Optimization method and system for cooperative occupation of our two machines in typical bullet borrowing scene

Abstract

The invention discloses an optimization method and system for our-party dual-computer cooperative occupation in a typical bomb borrowing scene, and relates to the technical field of aerospace, and the method comprises the steps: constructing a bomb borrowing air combat scene, and defining a dual-computer role task; respectively constructing a fighter three-degree-of-freedom dynamics and kinematics model, a weapon war area and non-escapable area model and a beyond-visual-range air combat situation reward and punishment situation function; carrying out enemy multi-step track prediction by adopting an LSTM model, and determining enemy state information under each time step; according to the enemy state information under each time step, adopting an improved particle swarm algorithm to solve and obtain an optimal occupation route; and according to the optimal occupied route, adopting an improved route planning algorithm to control the double aircraft to arrive at a specified position. According to the method and the device, dual-machine cooperative occupation optimization in a typical bullet borrowing scene can be realized, and a theoretical support is provided for multi-UCAV cooperative combat in an over-the-horizon air combat.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of aerospace technology, in particular to a method and system for optimizing the cooperative occupation of a typical borrowed missile scene by our two aircraft. BACKGROUND

[0002] With the development of airborne radar and missile technology, beyond-visual-range air combat (BVR) has become the core form of modern air combat. Multi-unmanned combat aircraft (UCAV) has shown significant effectiveness in cooperative air combat due to its low cost, high mobility, and zero casualties.

[0003] However, there are three limitations in the prior art: 1) Single scene: Most air combat decision-making models focus on single aircraft combat (one-on-one), and lack methods for multi-UCAV cooperative occupation.

[0004] 2) Lack of cooperative mechanism: Existing multi-aircraft cooperative research focuses on within-visual-range (WVR) dogfight or target assignment problems, ignoring the key step of cooperative occupation of attack positions.

[0005] 3) Insufficient dynamic modeling: Traditional methods (such as matrix game) do not incorporate the geometric constraints of the weapon engagement zone (WEZ), resulting in the selection of occupation areas being disconnected from the actual air combat dynamic characteristics (such as the coupled effects of distance, azimuth, and entry angle).

[0006] In summary, there is an urgent need to provide an optimization method and system for cooperative occupation of two aircraft to achieve optimization of cooperative occupation of two aircraft in a typical borrowed missile scene, solve the problems of single scene, lack of cooperative mechanism, and insufficient dynamic modeling in the prior art, and provide theoretical support for multi-UCAV cooperative combat in beyond-visual-range air combat. SUMMARY

[0007] The purpose of the present application is to provide an optimization method and system for cooperative occupation of two aircraft in a typical borrowed missile scene, which can achieve optimization of cooperative occupation of two aircraft in a typical borrowed missile scene and provide theoretical support for multi-UCAV cooperative combat in beyond-visual-range air combat.

[0008] To achieve the above-mentioned purpose, the present application provides the following solutions.

[0009] In a first aspect, the present application provides an optimization method for cooperative occupation of two aircraft in a typical borrowed missile scene, which includes the following steps.

[0010] A borrowed missile air combat scene is constructed, and the tasks of the two aircraft are defined; the two aircraft include a detection and guidance aircraft and a weapon launch aircraft, the task of the detection and guidance aircraft is to keep the enemy within the detectable range at all times, and the task of the weapon launch aircraft is to safely reach the designated position and state within the specified time to complete the missile launch.

[0011] construct a three-degree-of-freedom dynamics and kinematics model of the fighter, a weapon engagement zone and non-escapable zone model, and a beyond-visual-range air combat situation reward and punishment situation function based on the borrowed-ammunition battlefield scenario and the two-aircraft role mission; the three-degree-of-freedom dynamics and kinematics model of the fighter includes a particle kinematics equation and a dynamics equation, state variables of the particle kinematics equation and the dynamics equation include position, velocity component, and track angle parameters in an inertial coordinate system, control variables include tangential overload, normal overload, and roll angle; the weapon engagement zone and non-escapable zone model is a simulation model based on a weapon engagement zone and non-escapable zone; the beyond-visual-range air combat situation reward and punishment situation function includes an off-axis launch angle reward and a relative position reward based on a missile attack zone, and the weight of the off-axis launch angle reward and the relative position reward based on the missile attack zone is balanced in the form of an adjustment coefficient.

[0012] based on the three-degree-of-freedom dynamics and kinematics model of the fighter, the weapon engagement zone and non-escapable zone model, and the beyond-visual-range air combat situation reward and punishment situation function, enemy multi-step track prediction is performed using an LSTM model to determine enemy state information at each time step; the enemy state information includes position information and velocity information of the enemy.

