Individualized behavior modeling method for middle and long distance air combat pilots

Abstract

The invention discloses a middle-long distance air combat pilot personalized behavior modeling method. The method comprises the following steps that S1, air combat combat data and eye movement data of a pilot during one-to-one middle-long distance air combat according to middle-long distance air combat tactical regulations are collected through an air combat flight simulation system and an eye tracker system respectively, wherein the air combat combat data and the eye movement data are obtained when the pilot conducts one-to-one middle-long distance air combat according to the middle-long distance air combat tactical regulations; the eye movement data comprises position coordinates of an eye movement fixation point when the pilot makes an air battle decision; s2, selecting a plurality of battlefield situation data to divide a cockpit view area to obtain corresponding parameter areas; according to the position coordinates of the eye movement fixation point in the step S1, determining a parameter area where the fixation point of the pilot at each corresponding sampling moment is located; calculating the average value of the proportion of the number of fixation points of the pilot to the number of fixation points of the key decision points to the parameter areas in each stage of the air combat so as to determine pilot key decision behavior indexes; s3, quantitative analysis is carried out on the pilot key decision behavior indexes, distribution characteristics of the indexes are determined, and a pilot personalized behavior model is constructed.

Description

Technical Field

[0001] The invention relates to the technical field of air combat, and in particular to a method for modeling personalized behavior of pilots in medium- and long-range air combat. Background Art

[0002] In modern warfare, air combat is a key means to seize and protect air superiority. Air combat can be divided into close-range air combat and medium- and long-range air combat according to the distance between the two warring parties. The former refers to combat conducted by pilots of the two warring parties using infrared guided missiles and aircraft guns within the visual observation range, while the latter refers to combat conducted by pilots of both parties using active radar guided missiles outside the visual observation range. With the increase in the range of air-to-air missiles and the update of airborne radar equipment, the combat range of modern air combat is constantly expanding, and medium- and long-range air combat will become the main combat mode of future air combat.

[0003] The development and application of artificial intelligence technology have made the modern air combat environment more complex, with strong game confrontation, high real-time response, incomplete information, and uncertain boundaries, which have put forward more stringent requirements on equipment performance and pilot combat level. Air combat decision-making is the core of modern air combat. At present, air combat decision-making mainly relies on human pilots, who are the decision-makers of air combat. However, the battlefield situation of modern air combat is changing rapidly, and the battlefield data provided to pilots has exploded. At the same time, the time for pilots to analyze and make decisions has been sharply shortened, which has caused great difficulties for pilots' situation judgment and battlefield decision-making, making it almost impossible for pilots to make accurate tactical decisions in time by themselves, which inevitably leads to an increase in the error rate of pilots' decision-making and a decrease in the survivability of fighter jets, which ultimately affects the effectiveness of air combat and the success or failure of missions. In the increasingly complex modern air combat, the combat of pure manned fighters has become increasingly stretched and unable to meet the mission requirements of modern air combat.

[0004] Driven by the demand for air combat and aviation technology, unmanned combat equipment represented by unmanned fighter jets has developed rapidly, showing significant advantages over purely manned fighter jets in terms of combat efficiency, risk tolerance, survivability, and all-weather and all-domain combat capabilities in modern air combat surveillance, reconnaissance, and even combat. However, in the face of a highly dynamic and real-time air combat environment, existing technologies cannot support drones to make fully autonomous air combat decisions. Drones have not yet gotten rid of the current situation of "man in the loop", and their decision-making autonomy is insufficient, which limits their application in actual combat. Therefore, manned / unmanned collaborative combat will be the main form of future air combat.

[0005] The core technology of manned / unmanned collaborative air combat is human-machine adaptation, that is, the machine algorithm must adapt to the personalized behavior of pilots. The technical key to human-machine adaptation is the modeling of pilot personalized behavior. Modeling is usually divided into two steps: qualitatively determining the key decision-making behavior indicators of pilots in air combat, and quantitatively determining the distribution characteristics of the indicators. There are four main types of pilot behavior modeling methods at home and abroad: questionnaire-based, expert knowledge-based, data-driven, and online annotation-based methods. The questionnaire-based method is not online, resulting in insufficient indicator accuracy; the expert knowledge-based method directly determines the pilot's key decision-making behavior indicators based on expert knowledge in the field, weakening the pilot's personalized behavior characteristics; the data-driven method has doubts about the credibility of the indicators due to the general "black box" properties of the data-driven model; the online annotation method requires the pilot to distract his attention during the ring flight to annotate the indicators, which destroys the air combat mission loop. Therefore, it is urgent to study the personalized behavior modeling method of pilots in medium and long-range air combat that is online, more credible, and does not damage the mission loop.

