. Due to the mathematical characteristics of fuzzy sets, this model has a clear rule mapping when responding to LSO commands, thus explaining the working principle of the pilot-LSO loop. However, in describing the pilot's control behavior of the carrier-based aircraft, since the model output is a pulsed square wave signal and the control performance relies on fuzzy rule design, it has poor convergence in complex environments. Therefore, it is not suitable for explaining pilot models in complex environments. However, it can be applied to describe LSO control behavior. Summary of the Invention

[0007] This invention addresses the problems of various current pilot models in the carrier-based aircraft field, such as their failure to consider pilot control input constraints, their inability to describe overall trend-based control behavior, and their poor convergence in complex environments. This invention provides a prediction-based constrained pilot model. Based on parameters such as optical guidance information, centering deviation information, LSO voice command information, and distance from the ship, it establishes a set of prediction equations for trend prediction using specific calculation methods and equations. Under these specific constraints, an objective function is constructed to calculate the pilot's actual operational commands. This model helps pilots better predict future trends, reduces the risk of excessive control, and improves flight safety and stability. The invention also relates to a method for modeling a prediction-based constrained pilot model.

[0008] The technical solutions of the present invention are as follows:

[0009] A prediction-based constrained pilot model, characterized by comprising a ship-machine ring module, a fuzzy perception module, a pilot decision module, and a pilot execution module connected in sequence, wherein the pilot execution module is connected to the ship-machine ring module; the fuzzy perception module comprises a pilot altitude deviation perception submodule, a pilot lateral deviation perception submodule, and a pilot flight status perception submodule,

[0010] The aircraft ring module is used to provide optical guidance information, centering deviation information between the carrier aircraft and the centerline of the aircraft carrier, LSO voice command information, the distance between the carrier aircraft and the ideal landing point on the aircraft carrier, and the navigation parameters of the carrier aircraft;

[0011] The pilot altitude deviation perception submodule calculates the glide path altitude deviation angle to be corrected based on the optical guidance information and the LSO voice command information using a fuzzy optical perception model, and calculates a corrected altitude deviation estimate using an altitude deviation estimator based on the glide path altitude deviation angle, the distance to the ship, the system noise in the optical guidance information, and the observation noise when the pilot observes the glide path altitude deviation angle.

[0012] The pilot lateral deviation perception submodule calculates the carrier centerline lateral deviation angle based on the centering deviation information and the LSO voice command information using a fuzzy lateral perception model, and calculates the lateral deviation observation value based on the carrier centerline lateral deviation angle and observation noise; and then calculates the track deviation angle based on the speed and heading in the navigation parameters;

[0013] The pilot flight state perception submodule calculates the actual observation value of the pilot based on the observation noise, and calculates the flight state estimation value based on the actual observation value using the flight state estimator;

[0014] The pilot decision module constructs a desired trajectory sequence based on the corrected altitude deviation estimate, lateral deviation observation, and track deviation angle; obtains a state space model of the aircraft based on the carrier aircraft's landing reference motion state and a coefficient freezing method, discretizes the state space model to obtain a discrete state space equation, and obtains a prediction equation group based on the discrete state space equation; then constructs a prediction observation sequence based on the flight state estimate at each moment in the prediction equation group, and constructs a control sequence based on the pilot's desired command at each moment in the prediction equation group; and, under constraints established based on the value range of the desired command at each moment, establishes an objective function based on the desired trajectory sequence, the predicted observation sequence, and the control sequence; then, processes the objective function using quadratic programming to obtain a new objective function, and calculates the pilot's optimal desired command based on the new objective function using an interior point method;

[0015] The pilot execution module calculates the pilot's actual operation instruction based on the optimal expected instruction, and sends the actual operation instruction to the ship-aircraft ring module so that the carrier-based aircraft executes the actual operation instruction.

[0016] Preferably, in the pilot altitude deviation perception submodule, the process of calculating the corrected altitude deviation estimate specifically includes:

[0017] The altitude deviation value is calculated according to the glide path altitude deviation angle and the distance to the ship, and a state vector is constructed based on the distance to the ship and the altitude deviation value. A state transfer equation is established according to the state vector at the current moment, the state vector at the next moment after the current moment, the system noise, and the observation noise. The expression of the altitude deviation estimate is obtained according to the state transfer equation, and the corrected altitude deviation estimate is then calculated.

[0018] Preferably, in the pilot altitude deviation perception submodule, the fuzzy optical perception model includes a fuzzifier, a membership function, fuzzy reasoning, a fuzzy rule base and a defuzzifier.

[0019] Preferably, in the pilot altitude deviation perception submodule, the process of calculating the glide slope altitude deviation angle to be corrected specifically includes:

[0020] The fuzzifier converts the optical guidance information and LSO voice command information into a fuzzy set defined by a membership function and inputs the fuzzy set into a fuzzy inference. The fuzzy inference matches the input fuzzy set with the rules in the fuzzy rule base to generate an output fuzzy set corresponding to the input fuzzy set. The defuzzifier defuzzifies the output fuzzy set obtained by the fuzzy inference to obtain the glide slope altitude deviation angle to be corrected.

[0021] Preferably, in the pilot decision module, the constraint conditions established based on the value range of the pilot's desired instruction at each moment specifically include: obtaining the physical structure constraints of the aircraft based on the aircraft structure, and obtaining the maximum and minimum values ​​of the desired instruction based on the physical structure constraints of the aircraft, and establishing the constraint conditions based on the desired instruction at each moment, the maximum value of the desired instruction, and the minimum value of the desired instruction.