[0013] According to the enemy state information at each time step, an improved particle swarm algorithm is used to obtain an optimal occupation route.

[0014] According to the optimal occupation route, an improved track planning algorithm is used to control the two aircraft to reach a specified position.

[0015] Optionally, the expression of the three-degree-of-freedom dynamics and kinematics model of the fighter is as follows.

[0016] ; wherein, represents the position of the aircraft in the inertial coordinate system, wherein the aircraft includes a detection and guidance aircraft and a weapon launching aircraft; represents the components of the velocity of the aircraft in the three coordinate axes of the inertial coordinate system; respectively represent the track inclination angle, the track deflection angle, and the roll angle; and respectively represent the tangential overload along the velocity direction and the normal overload perpendicular to the velocity direction; represents the velocity of the aircraft; represents the gravitational acceleration; respectively represent the first-order derivatives of the velocity of the aircraft, the track inclination angle, and the track deflection angle with respect to time.

[0017] Optionally, the constraint condition of the control variable is as follows.

[0018] ; wherein, , respectively represent the minimum values of the tangential overload and the normal overload; , respectively represent the maximum values of the tangential overload and the normal overload; , respectively represent the maximum value and the minimum value of the roll angle.

[0019] Optionally, the expression of the over-the-horizon air combat situation reward-punishment situation function is as follows.

[0020] ; ; ; wherein, represents the over-the-horizon air combat situation reward-punishment situation function, is an off-axis launch angle, is a maximum off-axis launch angle, represents a reward of the off-axis launch angle, represents a relative position reward based on a missile attack area, , are respectively , adjustment coefficients of represents a distance between a weapon launcher and a weapon engagement area, represents an unescapable distance, represents an attackable distance.

[0021] Optionally, in the weapon engagement area and unescapable area model, the weapon engagement area is a spatial area centered on the enemy, the weapon engagement area satisfies a rule that if the enemy moves at a uniform speed in a straight line after the attack side launches a missile, the missile can hit, the boundary of the weapon engagement area is calculated by using a pattern search algorithm, and the range of the weapon engagement area dynamically changes with the azimuth angle and the entry angle of the enemy.

[0022] The unescapable area is a sub-area of the weapon engagement area, the unescapable area satisfies a rule that no matter how the enemy maneuvers to evade, the missile can destroy after the attack side launches the missile, and the unescapable area is a region that the enemy cannot escape from the missile damage envelope due to energy constraints or time constraints.

[0023] Optionally, according to the state information of the enemy at each time step, an improved particle swarm algorithm is used to obtain an optimal occupation route, and the method specifically includes the following steps.

[0024] According to the enemy state information at each time step, a multi-constraint optimization problem is constructed; constraint conditions of the multi-constraint optimization problem include: the enemy is in the front hemisphere region of the two aircrafts of the own side at the moment, the own side detection guidance aircraft observes the enemy at the moment, the weapon launch aircraft of the own side does not long-term stay in the weapon combat area of the enemy and does not enter the non-escapable area of the enemy; a target function of the multi-constraint optimization problem is to maximize the sum of the difference between the reward and the penalty at each moment.

[0025] Based on the multi-constraint optimization problem, an improved particle swarm algorithm is used to discretize the continuous problem in the form of space grid, define the particle as a binary group of position matrix and orientation matrix, and construct a three-subgroup cooperative mechanism to iteratively update the individual optimal solution and the group optimal solution, to obtain the optimal occupation route.

[0026] Optionally, the expression of the multi-constraint optimization problem is as follows.

[0027] ; Wherein, represents the sum of the difference between the reward and the penalty at each moment, represents the maximum time length to be solved, represents that the enemy is in the front hemisphere region of the two aircrafts of the own side at the moment, represents that the own side detection guidance aircraft observes the enemy at the moment, represents that the weapon launch aircraft of the own side does not long-term stay in the weapon combat area of the enemy and does not enter the non-escapable area of the enemy.

[0028] Optionally, according to the optimal occupation route, an improved path planning algorithm is used to control the two aircrafts to reach the specified position, which specifically includes the following steps.

[0029] Based on the optimal occupation route, a relative position description matrix is constructed, a mapping function is used to discretize the continuous relative distance of the two aircrafts into a direction label, and a current formation matrix and a target formation matrix are generated.

[0030] Based on the current formation matrix and the target formation matrix, a similarity matrix is constructed.