[0006] Eye tracking is a technology that can automatically and passively locate the center of the pupil and the position of the gaze point of the human eye online and estimate the direction of the line of sight. It is of great significance for studying the process and characteristics of human visual perception and cognitive processing. In medium and long-range air combat, about 80% of the information of the pilot is obtained through the visual channel, and its key decision-making behavior indicators are essentially the state parameters that the pilot focuses on when making decisions. Considering that eye tracking technology can realize online, non-destructive task loop, automatic tracking and positioning of the pilot's visual gaze point, and then determine the state parameters that the pilot focuses on when making air combat decisions, that is, its key decision-making behavior indicators, the present invention intends to use eye tracking technology to determine the key decision-making behavior indicators of pilots in medium and long-range air combat, and then construct a personalized behavior model for the pilot. While ensuring that the modeling process is non-destructive to the task loop and can be explained, the accuracy and credibility of the model are effectively improved. Summary of the invention

[0007] In view of the above-mentioned defects of the prior art, the present invention provides a method for modeling personalized behavior of pilots in medium and long-range air combat. The method is based on eye movement traction definition technology to construct a personalized behavior model of pilots in one-on-one medium and long-range air combat.

[0008] To achieve the above objectives, the present invention provides a method for modeling personalized behavior of mid-range and long-range air combat pilots, which comprises the following steps:

[0009] S1: collecting air combat engagement data and eye movement data of pilots when they conduct one-on-one medium- and long-range air combat according to medium- and long-range air combat tactical regulations through an air combat flight simulation system and an eye tracker system; the eye movement data includes the position coordinates of the eye movement fixation point of the pilots when making air combat decisions;

[0010] S2: Select several battlefield situation data to divide the cockpit visual field area to obtain corresponding parameter areas; determine the parameter area where the pilot's gaze point is located at each corresponding sampling moment according to the position coordinates of the eye movement gaze point in step S1; calculate the average value of the number of pilot's gaze points in each parameter area at each stage of air combat and each key decision point to determine the pilot's key decision behavior indicators;

[0011] S3: Conduct quantitative analysis on the pilot's key decision-making behavior indicators, determine the distribution characteristics of each indicator, and construct a personalized behavior model for the pilot, to provide support for building a manned / unmanned collaborative intelligent air combat algorithm framework that integrates the pilot's personalized behavior characteristics and improving the performance of the manned / unmanned collaborative intelligent air combat algorithm.

[0012] A further improvement of the present invention is that the air combat flight simulation system is established based on DCS World air combat simulation software; and the eye tracker is a non-contact eye tracker.

[0013] A further improvement of the present invention is that in step S2, the selected battlefield situation data includes: our aircraft speed, our aircraft altitude, our aircraft pitch angle, our aircraft yaw angle, our aircraft attack distance, our aircraft missile radar activation time, enemy aircraft speed, enemy aircraft altitude.

[0014] A further improvement of the present invention is that in step S2, the expression of the average value APFC of the number of fixation points in each parameter area is:

[0015]

[0016] Where: N i It represents the number of fixations in the ith parameter area at a certain stage or key decision point of an air combat, and n represents the number of times the pilot conducts one-on-one medium- and long-range air combat.

[0017] A further improvement of the present invention is that, in step S3, Gaussian distribution curve fitting is used to clarify the distribution characteristics of the key decision-making behavior indicators, and the mean μ and standard deviation σ of each key decision-making behavior indicator of the pilot are obtained.

[0018] The advantages of the present invention are as follows:

[0019] (1) Based on eye tracking technology, the average ratio of the number of fixations is used as an indicator to qualitatively determine the key decision-making behavior indicators of pilots at each combat phase and each key decision point in one-on-one medium- and long-range air combat online and non-destructively.

[0020] (2) Statistical methods can be used to quantitatively analyze the distribution characteristics of pilots' key decision-making behavior indicators in one-on-one medium- and long-range air combat, calculate the indicator distribution characteristic values, and construct a personalized pilot behavior model. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 : Schematic diagram of medium and long-range air combat tactics regulations.

[0022] Figure 2 : Schematic diagram of the main instrument area division in the cockpit.

[0023] Figure 3 : APFC grid heat map of S1-S5.

[0024] Figure 4 : APFC grid heat map of T1-T4.

[0025] Figure 5 : Pilots’ key decision-making behavior indicators in S1-S5 and T1-T4.

[0026] Figure 6 : Distribution histogram of pilots’ key decision-making behavior indicators.

[0027] Figure 7 : Gaussian distribution fitting curve diagram of pilot key decision-making behavior indicators. DETAILED DESCRIPTION

[0028] The following describes the embodiments of the present invention by specific examples, and those skilled in the art can easily understand other advantages and effects of the present invention from the contents disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and the details in this specification can also be modified or changed in various ways based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that the following embodiments and features in the embodiments can be combined with each other without conflict.

