A Pilot Decision-Making Inference Method Based on Dynamic Optimization of Multi-Level Fuzzy Branch Structure

Abstract

This invention relates to a pilot decision-making deduction method based on dynamic optimization of a multi-level fuzzy branch structure, comprising the following steps: 1) quantifying the set of decision states; 2) quantifying the set of fuzzy rules; 3) designing multiple key influencing characteristics such as hysteresis factors, pilot skill level stability factors, and search factors to establish a multi-level fuzzy branch structure relationship that reflects the pilot's decision-making logic, and using it as the pilot decision-making deduction mechanism in the human-machine system; 4) dynamically optimizing the multi-level decision logic and the decision-making deduction sequence results during the time-series recursion process. Compared with existing technologies, this invention introduces a logical thinking mode of human cognition mechanisms and decision-making characteristics by designing multiple decision levels and their rule changes, overcoming the shortcomings of traditional pilot models such as fixedness, static parameters, and poor transferability, while also possessing the advantages of adaptability to various aircraft types and wide application range.

Description

Technical Field

[0001] This invention relates to the field of control decision-making inference technology in human-machine systems, and in particular to a pilot decision-making inference method based on dynamic optimization of multi-level fuzzy branch structure. Background Technology

[0002] In the decision-making process of human-machine systems, numerous factors influence pilot decisions, such as the complexity of different decision-making levels, the pilot's cognitive abilities, and the layout design of the user interface. Within the pilot's decision-making process, determined by its logical structure, human factors influence decisions at different levels. Therefore, designing an effective decision-making logical deduction structure to express the pilot's decision-making process over time is crucial for analyzing human-machine system decisions. To reasonably represent the pilot's deduction process during target tracking and execution, and considering their inherent human characteristics, this paper designs an objective, efficient, and universally applicable decision-making framework, with its logical structure designed based on the characteristics of pilot decision-making.

[0003] Currently, the mainstream methods for expressing pilot decision-making processes include:

[0004] 1) Acquire pilots’ physiological indicators and behavioral actions through various sensors in experimental systems or real flight tests, such as heart rate, eye movement, electroencephalogram (EEG), electromyography (EMG), and hand gestures, and build models after data collection.

[0005] 2) Transfer the controller design to the pilot model, thereby treating the pilot as a real-time controller that completes flight actions, and solve and establish the model for the controller under the conditions of stability and robustness.

[0006] 3) Methods for designing qualitative models, such as the three-layer static structure of perception-decision-execution, can express the pilot's decision-making process and thus analyze it.

[0007] With the advancement of human-machine system technology and information, the design philosophy for aircraft is gradually converging towards a pilot-centric approach, starting from the pilot's decision-making patterns and designing a cockpit adapted to the pilot's decision-making process. Currently, none of the three pilot decision-making model methods mentioned above can fully reflect the pilot's decision-making reasoning logic and deduction process: While the physiological indicators and behavioral actions obtained in method 1) have some reference value in mechanized and semi-automated systems, they cannot accurately describe the complex causal effects of rule judgment and dynamic optimization; the pilot decision-making model established by designing a controller in method 2) inevitably suffers from staticity and lag; in method 3), the method of establishing a quantitative model to perceive the decision-making execution structure incorporates the pilot's perception of external states, cognitive decision-making about goals, and pilot decision-making actions into the model, but ignores the inherent connections between levels in the pilot's decision-making process. Furthermore, the assumptions of these methods are all based on the pilot being a standardized robot model, with little consideration for the pilot's human factors in the decision-making process, making them difficult to apply in practical human-machine systems. Therefore, regarding the pilot's logical judgment decision deduction process involved in this invention, the assumptions used in the above methods are not reasonable enough. Summary of the Invention

[0008] The purpose of this invention is to overcome the shortcomings of the existing technology and provide a pilot decision-making inference method based on dynamic optimization of multi-level fuzzy branch structure.