[0022] A prediction-based constrained pilot modeling method, characterized in that the method is performed by the model according to any one of claims 1 to 5, and the method comprises:

[0023] The aircraft ring module provides optical guidance information, centering deviation information between the carrier aircraft and the carrier centerline, LSO voice command information, the distance between the carrier aircraft and the ideal landing point on the carrier, and the navigation parameters of the carrier aircraft. The pilot altitude deviation perception submodule calculates the glide path altitude deviation angle to be corrected based on the optical guidance information and LSO voice command information using a fuzzy optical perception model. The altitude deviation estimator calculates the corrected altitude deviation estimate based on the glide path altitude deviation angle, the distance to the carrier, the system noise in the optical guidance information, and the observation noise when the pilot observes the glide path altitude deviation angle. The pilot lateral deviation perception submodule calculates the carrier centerline lateral deviation angle based on the centering deviation information and LSO voice command information using a fuzzy lateral perception model. The lateral deviation observation value is calculated based on the carrier centerline lateral deviation angle and the observation noise. The track deviation angle is then calculated based on the speed and heading in the navigation parameters. The pilot flight state perception submodule calculates the pilot's actual observation value based on the observation noise, and calculates the flight state estimate value based on the actual observation value using a flight state estimator.

[0024] The pilot decision module constructs a desired trajectory sequence based on the corrected altitude deviation estimate, lateral deviation observation, and track deviation angle. Based on the carrier-based aircraft's landing reference motion and using the coefficient freezing method, the aircraft's state-space model is obtained. The state-space model is discretized to obtain discrete state-space equations, and a set of prediction equations is derived from the discrete state-space equations. A prediction observation sequence is then constructed based on the flight state estimate at each moment in the prediction equations. A control sequence is then constructed based on the pilot's desired command at each moment in the prediction equations. Under constraints established based on the range of the desired command at each moment, an objective function is established based on the desired trajectory sequence, the prediction observation sequence, and the control sequence. Quadratic programming is then used to process the objective function to obtain a new objective function, and the pilot's optimal desired command is calculated based on the new objective function using the interior point method.

[0025] The pilot execution module calculates the pilot's actual operation instructions based on the optimal expected instructions, and sends the actual operation instructions to the ship-aircraft ring module so that the carrier-based aircraft executes the actual operation instructions.

[0026] Preferably, calculating the corrected altitude deviation estimate using an altitude deviation estimator based on the glide path altitude deviation angle, the distance from the ship, the system noise in the optical guidance information, and the observation noise when the pilot observes the glide path altitude deviation angle specifically includes:

[0027] The altitude deviation value is calculated according to the glide path altitude deviation angle and the distance to the ship, and a state vector is constructed based on the distance to the ship and the altitude deviation value. A state transfer equation is established according to the state vector at the current moment, the state vector at the next moment after the current moment, the system noise, and the observation noise. The expression of the altitude deviation estimate is obtained according to the state transfer equation, and the corrected altitude deviation estimate is then calculated.

[0028] Preferably, the fuzzy optical perception model and the fuzzy lateral perception model both include a fuzzifier, a membership function, fuzzy reasoning, a fuzzy rule base and a defuzzifier.

[0029] Preferably, calculating the glide slope height deviation angle to be corrected based on the optical guidance information and the LSO voice command information and using a fuzzy optical perception model specifically includes:

[0030] The fuzzifier converts the optical guidance information and LSO voice command information into a fuzzy set defined by a membership function and inputs the fuzzy set into a fuzzy inference. The fuzzy inference matches the input fuzzy set with the rules in the fuzzy rule base to generate an output fuzzy set corresponding to the input fuzzy set. The defuzzifier defuzzifies the output fuzzy set obtained by the fuzzy inference to obtain the glide slope altitude deviation angle to be corrected.

[0031] Preferably, the constraint conditions established based on the value range of the pilot's expected instruction at each moment specifically include: obtaining the physical structure constraints of the aircraft based on the aircraft structure, and obtaining the maximum and minimum values ​​of the expected instruction based on the physical structure constraints of the aircraft, and establishing the constraint conditions based on the expected instruction at each moment, the maximum value of the expected instruction, and the minimum value of the expected instruction.

[0032] The beneficial effects of the present invention are:

[0033] The present invention provides a prediction-based constrained pilot model, which is used to solve the problems of various pilot models in the current carrier-based aircraft field, such as poor convergence in complex environments, failure to consider the pilot's control input constraints, and inability to describe the control behavior of carrier-based aircraft pilots based on overall trends. The model includes a ship-aircraft ring module, a fuzzy perception module, a pilot decision module, and a pilot execution module connected in sequence. The pilot execution module is connected to the ship-aircraft ring module; the fuzzy perception module includes a pilot altitude deviation perception submodule, a pilot lateral deviation perception submodule, and a pilot flight status perception submodule. Each module works in coordination with each other. The ship-aircraft ring module provides optical guidance information, centering deviation information between the carrier-based aircraft and the aircraft carrier centerline, LSO voice command information, the distance between the carrier-based aircraft and the ideal landing point on the aircraft carrier, and The carrier-based aircraft's navigation parameters and the pilot altitude deviation perception submodule calculate the corrected glide path altitude deviation angle based on optical guidance information and LSO voice command information using a fuzzy optical perception model. This converts unstructured voice (LSO voice command information) and sensory data (optical guidance information) into structured deviation data, enabling the pilot to better understand the current state and deviation, improving their perception of flight status and enhancing the efficiency and accuracy of information processing, thereby improving the safety and reliability of the entire landing process. A specially designed altitude deviation estimator is then used to calculate the corrected altitude deviation estimate, effectively resolving the problem of observation deviation caused by observation noise when the pilot observes the glide path altitude deviation angle, and improving the accuracy of the pilot's information acquisition. The pilot's lateral deviation perception submodule uses a fuzzy lateral perception model to calculate the lateral deviation angle of the aircraft carrier's centerline, and uses a specific calculation method to calculate the lateral deviation observation value and track deviation angle, effectively reducing the pilot's operating burden, alleviating work pressure, and improving overall efficiency; the pilot's flight status perception submodule calculates the pilot's actual observation value based on the observation noise, and calculates the flight status estimation value based on the actual observation value and using a specifically designed flight status estimator. By converting complex state variables into intuitive estimation behaviors, pilots do not need to pay attention to changes in all variables, but only need to pay attention to key indicators, effectively reducing workload and improving work efficiency; at the same time, by estimating future trend changes, pilots can predict possible situations in advance and avoid potential risks.The pilot decision module constructs a desired trajectory sequence and discretizes the aircraft's state-space model to obtain discrete state-space equations. These equations generate a set of prediction equations, enabling pilots to better predict future trends and make better decisions, effectively improving flight safety and efficiency. An objective function is then established within constraints based on the range of desired commands at each moment. Quadratic programming is then used to process this objective function to obtain a new objective function. Based on this new objective function and the interior point method, the pilot's optimal desired command is calculated. By considering the pilot's control input constraints, the risk of excessive control can be reduced, thereby improving flight safety and stability. The execution module calculates the pilot's actual control command based on the optimal desired command and sends it to the aircraft-carrier loop module for execution. By accounting for the delay and noise associated with the pilot's physiological mechanisms, the module can more accurately simulate pilot behavior and estimate the pilot's control signals, thereby improving decision accuracy.