[0029] While some exemplary embodiments of the present invention have been described for the purpose of illustration, it should be understood that the present invention may be implemented in other ways not specifically shown in the drawings.

[0030] The present invention relies on the DCS World air combat simulation software to establish a highly realistic flight simulation environment, and conducts data collection and analysis based on this environment. The software demonstrates a high degree of restoration of real air combat in terms of simulated flight dynamics, avionics systems, and fire control systems, and has an operating environment that is extremely close to actual flight. At the same time, the present invention uses a certain type of non-contact binocular eye tracker to track the pilot's line of sight and collect eye movement data to avoid interference of the measuring instrument on the pilot.

[0031] The data collection of the present invention is carried out for 1 to 1 medium and long-range air combat. Both enemy and friendly aircraft are operated by human pilots with rich experience in medium and long-range air combat flight, and data recording is mainly carried out for more experienced pilots. Data collection is set as head-on relative flight of enemy and friendly aircraft, with an initial distance of 130 km, an initial speed of 0.6 Mach, and an initial (radio) altitude of 4 km. At the same time, in order to avoid the pilot's air combat decision-making behavior deviation caused by the type of aircraft, in the data collection, a pilot of our side flies a medium-sized twin-engine carrier-based third-generation aircraft (F) and another enemy pilot flies a light twin-engine third-generation aircraft (M) to conduct air combat confrontation. Detailed information on the two types of aircraft is shown in Table 1. The collected data types are shown in Table 2.

[0032] During data collection, the pilots did not fly completely freely, but strictly followed the medium and long-range air combat tactical regulations. Figure 1 Our pilot flies the fighter plane into the air combat battlefield, accelerates and flies level to approach the enemy in the S1 stage, and our plane continues to search and identify enemy targets; after our radar successfully detects the enemy's position and status and other information, the pilot will pull the stick to climb at T1, taking into account the energy characteristics of our plane such as speed, altitude, and maneuverability, and climb in the S2 stage to gain advantages in speed and altitude; after climbing to a certain altitude, the pilot pushes the stick to level at T2, and continues to fly level to approach the enemy in the S3 stage; when the distance between our plane and the enemy plane is shortened to a certain state, the pilot The first active radar-guided missile is launched at T3 to deter enemy aircraft. The missile launch can be carried out with a high throw or a horizontal throw trajectory. Subsequently, the pilot continues to fly level and approach the enemy aircraft in the S4 stage, and guides the deterrent missile. When the distance between our aircraft and the enemy aircraft is shortened to a certain state again, the pilot launches the second active radar-guided missile at T4 to kill the enemy aircraft. The missile launch generally chooses a horizontal throw trajectory. Subsequently, in the S5 stage, the pilot guides the lethal missile until the lethal missile shoots down the enemy aircraft or our aircraft takes an evasive maneuver, and the operation ends.

[0033] Considering that the enemy and our side have not yet entered the medium- and long-range air combat state in the initial short period of time, the present invention obtains the pilot's medium- and long-range air combat data set through time window interception, which lasts for 300 minutes.

[0034] Table 1-Basic parameters of two types of fighters included in the existing data of the present invention

[0035]

[0036] Table 2 - Data types of air combat and eye movement between the enemy and our side collected by the present invention

[0037]

[0038]

[0039] The data collected and intercepted in Table 2 are traversed frame by frame. In this embodiment, 8 battlefield situation data are selected to divide the cockpit field of view. For detailed information, see Table 3. For field of view area division, see Figure 2 .

[0040] Table-3 8 battlefield situation data tables

[0041]

[0042] In the study of the qualitative analysis of the key decision-making behavior indicators of pilots in mid- and long-range air combat, the battlefield situation and eye movement history measurement data collected in each air combat in the data set were downsampled at a frequency of 1Hz. Considering the visual field area corresponding to all 8 battlefield situation data, the average value of the proportion of fixation points APFC was selected as an indicator to measure the pilot's attention to different areas. The specific calculation method of APFC is:

[0043]

[0044] Among them, N i It represents the number of fixations in the ith parameter area at a certain stage or key decision point of an air combat, and n represents the number of times the pilot conducts one-on-one medium- and long-range air combat.

[0045] The larger the APFC value of an area, the more attention the pilot pays to the area when making air combat decisions, that is, the more attention the pilot pays to the battlefield situation data corresponding to the area. By comparing the APFC values ​​of each area of ​​the pilot in the five stages S1-S5 and the four key decision points T1-T4, the battlefield situation data that the pilot mainly pays attention to in each S stage and each T point can be determined, and then the key decision-making behavior indicators can be qualitatively analyzed. For the eight battlefield situation data, the APFC grid heat map of the pilot in the S1-S5 stage and the T1-T4 point is obtained (see Figure 3 and Figure 4 ).