[0031] Based on the similarity matrix, a position allocation optimization model is constructed, and the position allocation optimization model is solved to obtain an optimal warplane-target position mapping relationship; the optimal warplane-target position mapping relationship is a warplane-target position mapping relationship with the maximum matching degree.

[0032] According to the optimal warplane-target position mapping relationship, the tangential overload, the normal overload and the roll angle output by the three-degree-of-freedom dynamics and kinematics model of the warplane are combined as the control quantity to control the two aircrafts to reach the specified position, and time synchronization and collision avoidance of the two aircrafts are realized.

[0033] Optionally, the expression of the mapping function is as follows.

[0034] ; wherein, denotes the mapping function, denotes the front-back distance or left-right distance between the two aircrafts.

[0035] The expression of the position allocation optimization model is as follows.

[0036] ; ; wherein, and denote the target position and the target position , denote the similarity between the target position and the target position , is a selection function, denotes that a position is allocated to each detection and guidance aircraft or weapon launcher, denotes that a detection and guidance aircraft or weapon launcher is allocated to each position.

[0037] In a second aspect, the application provides an optimization system for our two aircrafts to cooperatively occupy positions in a typical borrowed missile scenario. The optimization system for our two aircrafts to cooperatively occupy positions in a typical borrowed missile scenario applies the optimization method for our two aircrafts to cooperatively occupy positions in a typical borrowed missile scenario. The optimization system for our two aircrafts to cooperatively occupy positions in a typical borrowed missile scenario comprises the following functional modules.

[0038] A scenario construction and role task definition module is configured to construct a borrowed missile air combat scenario and define tasks of two aircrafts. The two aircrafts include a detection and guidance aircraft and a weapon launcher. The task of the detection and guidance aircraft is to keep the enemy within a detectable range at all times. The task of the weapon launcher is to safely reach a designated position and state within a specified time to complete missile launch.

[0039] A model and function construction module is configured to construct a fighter three-degree-of-freedom dynamics and kinematics model, a weapon engagement zone and non-escapable zone model, and an over-the-horizon air combat situation reward and punishment situation function based on the dogfighting scene and the dual-aircraft role task, wherein the fighter three-degree-of-freedom dynamics and kinematics model includes a particle kinematics equation and a dynamics equation, state variables of the particle kinematics equation and the dynamics equation include position, velocity component, and track angle parameters in an inertial coordinate system, control variables include tangential overload, normal overload, and roll angle, the weapon engagement zone and non-escapable zone model is a simulation model based on a weapon engagement zone and a non-escapable zone, and the over-the-horizon air combat situation reward and punishment situation function includes an off-axis launch angle reward and a relative position reward based on a missile attack zone, and the weight of the off-axis launch angle reward and the relative position reward based on the missile attack zone is balanced in the form of an adjustment coefficient.

[0040] An enemy state prediction module is configured to perform multi-step track prediction of an enemy by using an LSTM model based on the fighter three-degree-of-freedom dynamics and kinematics model, the weapon engagement zone and non-escapable zone model, and the over-the-horizon air combat situation reward and punishment situation function, to determine enemy state information at each time step, and the enemy state information includes position information and velocity information of the enemy.

[0041] An optimal occupation route solving module is configured to solve an optimal occupation route by using an improved particle swarm algorithm according to the enemy state information at each time step.

[0042] A track planning module is configured to control the dual aircraft to reach a specified position by using an improved track planning algorithm according to the optimal occupation route.

[0043] According to the embodiments provided in the application, the following technical effects are achieved: The application provides a method and system for optimizing double-aircraft cooperative occupation in a typical borrowed missile scene. BRIEF DESCRIPTION OF DRAWINGS

[0044] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the drawings needed in the embodiments will be briefly introduced as follows. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without any creative effort on the basis of these drawings.

[0045] Figure 1 A flowchart of a method for optimizing double-aircraft cooperative occupation in a typical borrowed missile scene according to an embodiment of the present application.

[0046] Figure 2 A principle diagram of a method for optimizing double-aircraft cooperative occupation in a typical borrowed missile scene according to an embodiment of the present application.

[0047] Figure 3 An application scene diagram of a method for optimizing double-aircraft cooperative occupation in a typical borrowed missile scene according to an embodiment of the present application.

[0048] Figure 4 An LSTM network structure diagram according to an embodiment of the present application.

[0049] Figure 5 A schematic diagram of an LSTM trajectory prediction process is provided for an embodiment of the present application.

[0050] Figure 6 A schematic diagram of a particle swarm algorithm process is provided for an embodiment of the present application.

[0051] Figure 7 A schematic diagram of a path planning algorithm process is provided for an embodiment of the present application.