[0034] The present invention also relates to a prediction-based constrained pilot model modeling method. This method corresponds to the above-mentioned prediction-based constrained pilot model and can be understood as an implementation method of the above-mentioned prediction-based constrained pilot model. Based on parameters such as optical guidance information, centering deviation information, LSO voice command information, and distance to the ship, a prediction equation group for trend prediction is established by adopting specific calculation methods and equations, and an objective function is constructed under the established specific constraints to calculate the pilot's actual operation instructions. This can help pilots better predict future trends, reduce the risks caused by excessive operation of pilots, and improve flight safety and stability. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] Figure 1 It is a schematic diagram of the structure of the prediction-based constrained pilot model of the present invention.

[0036] Figure 2 It is a schematic diagram of the preferred structure of the prediction-based constrained pilot model of the present invention.

[0037] Figure 3 Schematic diagram of the working principle of the FLOLS system of the present invention.

[0038] Figure 4 It is a structural schematic diagram of the pilot altitude deviation perception submodule of the present invention.

[0039] Figure 5 It is a structural schematic diagram of the pilot lateral deviation perception submodule of the present invention.

[0040] Figure 6 It is a structural diagram of the flight status perception submodule of the present invention. DETAILED DESCRIPTION

[0041] The present invention will be described below with reference to the accompanying drawings.

[0042] The present invention relates to a prediction-based constrained pilot model, the structural diagram of which is shown in FIG. Figure 1 As shown, the system includes a carrier-based aircraft ring module, a fuzzy perception module, a pilot decision module, and a pilot execution module, which are connected in sequence. The fuzzy perception module is responsible for receiving and processing perception information such as optical guidance information, centering deviation information of the aircraft carrier centerline, the distance between the aircraft tail hook and the ideal landing point on the aircraft carrier, and LSO voice command information from the ship's surface. It simulates the process of the pilot estimating the deviation and desired control target based on the optical guidance information. The pilot decision module, based on the information solved by the fuzzy perception module, comprehensively predicts the flight trend and considers the pilot's control constraints to output the desired operation instruction. After the pilot execution module, it generates the final actual operation instruction. This solves the problems of various pilot models in the current carrier-based aircraft field, such as poor convergence in complex environments, failure to consider the pilot's control input constraints, and inability to describe the control behavior of carrier-based aircraft pilots based on overall trends. Among them, in order to make the model as close to the actual landing process as possible, the following assumptions are designed to make the pilot model as consistent as possible with the limits of human ability. The specific assumptions are as follows: 1) Throughout the process, the noise generated by the pilot's own physiological characteristics is the most common zero-mean white noise in natural organisms. The noise in each time interval is independent, and the noise intensity is linearly related to the task load. 2) The pilot will try to make the best possible maneuver, but only within a limited timeframe will their decisions conform to the Bellman equation. 3) The pilot has sufficient prior knowledge of voice, body sensation, and the dynamics of the aircraft to interpret. They will not make preemptive maneuvers for unknown disturbances. This model is suitable for pilots who are proficient in landing missions.

[0043] Specifically, if Figure 2 As shown in the schematic diagram of the preferred structure, the model includes a ship-aircraft ring module, a fuzzy perception module, a pilot decision module and a pilot execution module which are sequentially connected into a closed loop, and the pilot execution module is connected to the ship-aircraft ring module; the fuzzy perception module includes a pilot altitude deviation perception submodule, a pilot lateral deviation perception submodule and a pilot flight status perception submodule.

[0044] Among them, the ship-aircraft ring module is used to provide optical guidance information in FOLS (shipborne optical landing system), centering deviation information between the carrier-based aircraft and the centerline of the aircraft carrier, LSO voice command information, the distance between the tail hook of the carrier-based aircraft and the ideal landing point on the aircraft carrier, and the navigation parameters of the carrier-based aircraft, including speed and heading. In the landing mission, optical guidance information is the main way to obtain altitude deviation. The optical guidance information seen by the driver is a fixed reference light and a movable "ball", which is formed by the light of the reference light group and the aiming light group. The FLOL optical scene (that is, the optical guidance information in FOLLS) indicates to the driver the altitude deviation information between the carrier-based aircraft and the ideal glide path, such as Figure 3 When the FOLLS optical image seen by the pilot shows the ball aligned with the reference line (i.e., the reference light group), the carrier aircraft is on the ideal glide path; when the ball is above or below the reference line, the carrier aircraft is above or below the ideal glide path; when the distance between the ball and the reference line increases or decreases, the carrier aircraft is moving away from or closer to the ideal glide path. Pilots usually use the word "ball" to describe the current position of the carrier aircraft.

[0045] The pilot altitude deviation perception submodule calculates the glide path altitude deviation angle to be corrected based on the optical guidance information and LSO voice command information using a fuzzy optical perception model. It then uses the altitude deviation estimator to calculate the corrected altitude deviation estimate based on the glide path altitude deviation angle, distance to the ship, system noise in the optical guidance information, and observation noise when the pilot observes the glide path altitude deviation angle.