[0009] The objective of this invention can be achieved through the following technical solutions:

[0010] A pilot decision-making inference method based on dynamic optimization of multi-level fuzzy branch structure includes the following steps:

[0011] 1) Based on the basic structural components of the human-machine system, quantitatively express the set of decision states;

[0012] 2) Based on the judgment levels of the pilot's decision-making deduction, quantitatively express a multi-level set of fuzzy rules;

[0013] 3) Design self-terminating sequence search factor, long short-term memory factor, hysteresis factor, pilot skill level stability factor, state transition factor, and transition time for different decision-making levels. Consider the observable state data and controllable input decisions of the human-machine system, construct a formal expression of multi-level branch structure relationships, and realize a single-step instantiation mechanism for decision deduction.

[0014] 4) Perform hierarchical fuzzy rule dynamic optimization, that is, realize the decision inference mechanism through continuous dynamic iteration, and adjust the potential fuzzy rules of single-step instantiation decision.

[0015] 5) Based on the decision state set and the multi-level fuzzy rule set, the pilot's decision inference sequence results are derived through iterative decision inference and rule optimization mechanism.

[0016] In step 1), for the set of quantitatively expressed decision states, the metadata in the set of quantitatively expressed decision states is used to construct instances in the decision inference mechanism through the abstract representation of each level.

[0017] The instance construction in the aforementioned decision-making inference mechanism specifically includes:

[0018] Establish abstract state elements corresponding to route targets in the human-machine system;

[0019] Establish abstract state elements corresponding to observable instruments in the human-machine system;

[0020] Establish abstract state elements corresponding to the pilot's skill level in the human-machine system;

[0021] Establish abstract state elements corresponding to the human-machine system's control devices;

[0022] The quantified decision state set is the basic metadata for iterative decision deduction in a multi-level branching structure.

[0023] In step 2), for the quantification of the multi-level fuzzy rule set, the quantification of fuzzy decision logic rules at different levels is achieved by designing a classifier for route targets, a scheduler for observable instruments, a judge for pilot skill level stability factors, a scheduler for control equipment, and a judge for control method characteristics.

[0024] Step 2) specifically includes the following steps:

[0025] 21) Based on the trajectory elements of the route target and the flight status elements of the current aircraft system, establish a fuzzy judgment boundary set for target switching;

[0026] 22) Based on the selection set of observable instruments and the historical selection status of the instrument sequence, establish the instrument selection sequence;

[0027] 23) Based on the rate of change of the skill level stability factor, establish a fuzzy boundary set for pilots switching skill levels;

[0028] 24) Based on the controllable state elements provided by the controllable equipment and the historical selection state of the controllable equipment, establish a fuzzy scheduling set for the selection of the controllable equipment.

[0029] 25) Based on the state elements of the control method and the flight state elements directly associated with the current human-machine system, establish a fuzzy boundary set for the pilot to switch control methods according to the target;

[0030] 26) All the fuzzy rule elements covered in the above different levels will participate in the iterative dynamic adjustment as a rule set in the decision inference mechanism.

[0031] In step 3), the decision-making hierarchy includes the target decision-making layer, the observable instrument decision-making layer, the capability level characteristic layer, the control equipment decision-making layer, and the state judgment decision-making layer.

[0032] In step 3), a multi-level composite fuzzy inference engine considers the self-terminating sequence search factor, hysteresis factor, and long short-term memory factor, and correlates them with the transition probability and transition time of the state judgment layer. This enables the quantified rule logic engine to be applied to pilot decision-making deduction, obtaining effective control quantities for load cognitive decision-making. Specifically, this includes the following steps:

[0033] 31) Based on the data status output by the human-machine system at the current decision-making moment, calculate the cost function between the current target and the state, and complete the decision planning of the target layer according to the fuzzy rule set of the target decision layer;

[0034] 32) Introduce a self-terminating sequence search factor to evaluate the time cost in selecting instruments. Calculate the time spent and the final searched instrument status based on the instrument location to complete the decision planning of the instrument selection layer.

[0035] 33) Based on skill level characteristics, design a pilot's ability level factor classification for different flight missions, and provide input to the logic decision-maker of this layer;

[0036] 34) Based on the historical sequence of the control equipment, design the pilot weight factor of long short-term memory, and output the pilot's choice of control equipment at the decision layer according to the fuzzy decision rule of the control equipment selection layer, thereby determining the pilot's decision input to the human-machine system.