[0052] Figure 8 A schematic diagram of a structure of an optimization system for our double-machine cooperative occupation in a typical borrowed bomb scenario is provided for an embodiment of the present application. DETAILED DESCRIPTION

[0053] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only some of the embodiments of the present application, but not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work fall within the scope of protection of the present application.

[0054] In order to make the above-mentioned purposes, features and advantages of the present application more obvious and easy to understand, the present application will be further described in detail below with reference to the drawings and specific embodiments.

[0055] As shown in Figure 1 A typical borrowed bomb scenario optimization method for our double-machine cooperative occupation is mainly applied to cooperative occupation optimization and collision avoidance between at least two fighters (for example, a detection and guidance machine and a weapon launching machine) in a typical borrowed bomb scenario. The typical borrowed bomb scenario optimization method for our double-machine cooperative occupation specifically includes the following steps.

[0056] S1: A borrowed bomb air combat scenario is constructed, and double-machine role tasks are defined.

[0057] In this embodiment, the double-machine includes a detection and guidance machine and a weapon launching machine, the task of the detection and guidance machine is to keep the enemy within a detectable range at all times, and the task of the weapon launching machine is to safely reach a designated position and state within a specified time to complete missile launching.

[0058] S2: Based on the borrowed bomb air combat scenario and the double-machine role tasks, a three-degree-of-freedom dynamics and kinematics model of the fighter, a weapon engagement zone and non-escapable zone model, and a beyond-visual-range air combat situation reward and punishment situation function are constructed respectively.

[0059] In the embodiment, the three-degree-of-freedom dynamics and kinematics model of the fighter aircraft includes a particle kinematics equation and a dynamics equation, state variables of the particle kinematics equation and the dynamics equation include position, velocity component and track angle parameters in an inertial coordinate system, control variables include tangential overload, normal overload and roll angle; the weapon engagement zone and non-escapable zone model is a simulation model based on a weapon engagement zone and a non-escapable zone; the beyond-visual-range air combat situation reward and punishment situation function includes an off-axis launch angle reward and a relative position reward based on a missile attack zone, and the weight of the off-axis launch angle reward and the relative position reward based on the missile attack zone is balanced in the form of an adjustment coefficient.

[0060] S3: Based on the three-degree-of-freedom dynamics and kinematics model of the fighter aircraft, the weapon engagement zone and non-escapable zone model and the beyond-visual-range air combat situation reward and punishment situation function, an LSTM model is used to perform multi-step track prediction of the enemy, to determine enemy state information at each time step; the enemy state information includes position information and velocity information of the enemy.

[0061] S4: According to the enemy state information at each time step, an improved particle swarm algorithm is used to solve an optimal occupation route.

[0062] S5: According to the optimal occupation route, an improved track planning algorithm is used to control the two aircrafts to reach a specified position.

[0063] In the weapon engagement zone and non-escapable zone model constructed in step S2, the weapon engagement zone is a space region centered on the enemy, the weapon engagement zone satisfies a rule that if the enemy moves at a constant speed in a straight line after the attacking party launches a missile, the missile can hit, the boundary of the weapon engagement zone is calculated using a pattern search algorithm, and the range of the weapon engagement zone dynamically changes with the azimuth angle and the entry angle of the enemy. The non-escapable zone is a sub-region of the weapon engagement zone, the non-escapable zone satisfies a rule that no matter how the enemy maneuvers to evade, the attacking party can destroy after launching a missile, and the non-escapable zone is a region that the enemy cannot escape from the missile kill envelope due to energy constraints or time constraints.

[0064] In the embodiment, step S4 uses an improved particle swarm algorithm to solve an optimal occupation route according to the enemy state information at each time step, and specifically includes the following steps.

[0065] S41: According to the enemy state information at each time step, a multi-constraint optimization problem is constructed; constraint conditions of the multi-constraint optimization problem include that the enemy is at a time in a front hemisphere region of the two aircrafts, the detection and guidance aircraft of the attacking party observes the enemy at a time, the weapon launch aircraft of the attacking party does not stay in the weapon engagement zone of the enemy for a long time and does not enter the non-escapable zone of the enemy; an objective function of the multi-constraint optimization problem is to maximize the sum of the difference between rewards and punishments at each time.

[0066] S42: Based on the multi-constraint optimization problem, the improved particle swarm algorithm is used to discretize the continuous problem in the form of space grid, the particle is defined as a binary group of position matrix and orientation matrix, and the three-subgroup cooperative mechanism is constructed to iteratively update the individual optimal solution and the group optimal solution, and the optimal occupation route is obtained.