[0046] Specifically, if Figure 4 As shown, the fuzzifier in the fuzzy optical perception model first converts the optical guidance information and LSO voice command information into fuzzy sets defined by membership functions and then inputs them into fuzzy reasoning. Fuzzy sets are defined by membership functions, each with a clear boundary and shape that describes the degree of fuzziness of the input. Fuzzy reasoning matches the input fuzzy set with rules in a fuzzy rule base to generate an output fuzzy set corresponding to the input fuzzy set. The defuzzifier defuzzifies the output fuzzy set obtained by fuzzy reasoning to obtain the corrected glide slope altitude deviation angle Δe. This transforms the pilot's observation of the corrected glide slope altitude deviation angle Δe into a fuzzy reasoning process. The defuzzification process is designed according to the maximum membership principle, which selects the fuzzy set with the highest membership (defined by the membership function) as the final output.

[0047] At different landing distances, the pilot's gain in correcting the glide path height deviation angle Δe is also different. To express this difference, it is necessary to convert the glide path height deviation angle Δe into a height deviation value ΔH. Due to human physiological limitations, there is inevitably observation noise v when the pilot observes the glide path height deviation angle Δe, which leads to observation deviation. It is necessary to perform state estimation to obtain the corrected height deviation estimate ΔH. K This process is achieved through the altitude deviation estimator. The relationship between the altitude deviation angle Δe, the altitude deviation value ΔH and the distance from the ship Rx is:

[0048]

[0049] Then, the state vector X is constructed based on the distance from the ship and the height deviation value, and the According to the state vector X(k) at the current moment, the state vector X(k+1) at the next moment after the current moment, and the introduction of system noise w1 and observation noise v, the state transition equation of the FLOLS system can be established:

[0050] X(k+1)=A FLOLS X(k)+w1

[0051] Z=C FLOLS X(k)+v (2)

[0052] Where X is the state vector of the FLOLS system state transfer equation, A FLOLS is the state transfer matrix, C FLOLS is the observation matrix, k represents the current moment, and T is the transpose.

[0053] The expression of the altitude deviation estimate is obtained according to the state transition equation. The pilot's altitude deviation estimate (i.e., the corrected altitude deviation estimate) can be expressed as follows:

[0054]

[0055] In the above formula, F is the Kalman gain coefficient. According to the above formula (3), the corrected height deviation estimate ΔH can be calculated K .

[0056] The pilot lateral deviation perception submodule calculates the lateral deviation angle of the carrier centerline based on the centering deviation information and LSO voice command information using a fuzzy lateral perception model, and calculates the lateral deviation observation value based on the lateral deviation angle of the carrier centerline and observation noise; and then calculates the track deviation angle based on the speed and heading.

[0057] Specifically, in the landing mission, the lateral deviation is mainly measured by the deviation angle between the carrier-based aircraft and the aircraft carrier deck. The pilot can refer to the deck centerline and the landing officer (LSO) instructions to obtain the lateral centering deviation information. During the day, some pilots will also align the centerline by the aircraft carrier's wake. The centering deviation margin is higher than the height deviation. This is because the lateral safety envelope is wider than the longitudinal one. Existing technicians require the LSO to perform centering commands only when the glide path is qualified or slightly higher, and to be vigilant against sinking caused by instructions. This shows that in the landing mission, the priority of centerline alignment is second only to altitude. As Figure 5 As shown, the fuzzifier in the fuzzy lateral perception model first converts the lateral centering deviation information and LSO voice command information into fuzzy sets defined by membership functions and then inputs them into fuzzy inference. Fuzzy sets are defined by membership functions, and each fuzzy set has a clear boundary and shape, which is used to describe the degree of fuzziness of the input. Fuzzy inference matches the input fuzzy set with the rules in the fuzzy rule base to generate an output fuzzy set corresponding to the input fuzzy set. The defuzzifier defuzzifies the output fuzzy set obtained by fuzzy inference to obtain the lateral deviation angle of the aircraft carrier's centerline, thereby transforming the pilot's observation of the lateral deviation angle of the aircraft carrier's centerline into a fuzzy inference process. The defuzzification process is designed according to the maximum membership principle, that is, the fuzzy set with the highest membership (defined by the membership function) is selected as the final output.

[0058] In addition, since the difficulty of estimating errors by vision is particularly severe in the centering task, pilots tend to reduce the burden of operation through some pre-calculation, such as correcting the track angle of the carrier-based aircraft in advance by the speed and heading of the ship, and the speed and heading of the carrier-based aircraft. This makes the pilot's lateral deviation correction more dependent on the LSO voice command and the track angle based on prior knowledge. The corrected track angle The calculation formula is as follows:

[0059]

[0060] In the above formula, V ship Indicates the ship's speed, V K Indicates the track speed of the carrier-based aircraft, Install corners for the deck.

[0061] The pilot flight state perception submodule calculates the pilot's actual observation quantity based on the observation noise, and obtains the pilot's estimation of the flight state based on the actual observation quantity and the flight state estimator.

[0062] Specifically, if Figure 6As shown in Figure 1, trend assessment is a key aspect of pilot control technology, typically predicting and controlling trends within the next 2-3 seconds. This requires the pilot to maintain a comprehensive understanding of the aircraft's flight state. However, not all state variables are observable or readily noticeable. Generally speaking, apart from the mission-required position and angle of attack (or velocity), only changes in pitch and roll angles can be readily observed from the horizon. Other state variables do not occupy the pilot's attention and are therefore generally described as "stable" or "changing." Therefore, these variables are described in the pilot perception model as estimates based on observable quantities combined with prior knowledge of flight dynamics. Furthermore, the landing officer (LSO) also issues speed commands. While some commands are intuitive to adjust, such as "power," which directly corresponds to throttle control, there are also commands, such as adjusting the descent rate and drift speed, that require coordination between the rudder and throttle. To facilitate modeling, the necessary state observables are also converted into the fuzzy perception module.