[0037] 35) Based on the target layer, instrument selection layer, skill level characteristic layer and control equipment selection layer, the observable state variables and controllable decision variables are obtained, and the decision state variables are finally effectively estimated according to the state transition time and state transition probability.

[0038] In step 4), the dynamic optimization of hierarchical fuzzy rules is achieved by designing a cost function and feedback loop, and comparing the outputs of fuzzy inferencers at different levels with the human-machine system state and cost indicators. Specifically, this includes the following steps:

[0039] 41) Dynamic optimization of the flight target layer logic decision-maker: By combining the desired target with the threshold of the logic decision-maker, information obtained from the observable flight state is introduced to adjust the boundary range of the elements in the rule set.

[0040] 42) Dynamic optimization of the instrument selection layer logic scheduler: By comparing the number of planned instrument elements to be searched in the current decision-making process with the number of instrument elements to be searched in the historical memory, unnecessary instrument elements are avoided, thereby optimizing the self-end sequence search factor that affects the logic scheduling in this layer;

[0041] 43) Dynamic optimization of the ability level logic decision-maker: Consider the impact of the instantaneous changes in the current human-machine system state on the pilot's skill level stability factor, and thus adjust the boundaries associated with the state in the decision rule set;

[0042] 44) Dynamic optimization of the logic scheduler at the controllable equipment layer: As an optimization adjustment of the pilot's controllable decision input in the human-machine system, based on the changes in the pilot's skill level factor and considering the weight changes of the long short-term memory factor, the pilot's selection of controllable equipment based on flight status and personal skill level is simulated.

[0043] 45) Comprehensive optimization of the state decision layer logic inferencer: By combining the time delay factor of composite hierarchical logic inference and through the gradient change of the state, optimize the cost equation of the current objective, compare it with the expected minimum value, and optimize the transition time delay factor and transition probability of the current decision state.

[0044] In step 5), for the decision deduction sequence, based on revealing the element correlation in the pilot's cognitive decision deduction method, a recursive adaptive adjustment pilot decision deduction mechanism is constructed, and the set of fuzzy judgment rules on which the pilot's cognitive decision depends is optimized through a closed-loop feedback mechanism.

[0045] Step 5) specifically includes:

[0046] 51) Establish the correlation and potential correlation of multiple factors in various pilot cognitive decision-making mechanisms;

[0047] 52) It comprehensively reflects the hierarchical characteristics of decision-making in the pilot's decision-making simulation process and the cognitive logic structure in actual mission scenarios;

[0048] 53) By iteratively calculating, multiple intrinsic state variables are updated, thereby obtaining a stable decision output;

[0049] 54) The synchronously updated hierarchical fuzzy rules integrate variables such as lag factors, transfer factors, and pilot skill level factors involved in pilot cognitive decision-making deduction. Based on the dynamic adaptation of this rule set, the stability of the human-machine system in decision-making deduction is guaranteed.

[0050] Compared with the prior art, the present invention has the following advantages:

[0051] I. This invention overcomes the shortcomings of fixed pilot models and static human parameters in the prior art, and optimizes the robustness of pilot decision-making inference.

[0052] Second, by designing different decision-making levels and their rules to adapt to changes, this invention introduces a logical thinking pattern of human cognitive mechanisms and decision-making characteristics, thereby improving the accuracy of pilots' decision-making deductions.

[0053] Third, this invention introduces self-terminating sequence search factors, hysteresis factors, long short-term memory factors, and pilot skill level stability factors to measure the correlation of pilot decision-making factors at different decision-making levels, thereby avoiding the insufficiency of factors considered in existing decision-making techniques.

[0054] Fourth, this invention takes the construction of a multi-level fuzzy branch decision structure as its core, considers the observable state quantities and controllable decision quantities of the pilot in the human-machine system, and designs a Markov dynamic optimization mechanism for state transition probability and transition time, thereby improving the pilot's decision-making process. Attached Figure Description

[0055] Figure 1 This is a schematic diagram of the process of the present invention.

[0056] Figure 2 This is a quantitative diagram illustrating the states and rules of different decision-making levels in a decision-making simulation structure.