[0046] Figure 3 and Figure 4 In the process, the data types of battlefield situation that the pilots are mainly concerned about in the S1-S5 stage and T1-T4 point can be obtained according to the APFC value. Combined with the pilots' mid- and long-range air combat experience and air combat domain knowledge, the pilots' key decision-making behavior indicators for mid- and long-range air combat can be obtained (see Figure 5 ).

[0047] The present invention uses a histogram to perform a preliminary analysis on the distribution characteristics of key decision-making behavior indicators, and draws a distribution histogram of all key decision-making behavior indicators in this embodiment (see Figure 6 ). Figure 6The distribution of pilots' key decision-making behavior indicators all showed a bell shape with low ends and a high middle. It was preliminarily determined that the key decision-making behavior indicators approximately presented a Gaussian distribution.

[0048] Furthermore, the present invention adopts Gaussian distribution curve fitting to clarify the distribution characteristics of key decision-making behavior indicators, and draws Gaussian distribution curve fitting diagrams of all key decision-making behavior indicators (see Figure 7 ).according to Figure 7 It can be seen that all key decision-making behavior indicators of pilots obey Gaussian distribution.

[0049] According to statistical knowledge, the probability density function of Gaussian distribution is shown in formula (2).

[0050]

[0051] Where x is the sample value, μ is the sample mean, and σ is the sample standard deviation.

[0052] The present invention selects mean μ and standard deviation σ as the characteristic values ​​of the distribution of pilot key decision-making behavior indicators. The calculation method of the two is:

[0053]

[0054] Among them, x i It represents the i-th sample value of a certain type of indicator data, and n represents the number of samples of this type of indicator data.

[0055] The present invention calculates the mean μ and standard deviation σ of each key decision-making behavior index of the pilot, see Table 4.

[0056] Table 4 - Mean μ and standard deviation σ of each key decision-making behavior index of the pilot in the present invention

[0057]

[0058] According to the distribution mean and standard deviation of the pilots’ key decision behavior indicators in the five combat phases S1-S5 and the four key decision points T1-T4 in Table 4, the personalized behavior characteristics of pilots in 1-on-1 medium- and long-range air combat can be determined, and a personalized behavior model of pilots can be constructed based on the distribution characteristics of the key decision behavior indicators.

[0059] The above embodiments are merely illustrative of the principles and effects of the present invention, and are not intended to limit the present invention. Anyone familiar with the art may modify or alter the above embodiments without departing from the spirit and scope of the present invention. Therefore, all equivalent modifications or alterations made by a person of ordinary skill in the art without departing from the spirit and technical concept disclosed by the present invention shall still be covered by the claims of the present invention.

Claims

1. A method for modeling personalized behavior of mid-range and long-range air combat pilots, characterized in that: The following steps are involved: S1: collecting air combat engagement data and eye movement data of pilots when they conduct one-on-one medium- and long-range air combat according to medium- and long-range air combat tactical regulations through an air combat flight simulation system and an eye tracker system; the eye movement data includes the position coordinates of the eye movement fixation point of the pilots when making air combat decisions; S2: Select a number of battlefield situation data to divide the cockpit visual field area to obtain corresponding parameter areas; determine the parameter area where the pilot's gaze point is located at each corresponding sampling moment according to the position coordinates of the eye movement gaze point in step S1; Calculate the average ratio of the number of pilots' gaze points on each parameter area at each stage of air combat and each key decision point to determine the pilot's key decision-making behavior indicators; S3: Conduct quantitative analysis on pilots’ key decision-making behavior indicators, determine the distribution characteristics of each indicator, and construct a personalized behavior model for pilots.

2. The method for modeling personalized behavior of mid-range and long-range air combat pilots according to claim 1, characterized in that: The air combat flight simulation system is established based on DCS World air combat simulation software; the eye tracker is a non-contact eye tracker.

3. The method for modeling personalized behavior of mid-range and long-range air combat pilots according to claim 1, characterized in that: In step S2, the selected battlefield situation data include: our aircraft speed, our aircraft altitude, our aircraft pitch angle, our aircraft yaw angle, our aircraft attack distance, our aircraft missile radar opening time, enemy aircraft speed, enemy aircraft altitude.

4. The method for modeling personalized behavior of mid-range and long-range air combat pilots according to claim 1, characterized in that: In step S2, the expression of the average APFC of the number of fixations in each parameter area is: Where: N i represents the number of fixation points in the i-th parameter area, and n represents the number of times the pilot conducts one-on-one medium- and long-range air combat.

5. The method for modeling personalized behavior of mid-range and long-range air combat pilots according to claim 1, characterized in that: In step S3, Gaussian distribution curve fitting is used to clarify the distribution characteristics of key decision-making behavior indicators, and the mean μ and standard deviation σ of each key decision-making behavior indicator of the pilot are obtained.

Images