[0067] In this embodiment, step S5 controls the double aircraft to reach the specified position according to the optimal occupation route by using the improved path planning algorithm, specifically including the following steps.

[0068] S51: Based on the optimal occupation route, a relative position description matrix is constructed, the mapping function is used to discretize the continuous relative distance of the double aircraft into a direction label, and the current formation matrix and the target formation matrix are generated.

[0069] S52: Based on the current formation matrix and the target formation matrix, a similarity matrix is constructed.

[0070] S53: Based on the similarity matrix, a position allocation optimization model is constructed, and the position allocation optimization model is solved to obtain an optimal aircraft-target position mapping relationship; the optimal aircraft-target position mapping relationship is the aircraft-target position mapping relationship with the maximum matching degree.

[0071] S54: According to the optimal aircraft-target position mapping relationship, the tangential overload, the normal overload and the roll angle output by the three-degree-of-freedom dynamics and kinematics model of the aircraft are combined as the control quantity to control the double aircraft to reach the specified position, and the time cooperation and collision avoidance of the double aircraft are realized.

[0072] In order to make the technical scheme of the present application more clear, the specific implementation process of the technical scheme of the present application will be described in the form of examples as follows.

[0073] Figure 2The principle of the optimization method of the cooperative occupation of the double aircraft of the application in a typical borrowed bomb scene is shown. It is assumed that the numbers of the double aircraft of the application are UCAV1 (a detection guidance aircraft) and UCAV2 (a weapon launching aircraft). First, the target position and heading are determined through the cooperative occupation algorithm. Then, the current position (x1, y1) of UCAV1 and the current position (x2, y2) of UCAV2 are obtained through the formation switching algorithm, respectively. Then, the control amount is calculated, respectively. Then, target position 1 and target position 2 are allocated to UCAV1 and UCAV2, respectively. Then, UCAV1 and UCAV2 are controlled to reach the specified positions in a real-time control manner, so as to realize the time cooperation and collision avoidance of the double aircraft. In the case of determining the tactics, the autonomous decision-making ability of the air combat of the application is significantly improved. A complete set of tactics can be completed according to the needs, while the threat of the enemy can be maximally avoided. The striking and survival ability of the application in the tactical cooperation in the autonomous combat is significantly improved. The method specifically includes the following implementation steps.

[0074] S1: constructing a borrowed bomb air combat scene and defining the tasks of the double aircraft.

[0075] In this embodiment, the borrowed bomb air combat scene constructed in step S1 is a typical borrowed bomb attack scene, which involves a fighter aircraft. One of the fighter aircrafts is a detection guidance aircraft, and the target of the detection guidance aircraft is to make the enemy always within the detectable range of the application. The other fighter aircraft is a weapon launching aircraft, and the target of the weapon launching aircraft is to safely reach the specified position and state for missile launching within the specified time.

[0076] Figure 3 The application scene of the optimization method of the cooperative occupation of the double aircraft of the application in a typical borrowed bomb scene is shown. The spatial position of the cooperative occupation of the double aircraft and the reaching of the specified position according to the occupation route are demonstrated.

[0077] S2: based on the borrowed bomb air combat scene and the tasks of the double aircraft, a three-degree-of-freedom dynamics and kinematics model of the fighter aircraft, a weapon engagement zone and non-escapable zone model, and a beyond-visual-range air combat situation reward and punishment situation function are constructed, respectively. The three-degree-of-freedom dynamics and kinematics model of the fighter aircraft includes a particle kinematics equation and a dynamics equation. The state variables of the particle kinematics equation and the dynamics equation include the position, velocity component and track angle parameter in the inertial coordinate system. The control amount includes the tangential overload, the normal overload and the roll angle. The weapon engagement zone and non-escapable zone model is a simulation model based on the weapon engagement zone and non-escapable zone. The beyond-visual-range air combat situation reward and punishment situation function includes a off-axis launch angle reward and a relative position reward based on a missile attack zone. The weights of the off-axis launch angle reward and the relative position reward based on the missile attack zone are balanced in the form of an adjustment coefficient.

[0078] In this embodiment, the step S2 is specifically implemented as follows: firstly, three-dimensional dynamics and kinematics modeling of the fighter is performed, that is, a three-dimensional dynamics and kinematics model of the fighter is established. When studying the air combat maneuver decision of the aircraft, in order to accurately describe the motion track and maneuver characteristics of the aircraft, three-dimensional particle kinematics modeling of the aircraft in three-dimensional space is performed.