[0057] Figure 3 This is a schematic diagram illustrating the logical connections and detailed operation between different decision-making levels in a static decision simulation.

[0058] Figure 4 This is a schematic diagram illustrating the dynamic optimization of the logical structure layer during the decision-making process. Detailed Implementation

[0059] The present invention will now be described in detail with reference to the accompanying drawings and specific embodiments.

[0060] Example

[0061] like Figure 1 As shown, this invention provides a pilot decision-making inference method based on dynamic optimization of a multi-level fuzzy branch structure, including the following:

[0062] (1) In this embodiment, the human-machine system data of the aircraft and the pilot will be loaded into the pilot's decision-making process during operation;

[0063] (2) The human-machine system data includes aircraft operating status data: altitude, speed, vertical speed, attitude angle, azimuth, angle of attack and sideslip angle; cockpit instrument data: instruments with readable parameters, such as PFD and ND; controllable equipment such as control stick, pedals and throttle; pilot status data: pilot control variables; the above data is the basis for the generation of decision-making and deduction sequences in this invention. In this invention, the human-machine system specifically refers to an aircraft that can be controlled by a pilot.

[0064] (3) The hierarchical data with different logical structures include the set of quantified actions and states and the set of quantified rules, and the data of this structure layer will be dynamically updated throughout the recursive pilot decision-making process;

[0065] (4) The quantified action and state set includes: elements related to the pilot's action set, classification of elements for judging flight target states, element division of readable instruments, and abstraction of elements of pilot-controllable equipment; the quantification rule set includes the classification of judicious targets at different levels; to enable technicians to have a more comprehensive understanding of the channel quantification process, relevant details are provided in... Figure 2 Explanation will be provided in the following section;

[0066] (5) In the decision-making process, the target layer, instrument layer, equipment layer, skill level layer, and method characteristic layer together form the complete logical deduction structure of the pilot's decision-making. The logical relationship between these different decision-making levels is a sequential association relationship. After each layer completes the decision-making judgment, the result is passed to the next decision-making level for use.

[0067] (6) The self-terminating sequence search factor, hysteresis factor, long short-term memory factor, and pilot skill level stability factor play roles in different decision-making levels. At the same time, the hierarchical optimization mechanism optimizes the transition probability and transition time by saving the results of different decision-making levels in the recursive decision-making process, thereby adjusting the parameter factors in the decision-making structure.

[0068] (7) The recursive method specifically refers to using a multi-level pilot decision-making logic structure as a model, based on the time sequence, and dynamically optimizing the rule parameters to obtain the output of the decision deduction sequence. To provide technicians with a more comprehensive understanding of the workload and automation level evaluation mechanism of this invention, relevant details are provided below. Figure 3 and Figure 4 The explanation is provided below.

[0069] like Figure 2 As shown, the main steps in quantifying pilot status and rules are:

[0070] (1) Establishing the hierarchy that influences pilot decision-making can be done in two steps. The first step is to quantify the input and output elements of the decision-making hierarchy; the second step is to quantify the judgment process within the corresponding decision-making hierarchy.

[0071] (2) The information obtained by the pilot from the instruments includes visual observation of aircraft instrument information, visual information, auditory information, and somatosensory information (gravity perception, acceleration perception, etc.). Among them, visual information is directly related to the pilot's control behavior and requires the pilot to actively allocate certain attention resources for observation. The other information can be passively obtained and play a compensating role. The state quantities that can be searched from the instruments in the instrument layer are flight altitude, speed, attitude angle, heading, latitude and longitude.

[0072] (3) According to the importance of their impact on pilot operation, they are divided into three categories from high to low: direct impact, indirect impact, and weak impact. Direct impact: speed, position, attitude angle; indirect impact: acceleration, angular velocity, etc.; weak impact: fuel gauge, thermometer, auxiliary system status indication, etc.

[0073] (4) The set of instruments observable by the pilot is quantified as {st1, st2, ..., st N The flight-related parameters defined in different instruments vary.