[0063] The pilot's overall control of the flight trend can be regarded as an estimation behavior. Assume that the state space model of the aircraft is:

[0064]

[0065] y=Cx+Du p (5)

[0066] Where x is the state vector of the aircraft. The initial state vector x0 of the aircraft is obtained from the carrier-based aircraft landing reference motion state (the reference motion state includes the attitude angle, speed and glide path of the carrier-based aircraft); u p is the actual operation instruction of the pilot, the initial actual operation instruction u p0 It is obtained from the carrier-based aircraft landing reference motion; A is the state matrix, B is the input matrix, w2 is the noise excitation function (including all noises including system noise, disturbance noise, etc.), E is the disturbance matrix, C is the observation matrix, and D is the direct transfer matrix.

[0067] The actual observation quantity of the pilot is calculated based on the observation noise, and the actual observation quantity of the pilot is:

[0068] y p =C Pilot x+v y (6)

[0069] Among them, C Pilot is the pilot observation matrix, v y is the observation noise, and x is the state vector of the aircraft.

[0070] y pIt is the state quantity that the pilot can actually observe (the actual observation quantity of the pilot), which changes with the working conditions. The pilot's perception of the flight state can be regarded as an estimation behavior. This process is realized by the flight state estimator. Therefore, the pilot's estimation of the flight state can be calculated based on the actual observation quantity and the flight state estimator. The input required for estimating the flight state includes the observable quantity y p (the actual observation of the pilot) and the pilot's own expected instruction u c , assuming that the noise in the process is zero-mean white noise, we have:

[0071]

[0072] in, represents the state observation value,

[0073] Σ is the solution of the following Ricatti equation:

[0074]

[0075] Where W is the noise matrix. Based on formula (6), the pilot’s estimate of the flight state can be obtained: (i.e., flight status estimate).

[0076] The pilot decision module constructs a desired trajectory sequence based on the corrected altitude deviation estimate, lateral deviation observation, and track deviation angle. Based on the carrier-based aircraft's landing reference motion and using the coefficient freezing method, the aircraft's state-space model is obtained. The state-space model is discretized to obtain discrete state-space equations, and a set of prediction equations is obtained based on the discrete state-space equations. A prediction observation sequence is then constructed based on the flight state estimate at each moment in the prediction equation set, and a control sequence is constructed based on the pilot's desired command at each moment in the prediction equation set. Under constraints established based on the value range of the desired command at each moment, an objective function is established based on the desired trajectory sequence, the prediction observation sequence, and the control sequence. The objective function is then processed using the quadratic programming method to obtain a new objective function, and the pilot's optimal desired command is calculated based on the new objective function using the interior point method.

[0077] The pilot decision module describes how the pilot makes decisions based on the information processed by the fuzzy perception module and outputs the expected operation instructions u cprocess. In addition, in order to establish a more realistic pilot decision model, this application refers to the opinions of real pilots. According to the pilots' description, the judgment of trends is the focus of pilot control technology. The pilots emphasized the importance of anticipation (i.e., trend prediction). The flight captain described the anticipation capability as: the pilot's ability to understand the aerodynamics currently occurring and to make accurate predictions about future trends. Usually, pilots predict trend changes in the next 2-3s and complete operations. Therefore, this application establishes a pilot decision module based on predictive features. Specifically, first, the expected trajectory sequence Y is constructed based on the corrected altitude deviation estimate, lateral deviation observation, and track deviation angle. tracjtory Then, based on the carrier-based aircraft's landing reference motion and using the coefficient freezing method, the aircraft's state space model is obtained and expressed as follows:

[0078]

[0079] y=Cx+Du p (9)

[0080] Where x is the state vector of the aircraft. The initial state vector x0 of the aircraft is obtained from the carrier-based aircraft landing reference motion state (the reference motion state includes the attitude angle, speed and glide path of the carrier-based aircraft); u p is the actual operation instruction of the pilot, the initial actual operation instruction u p0 It is obtained from the carrier-based aircraft landing reference motion; w is the noise excitation function (including all noises including system noise, disturbance noise, etc.), A is the state matrix, B is the input matrix, E is the disturbance matrix, C is the observation matrix, and D is the direct transfer matrix.

[0081] The state space model is discretized to obtain the discrete state space equation, which is expressed as follows:

[0082] x(k+1)=Ax(k)+Bu c (k)+Ew(k)

[0083] y=Cx(k)+Du c (k) (10)

[0084] The prediction equations can be obtained by recursively deducing the discrete state space equations:

[0085] x(k+2)=A 2 x(k)+ABu c (k)+Bu c (k+1)+AEw

[0086]

[0087] …

[0088]

[0089] In the above formula, the prediction equations derived recursively based on the discrete state space equations represent the pilot's prediction behavior of the flight state trend, P is the pilot's prediction time domain, and k represents the current moment.

[0090] After obtaining the prediction equations, the flight state estimation value at each moment in the prediction equations is That is, x(k) in the prediction equation group (10) constructs the prediction observation sequence Y p , and according to the expected instruction u of the pilot at each moment in the prediction equation group c Construct the control sequence ΔU(k).