[0079] In order to focus on the air combat maneuver decision, the model is established based on the following conditions in this embodiment.

[0080] (1) During the motion of the aircraft, the influence of air flow rate on the motion of the aircraft is not considered.

[0081] (2) The environmental gravity acceleration g is always kept unchanged, and the influence of environmental factors such as height and air density on acceleration is not considered.

[0082] (3) The influence of the size of the aircraft on the motion is ignored, and the aircraft is regarded as a movable particle.

[0083] (4) The aircraft has no sideslip phenomenon during the motion.

[0084] (5) During the motion of the aircraft, the ground reference frame is always kept stationary, and the earth rotation is not considered.

[0085] Based on the above conditions, the three-dimensional particle dynamics equation of the aircraft in three-dimensional space can be expressed as follows.

[0086] (1) wherein, represents the position of the aircraft in the inertial coordinate system; represents the velocity of the aircraft on the three coordinate axes; respectively represents the flight path angle, the flight path angle and the roll angle; and respectively represent the tangential overload along the velocity direction and the normal overload perpendicular to the velocity direction; respectively represent the first order derivative of the velocity of the aircraft, the flight path angle and the flight path angle with respect to time. In formula (1), the first three terms are the kinematics model of the aircraft particle, and the last three terms are the dynamics model of the aircraft; the state variables include and ; the control variables include and .

[0087] In this embodiment, it can be known from the three-dimensional particle dynamics equation of the aircraft in three-dimensional space that the maneuver action of the aircraft can be represented by and Three control variables are used to control the aircraft maneuver. To ensure the effectiveness of the research, the maneuver decision made by reinforcement learning should be consistent with the characteristics of the real maneuver model. Therefore, the internal structure strength of the aircraft, the normal working conditions of various sensors, and other factors need to be considered in the design. The control variables need to be constrained, as shown in the following formula.

[0088] (2) The maximum and minimum values of the roll angle are represented by and respectively. Combined with the physical limit that the aircraft can withstand (i.e., the maximum overload during the design of the aircraft), the embodiment sets .

[0089] In this embodiment, the weapon engagement zone and the non-escapable zone model of the fighter aircraft need to be constructed, as follows.

[0090] In the beyond-visual-range air combat decision model, the weapon engagement zone refers to the attackable zone, which is defined as a spatial region centered on the target. When the attacker launches a missile from this region, if the target maintains a constant linear motion, the missile can effectively hit the target. The boundary of WEZ is accurately calculated by a pattern search algorithm, and its range changes dynamically with the target azimuth (i.e., the off-axis launch angle) and the entry angle : when and (head-on attack), the WEZ range is the largest, and when approaches the maximum off-axis launch angle , the WEZ range significantly shrinks. The non-escapable zone (NEZ) is a sub-region inside WEZ. When the attacker launches a missile from NEZ, regardless of the target's evasion maneuver, the missile can guarantee to destroy the target. The physical nature of NEZ is the core region of the target that cannot escape the missile's kill envelope due to energy or time constraints, and its range is significantly smaller than WEZ.

[0091] In this embodiment, the beyond-visual-range air combat situation reward-punishment situation function constructed in step S2 includes the off-axis launch angle reward and the relative position reward based on the missile attack zone, as shown in the following formula.

[0092] (3) (4)​​​​​​​​​​​ (5) in, The reward and punishment situation function represents the situational awareness of beyond-visual-range air combat. Off-axis launch angle, For the maximum off-axis launch angle, This indicates the bonus for the off-axis launch angle. This indicates a reward based on the relative position of the missile's attack zone. , They are respectively , The adjustment coefficients for both depend on whether the reward is more biased towards off-axis launch angle or distance. Indicates the distance between the weapon launcher and the weapon engagement zone. Indicates the distance from which escape is impossible. Indicates the attack range.

[0093] S3: Based on the three-degree-of-freedom dynamics and kinematics model of the fighter jet, the weapon engagement zone and no-escape zone model, and the beyond-visual-range air combat situation reward and punishment function, the LSTM model is used to predict the enemy's multi-step trajectory and determine the enemy's state information at each time step; the enemy's state information includes the enemy's position information and velocity information.

[0094] In this embodiment, step S3 is implemented as follows.

[0095] Assuming the plane is currently in At any time, and Historical position before the moment Speed ​​information and control quantity Training is performed using LSTM. The target position state at different prediction times is obtained by using position and velocity information. .