[0074] (5) In the equipment layer, the equipment that the pilot is allowed to operate is divided into five categories: control stick, foot pedal, throttle, other equipment and individual air controls;

[0075] (6) When quantifying the fuzzy selection rules in the equipment layer, the pilot will use the target judgment result and the observable state quantity obtained from the instrument layer as the input variables of the fuzzy rules; and use the controllable equipment finally selected by the pilot in the equipment control layer as the quantization output result of that layer.

[0076] (7) At the skill level level, the ratio of the pilot’s deviation from the aforementioned control to the control target is taken into account for the change in the pilot’s control deviation.

[0077] (8) The quantitative input in the skill level layer is the judgment output of the target layer, and the quantitative output result is the ability level stability factor of different levels;

[0078] (9) In the method characteristic layer, the control ratio caused by the target deviation in the pilot decision-making simulation is used as the control variable that disturbs the pilot; the input ratio of the pilot to the control quantity of a certain type is quantified.

[0079] (10) In the method characteristic layer, the aforementioned output of the equipment selection result, skill level stability factor and the judgment output of the target layer are used as the quantitative input elements of this layer; the obtained control ratio is used as the quantitative output result of this layer;

[0080] like Figure 3 As shown, the main steps of a static instance that can be considered an independent pilot decision-making simulation process are as follows:

[0081] (1) The model for acquiring visual information is not fixed throughout the flight process. During takeoff and landing, the aircraft state changes rapidly and the requirements for flight accuracy are high. Pilots must pay close attention to information directly related to flight performance, such as altitude, speed, and heading angle. During this process, the pilot's effective information acquisition frequency is the highest. However, during stable processes such as altitude hold-cruise, the pilot can acquire information at a lower frequency and can divert some attention to observe secondary instrument information such as fuel gauge and temperature gauge. Therefore, the visual information acquisition part of the pilot model is directly related to the current flight mission.

[0082] (2) Introducing a self-terminating sequence search factor The time it takes for a pilot to scan multiple identical instruments can be approximated by a linear equation: Where ST is the total search time, i.e., the total time spent acquiring instrument information; a p It is the inherent time constant, including head-turning time, attention shifting time, etc.; b i is the time constant for the search and dwell time of the i-th instrument; N is the number of instruments;

[0083] (3) Obtain the self-terminating sequence search factor after each search decision. The state values ​​obtained from the search will be saved as historical information h;

[0084] (4) Define the probability of a pilot selecting different instruments during a search, p = {p1, p2, ..., p...} N},at the same time Introducing a long short-term memory factor l, the historical device selection probability is adjusted to...

[0085] (5) From the different selected instruments and equipment, obtain the observable state quantities u1={x1,x2,...,x} selected by the pilot. m}=∑p i 'st i ;

[0086] (6) Define the output state elements of the target decision layer as three categories: near, far, and nearby; define the set of input variables of the target decision layer related to these three categories of state elements {x1, x2, ..., x...} m}; Calculate the deviation between the observable state elements and the target flight point state elements at this time, Δ=∑||x i -y i || 2 The deviation value is then used as an input variable for the target layer.

[0087] (7) Define the membership functions of the target layer input and output variables according to the rules of target decision-making. Where c is the mean and σ is the standard deviation;

[0088] (8) The fuzzy judgment result u1 of the target decision layer can be output based on the membership function;

[0089] (9) In the target decision-making layer, the ratio of the current state to the expected target point is defined as the decision factor ss = ∑||xy|| / ∑||y||.

[0090] (10) The choice of equipment layer also depends on the long short-term memory factor. The pilot's choice of equipment will depend on the frequency of equipment use at the aforementioned points in time.

[0091] (11) where q = {q1, q2, ..., q M} is defined as the probability of choosing different devices, and M is the number of devices in the device decision layer; in the device layer, each device is defined as having a corresponding manipulation quantity;

[0092] (12) Based on probability q i By selecting the size, the output result u2 of the current device decision layer is obtained;

[0093] (13) In the skill level layer, the introduced skill level stability factor dynamically assigns weights to the deviations caused by stability in the pilot decision-making process.