[0091] Furthermore, constraints are common in pilots' work environments. Some studies have characterized pilot control behavior as a highly constrained optimal linear controller. During carrier-based aircraft landings, military thrust is typically maintained at around 85%. To maintain flight stability, changes in control variables must be kept as small as possible. Based on pilot experience, a command input of 1% to 2% is generally considered appropriate, while adjustments exceeding 5% are considered inappropriate and coarse, limiting the pilot's control rate. Furthermore, the control resources available to the pilot vary at each sampling time, significantly impacting their decision-making. This not only restricts their control performance but also introduces the intelligent nature of control margin-based allocation. The pilot model developed in this paper has the advantage of explicitly addressing constraints. By incorporating control variable constraints as conditions into performance indicators, the pilot's control behavior is described as a planning problem. The small-disturbance model established under conventional operating points typically retains only 10% to 15% of the throttle control resource margin. Using hard constraints in the modeling approach could result in the control variable being permanently at its maximum value. Therefore, it is essential to consider control input constraints. First, the physical constraints of the aircraft are obtained based on the aircraft structure. The maximum and minimum values ​​of the expected instructions are obtained based on the physical constraints of the aircraft. Constraints are established based on the expected instructions, the maximum value of the expected instructions, and the minimum value of the expected instructions at each moment. The constraints of the control input are expressed as follows:

[0092] u min (k+i)≤u(k+i)≤u max (k+i) (12)

[0093] Under the constraints established above, the objective function is established according to the expected trajectory sequence, the predicted observation sequence and the control sequence. The objective function J is shown as follows:

[0094] J=||Q*(Y p (k+1|k)-Ytracjtory (k+1|k))|| 2 +||R*ΔU(k)||2 (13)

[0095] In the above formula, Y p is the predicted observation sequence, Y tracjtory is the desired trajectory sequence, ΔU(k) ​​is the control sequence, Q is the state weight matrix, and R is the control weight matrix.

[0096] The objective function includes the degree to which the flight trend deviates from the mission and the degree of consumption of control resources. Based on the above function, the closed-loop pilot optimal control can be derived as:

[0097]

[0098] in:

[0099] E(k+1)=R(k+1)-S x Δx(k)-Iy c (k)-S d Δd(k) (15)

[0100]

[0101] Where I represents the identity matrix and m is the control time domain.

[0102] It should be noted that after adding constraints, the objective function J contains inequalities, and the optimal pilot gain can no longer be obtained by solving the Riccati equation. This is a typical QP problem, so it needs to be considered from an optimization perspective. Therefore, based on the above assumption 2), it is assumed that the pilot will optimize the performance function at each sampling moment. In a physical sense, the pilot gain obtained by the optimization method is essentially a control strategy that minimizes the overall deviation of the flight state by combining the pilot's experience with his own estimation of the flight trend. Selecting appropriate Q and R will prevent inhuman control behavior. The quadratic programming (QP) problem is a classic type of optimization problem. The relevant numerical optimization methods are relatively mature and have been solved. Therefore, the pilot's decision-making problem can be solved by converting it into the standard form of the QP problem. The standard form of the QP problem is:

[0103]

[0104] z represents the solution to be found, which is the pilot's decision problem ΔU(k). The prediction equation is transformed into:

[0105]

[0106] Then the objective function J is transformed into:

[0107]

[0108] At this point, the pilot's decision problem is transformed into a standard form of a QP problem, that is, the objective function J is processed using the constrained quadratic programming standard form to obtain a new objective function The optimization numerical calculation of the QP problem is solved by the interior point method to obtain the pilot's optimal expected instruction u c (Also called the optimal expected operation instruction u c ).

[0109] The pilot execution module calculates the pilot's actual operation instructions based on the optimal expected instructions, and sends the actual operation instructions to the ship-aircraft ring module so that the carrier-based aircraft executes the actual operation instructions.

[0110] Specifically, if Figure 6 As shown, the expected instruction u in the pilot's mind c The actual operation instruction u with the actual output p There are always errors, which are caused by ambiguity and non-repeatability due to human physiological limitations. The role of the pilot execution module is to establish an error model caused by physiological noise and describe the expected command u c To actual operation instructions u p In order to describe the pilot's final action output, the method of superimposing the action error on the basis of the pilot's expected action is adopted. This method not only conforms to the output principle of the actual pilot's final action, but also is easy to implement from a mathematical point of view. Therefore, the pilot's actual operation instruction u p It can be expressed as follows:

[0111] u p =u c +B d *w (20)

[0112] Where B d is the covariance matrix of the zero-mean white noise perturbation, and w is the zero-mean white noise excitation function. Scientific research shows that the normal distribution is the most common form of biological and physiological noise in nature. Therefore, according to Assumption 1), the pilot's physiological noise also satisfies the normal distribution relationship, so w is the zero-mean white noise excitation function.

[0113] The present invention further relates to a method for modeling a prediction-based constrained pilot model. This method corresponds to the above-mentioned prediction-based constrained pilot model and can be understood as a method for implementing the prediction-based constrained pilot model. The method includes:

[0114] The aircraft ring module provides optical guidance information, centering deviation information between the carrier aircraft and the carrier centerline, LSO voice command information, the distance between the carrier aircraft and the ideal landing point on the carrier, and the navigation parameters of the carrier aircraft. The pilot altitude deviation perception submodule calculates the glide path altitude deviation angle to be corrected based on the optical guidance information and LSO voice command information using a fuzzy optical perception model. The altitude deviation estimator calculates the corrected altitude deviation estimate based on the glide path altitude deviation angle, the distance to the carrier, the system noise in the optical guidance information, and the observation noise when the pilot observes the glide path altitude deviation angle. The pilot lateral deviation perception submodule calculates the carrier centerline lateral deviation angle based on the centering deviation information and LSO voice command information using a fuzzy lateral perception model. The lateral deviation observation value is calculated based on the carrier centerline lateral deviation angle and the observation noise. The track deviation angle is then calculated based on the speed and heading in the navigation parameters. The pilot flight state perception submodule calculates the pilot's actual observation value based on the observation noise, and calculates the flight state estimate value based on the actual observation value using a flight state estimator.

[0115] The pilot decision module constructs a desired trajectory sequence based on the corrected altitude deviation estimate, lateral deviation observation, and track deviation angle. Based on the carrier-based aircraft's landing reference motion and using the coefficient freezing method, the aircraft's state-space model is obtained. The state-space model is discretized to obtain discrete state-space equations, and a set of prediction equations is derived from the discrete state-space equations. A prediction observation sequence is then constructed based on the flight state estimate at each moment in the prediction equations. A control sequence is then constructed based on the pilot's desired command at each moment in the prediction equations. Under constraints established based on the range of the desired command at each moment, an objective function is established based on the desired trajectory sequence, the prediction observation sequence, and the control sequence. Quadratic programming is then used to process the objective function to obtain a new objective function, and the pilot's optimal desired command is calculated based on the new objective function using the interior point method.