[0096] In this embodiment, the LSTM network structure is as follows: Figure 4 As shown, in At any given moment, a single layer of a network consists of two information flows, one above the other. From... arrive The information flow represents the transmission of cell states, and the entire line interacts linearly with the information flow below through three gating structures. The gating structures allow information to flow selectively, flowing upwards... arrive The information flow deletes or adds information about the cell state. In the gating structure... Activation function layer and The activation function layer can convert the input to (0, 1) and (-1, 1) respectively, and generate the weight of the input data, so as to filter the input data. There are 3 gate structures in each LSTM network to control the cell state.

[0097] (1) The forget gate.

[0098] The expression of the forget gate is as follows.

[0099] (6) Wherein, The layer obtains a value between 0 and 1 as the probability of forgetting the previous cell state by the last time hidden state and the input at the time The decay coefficient of memory is regarded as . is a two-dimensional matrix, which is a parameter that the LSTM model needs to learn. This matrix contains the knowledge learned by the model and is used to determine how to calculate which information should be forgotten according to the past hidden state and the current input. It defines the importance weight of the input information combination. is a vector, which is also a parameter that the LSTM model needs to learn, adding an offset (constant term) to the calculation of the forget gate. This allows the LSTM model to still output a non-zero value when the input weighted sum is zero (for example, by default, it tends to retain or forget some information), thereby increasing the flexibility of the model.

[0100] (2) The input gate.

[0101] Part of the input gate linearly combines and the hidden state at the last time and obtains through the layer activation, which determines which information needs to be updated, and these information is the part selected to be forgotten in the forget gate. Another part will and generate a vector through a layer, which is the alternative content for updating. Then the two parts are combined to update the state to . The formula for the input gate update process is as follows.

[0102] (7) (8) (9) wherein,​​ is the input gate vector, which controls the degree of preservation of new information (0~1); is the input gate weight matrix, which learns how to combine historical information and current input; is the input gate bias vector, which increases the flexibility of the model; is the candidate state weight matrix, which learns the combination of new information; is the candidate state bias vector, which increases the flexibility of the model; represents convolution.

[0103] (3) Output gate.

[0104] (10) (11) wherein, is the output gate vector, which controls the degree of output of the cell state; is the output gate weight matrix, which is the output weight; is the output gate bias vector; is the current hidden state, which is the final output of the LSTM model and also the input of the next step. In addition, and , , represent the weight and bias matrix of the output, respectively, which are also parameters that need to be learned in training. Since the information of the previous cycle is used in each cycle and , each output state is affected by the previous state, so the LSTM model has the ability to remember long-term historical information.

[0105] Figure 5 The flow of LSTM trajectory prediction is shown. First, the dataset is obtained, then the obtained dataset is divided and normalized, then it is judged whether the network structure parameters (initial parameters) are fitted, if so, it is further judged whether the precision is met, if so, the network model is further saved, then the test set is input, and the output result is predicted. If not fitted, adjust and optimize the network structure parameters, continue to judge whether it is fitted, until it is fitted; if the precision is not met, adjust and optimize the network structure parameters, continue to judge whether it is fitted and whether the precision is met, until it is fitted and the precision is met.

[0106] S4: According to the enemy state information at each time step, an improved particle swarm algorithm is used to obtain the optimal occupation route.

[0107] In this embodiment, the specific implementation method of step S4 is as follows.

[0108] The target point is calculated by improved particle swarm optimization (PSO). Particle swarm optimization is a swarm intelligence-based optimization algorithm inspired by the foraging behavior of bird flocks, each particle represents a candidate solution in the solution space, and the particle dynamically adjusts the flight direction and speed according to the individual historical optimal solution (pbest) and the group historical optimal solution (gbest), Figure 6 The basic framework flowchart of the particle swarm optimization algorithm is shown, which first initializes the particle swarm and calculates the particle fitness, then judges whether the termination condition is met, if yes, outputs the optimal solution, otherwise updates the individual optimal solution and the group optimal solution, then updates the particle speed and position, and calculates the particle fitness again according to the updated particle speed and position, so as to achieve the purpose of dynamically adjusting the flight direction and speed.

[0109] The core steps of the improved particle swarm optimization algorithm for the typical borrowed field scenario of dual-machine cooperative occupation are as follows.

[0110] First, the occupation space is divided into a grid, which facilitates the conversion of continuous position space into a discrete combination problem that is easy to solve, i.e., a multi-constrained optimization problem, so that both discrete variables (occupation grid positions) and continuous variables (UCAV orientation angles) can be optimized simultaneously while satisfying the constraint conditions of avoiding missile attack areas. The constraint conditions of this multi-constrained optimization problem include that the enemy is in the front hemisphere of the dual-machine at the moment, the detection and guidance machine observes the enemy at the moment, and the weapon launching machine does not stay in the weapon combat area of the enemy for a long time and does not enter the non-escapable area of the enemy. The objective function of the multi-constrained optimization problem is to maximize the sum of the difference between the reward and the penalty at each moment.