[0094] (14) Define the ability stabilization factor of input variables The output variable is the mean ω of the decision bias distribution, with the mean range limited to [ω]. min ,ω max Between; the membership function of fuzzy rules in the skill level layer is defined as an S-shaped membership function. Where a is the slope and c is the inflection point of the function;

[0095] (15) The judgment result u3 of the skill level layer can be fitted according to the membership function;

[0096] (16) In the method characteristic layer, the target factor calculated above is used as the input variable, and the pilot's ability stability level factor obtained above is used as the input variable. Figure 2 The skill level dynamic stability factor) is used as the median value of the parameter c in the membership function, and a and b are the upper and lower stability threshold boundaries of the rule judgment function, respectively.

[0097] (17) Define the membership function of the pilots in this layer according to the rules of the method characteristic layer.

[0098] (18) Based on the membership function, x can be fitted and solved to obtain the judgment result u4 of the method characteristic layer;

[0099] (19) Based on the above, the obtained U t ={u1,u2,u3,u4}, to obtain the decision-making and controllable result δ under the influence of the pilot's ability level, the current observable state of the pilot, and the controllable input selected by the pilot;

[0100] (20) There is a human-induced time delay in the process of pilots acquiring aircraft instrument or visual information. It takes a certain amount of time for pilots to observe each aircraft instrument and obtain information. This part uses a self-terminating sequence search factor. Expression; simultaneously, a hysteresis factor is introduced into the equipment operation during the decision-making process. This indicates the pilot's time of operation with the equipment;

[0101] (21) The obtained control quantity Total time delay Input the data into the aircraft simulation system to obtain the corresponding aircraft system state response;

[0102] (22) Detect whether the target is within the set flight area and whether a target switch has occurred;

[0103] (23) Check whether the current proportion of flight targets has met the requirements, and obtain the cost factor to provide to the skill level layer;

[0104] (24) The pilot's dynamic decision-making simulation does not rely on the assumption that the pilot's subjective workload is directly related to the pilot's operational performance. Instead, it derives the final decision control input from the pilot to the aircraft based on the reasoning structure relationships at different decision-making levels. Figure 3 The decision-making logic relationships between the various factors and related variables listed.

[0105] like Figure 4 As shown, the steps of pilot dynamic decision-making deduction and fuzzy branch structure adaptive optimization are as follows:

[0106] (1) Dynamic optimization of multi-level fuzzy branch structure in pilot decision deduction mainly considers the influence of time-varying factors of human-machine system state on pilot decision-making.

[0107] (2) In Figure 4 In this context, dynamic optimization of fuzzy branching structures refers to the process by which a pilot's multi-level logical decision-making structure, through a recursive time progression, optimizes the various components involved in the decision-making structure. Figure 3 Adjust the factor variables in the data;

[0108] (3) Among them, the judgment of the sub-objective is to calculate the cost function by predicting the transition cost of the future state and considering the historical decision cost function. Through optimization, the dynamic optimization of the structure is completed.

[0109] (4) Calculate the pilot's decision cost function w = ∑||xy|| after completing the current decision action. 2 ;

[0110] (5) Considering the long short-term memory factor l, calculate the cumulative cost function from the start of the decision to the current time t.

[0111] (6) Based on the normalized weight ratios of different levels in the decision-making layer, optimize the gradient in the direction of solving the decision parameters.

[0112] (7) Within the decision history segment, the past costs are combined with the possible future costs, and the currently obtainable rule parameters are solved by the minimax algorithm. Where the function L(·) represents the future decision cost under the current gradient parameter optimization, and α represents the different consideration weights of historical and future rewards;

[0113] (8) For the instrument selection layer and the equipment decision layer, the optimization of adaptive rule parameters is related to the historical selection information of readable parameter instruments and controllable parameter equipment; considering the selection probabilities p and q of instruments and equipment respectively, the change in their probabilities is as follows: Here are symbols This represents the principle of satisfying maximum and minimum functional coverage between historical information and the current state;

[0114] (9) Based on the rule parameters obtained from the formula, the pilot decision-making model will use the updated rule parameters and weight coefficients in the next decision-making process;

[0115] (10) The constructed multi-level dynamic optimization fuzzy rule model establishes the direct or indirect dependence of pilot judgments on the aforementioned decisions among decision-making levels. The observation model of decision-making at each level has multiple measurement indicators. Furthermore, the factor variables introduced between the corresponding levels have a certain degree of applicability and practicality for the incomplete structural model considering human factors. Due to the existence of measurement errors, the correlation coefficient of the variable variables changes at a predetermined ratio.