[0116] The pilot execution module calculates the pilot's actual operation instructions based on the optimal expected instructions, and sends the actual operation instructions to the ship-aircraft ring module so that the carrier-based aircraft executes the actual operation instructions.

[0117] Preferably, calculating the corrected altitude deviation estimate using an altitude deviation estimator based on the glide path altitude deviation angle, the distance from the ship, the system noise in the optical guidance information, and the observation noise when the pilot observes the glide path altitude deviation angle specifically includes:

[0118] The altitude deviation value is calculated according to the glide path altitude deviation angle and the distance to the ship, and a state vector is constructed based on the distance to the ship and the altitude deviation value. A state transfer equation is established according to the state vector at the current moment, the state vector at the next moment after the current moment, the system noise, and the observation noise. The expression of the altitude deviation estimate is obtained according to the state transfer equation, and the corrected altitude deviation estimate is then calculated.

[0119] Preferably, the fuzzy optical perception model and the fuzzy lateral perception model both include a fuzzifier, a membership function, fuzzy reasoning, a fuzzy rule base and a defuzzifier.

[0120] Preferably, calculating the glide slope height deviation angle to be corrected based on the optical guidance information and the LSO voice command information and using a fuzzy optical perception model specifically includes:

[0121] The fuzzifier converts the optical guidance information and LSO voice command information into a fuzzy set defined by a membership function and inputs the fuzzy set into a fuzzy inference. The fuzzy inference matches the input fuzzy set with the rules in the fuzzy rule base to generate an output fuzzy set corresponding to the input fuzzy set. The defuzzifier defuzzifies the output fuzzy set obtained by the fuzzy inference to obtain the glide slope altitude deviation angle to be corrected.

[0122] Preferably, the constraint conditions established based on the value range of the pilot's expected instruction at each moment specifically include: obtaining the physical structure constraints of the aircraft based on the aircraft structure, and obtaining the maximum and minimum values ​​of the expected instruction based on the physical structure constraints of the aircraft, and establishing the constraint conditions based on the expected instruction at each moment, the maximum value of the expected instruction, and the minimum value of the expected instruction.

[0123] The present invention provides an objective and scientific prediction-based constrained pilot model and modeling method. Based on parameters such as optical guidance information, centering deviation information, LSO voice command information, and distance to the ship, a prediction equation group for trend prediction is established by adopting specific calculation methods and equations. An objective function is constructed under the established specific constraints to calculate the pilot's actual operation instructions. This can help pilots better predict future trends, reduce the risks caused by excessive operation of pilots, and improve flight safety and stability.

[0124] It should be noted that the specific embodiments described above can enable those skilled in the art to more fully understand the present invention, but do not limit the present invention in any way. Therefore, although this specification has described the present invention in detail with reference to the drawings and embodiments, those skilled in the art should understand that the present invention can still be modified or replaced with equivalents. In short, all technical solutions and improvements that do not depart from the spirit and scope of the present invention should be included in the scope of protection of the patent for the present invention.

Claims

1. A prediction-based constrained pilot model, characterized in that It includes a ship-machine ring module, a fuzzy perception module, a pilot decision module and a pilot execution module connected in sequence. The pilot execution module is connected to the ship-machine ring module; the fuzzy perception module includes a pilot altitude deviation perception submodule, a pilot lateral deviation perception submodule and a pilot flight status perception submodule. The aircraft ring module is used to provide optical guidance information, centering deviation information between the carrier aircraft and the centerline of the aircraft carrier, LSO voice command information, the distance between the carrier aircraft and the ideal landing point on the aircraft carrier, and the navigation parameters of the carrier aircraft; The pilot altitude deviation perception submodule calculates the glide path altitude deviation angle to be corrected based on the optical guidance information and the LSO voice command information using a fuzzy optical perception model, and calculates a corrected altitude deviation estimate using an altitude deviation estimator based on the glide path altitude deviation angle, the distance to the ship, the system noise in the optical guidance information, and the observation noise when the pilot observes the glide path altitude deviation angle. The pilot lateral deviation perception submodule calculates the carrier centerline lateral deviation angle based on the centering deviation information and the LSO voice command information using a fuzzy lateral perception model, and calculates the lateral deviation observation value based on the carrier centerline lateral deviation angle and observation noise; and then calculates the track deviation angle based on the speed and heading in the navigation parameters; The pilot flight state perception submodule calculates the actual observation value of the pilot based on the observation noise, and calculates the flight state estimation value based on the actual observation value using the flight state estimator; The pilot decision module constructs a desired trajectory sequence based on the corrected altitude deviation estimate, lateral deviation observation, and track deviation angle; obtains a state space model of the aircraft based on the carrier aircraft's landing reference motion state and a coefficient freezing method, discretizes the state space model to obtain a discrete state space equation, and obtains a prediction equation group based on the discrete state space equation; then constructs a prediction observation sequence based on the flight state estimate at each moment in the prediction equation group, and constructs a control sequence based on the pilot's desired command at each moment in the prediction equation group; and, under constraints established based on the value range of the desired command at each moment, establishes an objective function based on the desired trajectory sequence, the predicted observation sequence, and the control sequence; then, processes the objective function using quadratic programming to obtain a new objective function, and calculates the pilot's optimal desired command based on the new objective function using an interior point method; The pilot execution module calculates the pilot's actual operation instruction based on the optimal expected instruction, and sends the actual operation instruction to the ship-aircraft ring module so that the carrier-based aircraft executes the actual operation instruction.