Claims

1. A pilot decision-making inference method based on dynamic optimization of multi-level fuzzy branch structure, characterized in that, Includes the following steps: 1) Based on the basic structural components of the human-machine system, quantitatively express the set of decision states; 2) Based on the judgment levels of the pilot's decision-making deduction, quantitatively express a multi-level set of fuzzy rules; 3) For different decision-making levels, design self-terminating sequence search factors, long short-term memory factors, hysteresis factors, pilot skill level stability factors, state transition factors, and transition times. Consider the observable state data and controllable input decisions of the human-machine system, construct a formal expression of multi-level branch structure relationships, and realize a single-step instantiation mechanism for decision deduction. The decision-making level data with different logical structures includes quantified action and state sets and quantified rule sets, and the decision-making level data is dynamically updated throughout the recursive pilot decision deduction process. 4) Perform hierarchical fuzzy rule dynamic optimization, that is, realize the decision inference mechanism through continuous dynamic iteration, and adjust the potential fuzzy rules of single-step instantiation decision. 5) Based on the decision state set and the multi-level fuzzy rule set, the pilot's decision inference sequence results are derived through iterative decision inference and rule optimization mechanism.

2. The pilot decision-making inference method based on dynamic optimization of multi-level fuzzy branch structure according to claim 1, characterized in that, In step 1), for the set of quantitatively expressed decision states, the metadata in the set of quantitatively expressed decision states is used to construct instances in the decision inference mechanism through the abstract representation of each level.

3. The pilot decision-making inference method based on dynamic optimization of multi-level fuzzy branch structure according to claim 2, characterized in that, The instance construction in the aforementioned decision-making inference mechanism specifically includes: Establish abstract state elements corresponding to route targets in the human-machine system; Establish abstract state elements corresponding to observable instruments in the human-machine system; Establish abstract state elements corresponding to the pilot's skill level in the human-machine system; Establish abstract state elements corresponding to the human-machine system's control devices; The quantified decision state set is the basic metadata for iterative decision deduction in a multi-level branching structure.

4. The pilot decision-making inference method based on dynamic optimization of multi-level fuzzy branch structure according to claim 1, characterized in that, In step 2), for the quantification of the multi-level fuzzy rule set, the quantification of fuzzy decision logic rules at different levels is achieved by designing a classifier for route targets, a scheduler for observable instruments, a judge for pilot skill level stability factors, a scheduler for control equipment, and a judge for control method characteristics.

5. The pilot decision-making inference method based on dynamic optimization of multi-level fuzzy branch structure according to claim 4, characterized in that, Step 2) specifically includes the following steps: 21) Based on the trajectory elements of the route target and the flight status elements of the current aircraft system, establish a fuzzy judgment boundary set for target switching; 22) Based on the selection set of observable instruments and the historical selection status of the instrument sequence, establish the instrument selection sequence; 23) Based on the rate of change of the skill level stability factor, establish a fuzzy boundary set for pilots switching skill levels; 24) Based on the controllable state elements provided by the controllable equipment and the historical selection state of the controllable equipment, establish a fuzzy scheduling set for the selection of the controllable equipment. 25) Based on the state elements of the control method and the flight state elements directly associated with the current human-machine system, establish a fuzzy boundary set for the pilot to switch control methods according to the target; 26) All the fuzzy rule elements covered in the above different levels will participate in the iterative dynamic adjustment as a rule set in the decision inference mechanism.

6. The pilot decision-making inference method based on dynamic optimization of multi-level fuzzy branch structure according to claim 1, characterized in that, In step 3), the decision-making hierarchy includes the target decision-making layer, the observable instrument decision-making layer, the capability level characteristic layer, the control equipment decision-making layer, and the state judgment decision-making layer.

7. The pilot decision-making inference method based on dynamic optimization of multi-level fuzzy branch structure according to claim 6, characterized in that, In step 3), a multi-level composite fuzzy inference engine considers the self-terminating sequence search factor, hysteresis factor, and long short-term memory factor, and correlates them with the transition probability and transition time of the state judgment layer. This enables the quantified rule logic engine to be applied to pilot decision-making deduction, obtaining effective control quantities for load cognitive decision-making. Specifically, this includes the following steps: 31) Based on the data status output by the human-machine system at the current decision-making moment, calculate the cost function between the current target and the state, and complete the decision planning of the target layer according to the fuzzy rule set of the target decision layer; 32) Introduce a self-terminating sequence search factor to evaluate the time cost in selecting instruments. Calculate the time spent and the final searched instrument status based on the instrument location to complete the decision planning of the instrument selection layer. 33) Based on skill level characteristics, design a pilot's ability level factor classification for different flight missions, and provide input to the logic decision-maker of this layer; 34) Based on the historical sequence of the control equipment, design the pilot weight factor of long short-term memory, and output the pilot's choice of control equipment at the decision layer according to the fuzzy decision rule of the control equipment selection layer, thereby determining the pilot's decision input to the human-machine system. 35) Based on the target layer, instrument selection layer, skill level characteristic layer and control equipment selection layer, the observable state variables and controllable decision variables are obtained, and the decision state variables are finally effectively estimated according to the state transition time and state transition probability.

8. The pilot decision-making inference method based on dynamic optimization of multi-level fuzzy branch structure according to claim 1, characterized in that, In step 4), the dynamic optimization of hierarchical fuzzy rules is achieved by designing a cost function and feedback loop, and comparing the outputs of fuzzy inferencers at different levels with the human-machine system state and cost indicators. Specifically, this includes the following steps: 41) Dynamic optimization of the flight target layer logic decision-maker: By combining the desired target with the threshold of the logic decision-maker, information obtained from the observable flight state is introduced to adjust the boundary range of the elements in the rule set. 42) Dynamic optimization of the instrument selection layer logic scheduler: By comparing the number of planned instrument elements to be searched in the current decision-making process with the number of instrument elements to be searched in the historical memory, unnecessary instrument elements are avoided, thereby optimizing the self-end sequence search factor that affects the logic scheduling in this layer; 43) Dynamic optimization of the ability level logic decision-maker: Consider the impact of the instantaneous changes in the current human-machine system state on the pilot's skill level stability factor, and thus adjust the boundaries associated with the state in the decision rule set; 44) Dynamic optimization of the logic scheduler at the controllable equipment layer: As an optimization adjustment of the pilot's controllable decision input in the human-machine system, based on the changes in the pilot's skill level factor and considering the weight changes of the long short-term memory factor, the pilot's selection of controllable equipment based on flight status and personal skill level is simulated. 45) Comprehensive optimization of the state decision layer logic inferencer: By combining the time delay factor of composite hierarchical logic inference and through the gradient change of the state, optimize the cost equation of the current objective, compare it with the expected minimum value, and optimize the transition time delay factor and transition probability of the current decision state.

9. The pilot decision-making inference method based on dynamic optimization of multi-level fuzzy branch structure according to claim 1, characterized in that, In step 5), for the decision deduction sequence, based on revealing the element correlation in the pilot's cognitive decision deduction method, a recursive adaptive adjustment pilot decision deduction mechanism is constructed, and the set of fuzzy judgment rules on which the pilot's cognitive decision depends is optimized through a closed-loop feedback mechanism.

10. The pilot decision-making inference method based on dynamic optimization of multi-level fuzzy branch structure according to claim 9, characterized in that, Step 5) specifically includes: 51) Establish the correlation and potential correlation of multiple factors in various pilot cognitive decision-making mechanisms; 52) It comprehensively reflects the hierarchical characteristics of decision-making in the pilot's decision-making simulation process and the cognitive logic structure in actual mission scenarios; 53) By iteratively calculating, multiple intrinsic state variables are updated, thereby obtaining a stable decision output; 54) The synchronously updated hierarchical fuzzy rules integrate the variables involved in pilot cognitive decision-making, such as lag factors, transfer factors, and pilot skill level factors. Based on the dynamic adaptation of this rule set, the stability of the human-machine system in decision-making is guaranteed.

Images