2. The prediction-based constrained pilot model according to claim 1, wherein: In the pilot altitude deviation perception submodule, the process of calculating the corrected altitude deviation estimate specifically includes: The altitude deviation value is calculated according to the glide path altitude deviation angle and the distance to the ship, and a state vector is constructed based on the distance to the ship and the altitude deviation value. A state transfer equation is established according to the state vector at the current moment, the state vector at the next moment after the current moment, the system noise, and the observation noise. The expression of the altitude deviation estimate is obtained according to the state transfer equation, and the corrected altitude deviation estimate is then calculated.

3. The prediction-based constrained pilot model of claim 1, wherein: In the pilot altitude deviation perception submodule, the fuzzy optical perception model includes a fuzzifier, a membership function, fuzzy reasoning, a fuzzy rule base, and a defuzzifier.

4. The prediction-based constrained pilot model of claim 3, wherein: In the pilot altitude deviation perception submodule, the glide slope altitude deviation angle calculation process to be corrected specifically includes: The fuzzifier converts the optical guidance information and LSO voice command information into a fuzzy set defined by a membership function and inputs the fuzzy set into a fuzzy inference. The fuzzy inference matches the input fuzzy set with the rules in the fuzzy rule base to generate an output fuzzy set corresponding to the input fuzzy set. The defuzzifier defuzzifies the output fuzzy set obtained by the fuzzy inference to obtain the glide slope altitude deviation angle to be corrected.

5. The prediction-based constrained pilot model of claim 1, wherein: In the pilot decision module, the constraints established based on the range of values ​​of the pilot's desired instruction at each moment specifically include: obtaining the aircraft's physical structural constraints based on the aircraft structure, and obtaining the maximum and minimum values ​​of the desired instruction based on the aircraft's physical structural constraints, and establishing the constraints based on the desired instruction at each moment, the maximum value of the desired instruction, and the minimum value of the desired instruction.

6. A method for modeling a constrained pilot model based on prediction, characterized in that: The method is performed by the model according to any one of claims 1 to 5, and the method includes: The aircraft ring module provides optical guidance information, centering deviation information between the carrier aircraft and the carrier centerline, LSO voice command information, the distance between the carrier aircraft and the ideal landing point on the carrier, and the navigation parameters of the carrier aircraft. The pilot altitude deviation perception submodule calculates the glide path altitude deviation angle to be corrected based on the optical guidance information and LSO voice command information using a fuzzy optical perception model. The altitude deviation estimator calculates the corrected altitude deviation estimate based on the glide path altitude deviation angle, the distance to the carrier, the system noise in the optical guidance information, and the observation noise when the pilot observes the glide path altitude deviation angle. The pilot lateral deviation perception submodule calculates the carrier centerline lateral deviation angle based on the centering deviation information and LSO voice command information using a fuzzy lateral perception model. The lateral deviation observation value is calculated based on the carrier centerline lateral deviation angle and the observation noise. The track deviation angle is then calculated based on the speed and heading in the navigation parameters. The pilot flight state perception submodule calculates the pilot's actual observation value based on the observation noise, and calculates the flight state estimate value based on the actual observation value using a flight state estimator. The pilot decision module constructs a desired trajectory sequence based on the corrected altitude deviation estimate, lateral deviation observation, and track deviation angle. Based on the carrier-based aircraft's landing reference motion and using the coefficient freezing method, the aircraft's state-space model is obtained. The state-space model is discretized to obtain discrete state-space equations, and a set of prediction equations is derived from the discrete state-space equations. A prediction observation sequence is then constructed based on the flight state estimate at each moment in the prediction equations. A control sequence is then constructed based on the pilot's desired command at each moment in the prediction equations. Under constraints established based on the range of the desired command at each moment, an objective function is established based on the desired trajectory sequence, the prediction observation sequence, and the control sequence. Quadratic programming is then used to process the objective function to obtain a new objective function, and the pilot's optimal desired command is calculated based on the new objective function using the interior point method. The pilot execution module calculates the pilot's actual operation instructions based on the optimal expected instructions, and sends the actual operation instructions to the ship-aircraft ring module so that the carrier-based aircraft executes the actual operation instructions.

7. The prediction-based constrained pilot modeling method according to claim 6, characterized in that: Based on the glide path altitude deviation angle, the distance to the ship, the system noise in the optical guidance information, and the observation noise when the pilot observes the glide path altitude deviation angle, the altitude deviation estimator is used to calculate the corrected altitude deviation estimate, which specifically includes: The altitude deviation value is calculated according to the glide path altitude deviation angle and the distance to the ship, and a state vector is constructed based on the distance to the ship and the altitude deviation value. A state transfer equation is established according to the state vector at the current moment, the state vector at the next moment after the current moment, the system noise, and the observation noise. The expression of the altitude deviation estimate is obtained according to the state transfer equation, and the corrected altitude deviation estimate is then calculated.

8. The prediction-based constrained pilot modeling method according to claim 6, characterized in that: The fuzzy optical perception model and the fuzzy lateral perception model both include a fuzzifier, a membership function, fuzzy reasoning, a fuzzy rule base and a defuzzifier.

9. The prediction-based constrained pilot modeling method according to claim 8, characterized in that: The glide slope altitude deviation angle to be corrected is calculated based on the optical guidance information and the LSO voice command information and the fuzzy optical perception model, specifically including: The optical guidance information and the LSO voice command information are converted into fuzzy sets defined by membership functions by the fuzzifier and input into fuzzy reasoning, which matches the input fuzzy sets with rules in a fuzzy rule base to generate output fuzzy sets corresponding to the input fuzzy sets; The defuzzifier performs defuzzification processing on the output fuzzy set obtained by fuzzy inference to obtain the glide slope height deviation angle to be corrected.

10. The prediction-based constrained pilot model building method according to claim 6, characterized in that: The constraints established based on the range of values ​​of the pilot's expected instructions at each moment specifically include: obtaining the aircraft's physical structural constraints based on the aircraft structure, and obtaining the maximum and minimum values ​​of the expected instructions based on the aircraft's physical structural constraints, and establishing constraints based on the expected instructions at each moment, the maximum value of the expected instructions, and the minimum value of the expected instructions.

Images