[0111] The position matrix and the orientation matrix are defined as follows.

[0112] (12) (13) wherein, is the i-th element in the orientation matrix , and the superscript represents transposition.

[0113] A three-subgroup cooperative mechanism is adopted, specifically, in the improved particle swarm optimization algorithm, the three-subgroup cooperative mechanism realizes global optimization through functional complementation, the exploration subgroup completely randomly generates new particles (position matrix with random 1, heading in Uniform sampling (accounting for 30% of the total number of particles) ensures the diversity of the solution space and avoids premature convergence; the empirical inheritance subgroup is based on the previous generation of particles and subjected to bounded perturbation, as shown below.

[0114] (14) in, The grid coordinates (row and column indices) represent the current position of the particle, and the values ​​are based on the positions of non-zero elements in the position matrix of the previous generation of particles; This represents the new grid coordinates after the perturbation, and the values ​​are based on the new position matrix. index; This is a function to generate random integers within a range; This indicates the maximum positional perturbation step size (in grid units).

[0115] In this embodiment, the heading is... Adjust and truncate to It only accepts updates that improve fitness to preserve historical experience. Among these, This represents the original heading angle, and its value is based on the heading matrix of the previous generation of particles. The value in; Indicates the new heading angle after the disturbance; This represents the maximum heading disturbance amplitude (in degrees). The elite development subgroup focuses on the global optimum and samples locally within its grid neighborhood (neighborhood radius). ,in, The neighborhood radius, (for grid width), and finely adjust the heading ( The range is completely reset in each generation to create new particles in the neighborhood for in-depth development. The three subgroups unify the synchronization direction, and the low-fitness particles in the experience inheritance subgroup are replaced by high-fitness random particles in the exploration subgroup to prevent stagnation. The three subgroups work together in a 1:1:1 resource ratio to break through the local optimum of the MINLP (Mixed Integer Nonlinear Programming) problem.

[0116] S5: Based on the optimal positioning route, an improved trajectory planning algorithm is used to control the two aircraft to reach the designated position.

[0117] In this embodiment, step S5 is implemented as follows.

[0118] Formation switching is the first step in the trajectory planning algorithm. In this algorithm, the relative position description matrix is ​​the core tool; its essence is to quantify the formation structure by discretizing directional relationships, effectively reducing the computational complexity of trajectory planning. First, a mapping function is defined. As shown in the following formula.

[0119] (15) where, represents the mapping function, represents the front-back distance or left-right distance between two aircrafts, including the front-back distance and the left-right distance . The output 1 of the mapping function represents the positive direction, 0 represents the coincidence, and -1 represents the negative direction. The relative positions between aircrafts in the continuous space (such as represents the front-back distance, represents the left-right distance) are compressed into discrete direction labels. Based on this, the current formation matrix is constructed, and the current formation matrix is a matrix, where each non-diagonal element is composed of a two-tuple , which completely records the direction relationship of the jth aircraft relative to the ith aircraft, for example represents is on the right front of , and the diagonal element is always ; the target formation matrix is constructed synchronously, which has the same structure as but is calculated based on the ideal geometric relationship of the target position, forming a mathematical abstract expression of the formation structure.

[0120] Figure 7 The flow of the path planning algorithm is shown, which first calculates the current formation matrix and the target formation matrix according to the input data, then constructs the similarity matrix and solves the position assignment problem, and finally applies the consistency control protocol to make the UCAV reach the target position synchronously. The position assignment problem is converted into an optimization task through the similarity matrix . When calculating the similarity between , the first row of the current formation matrix (describing the direction relationship between the detection guidance aircraft and the weapon launcher) and the first row of the target formation matrix (required direction relationship of the target position ) are extracted, and the number of two-tuples that are completely matched is counted as . From this, the position assignment optimization model is constructed, which is represented as follows.

[0121] (16) (17) where, and ​​​​​respectively represent target positions and target positions , represent the similarity between target positions and target positions , is a selection function taking 0 or 1 from (0, 1), represents assigning a position to each detection guidance machine or weapon launcher, represents assigning a detection guidance machine or weapon launcher to each position.

[0122] The essence is to seek an optimal warplane-target position mapping relationship, and the discretization processing significantly reduces the calculation complexity.

[0123] (18) wherein, represents an output control quantity, and are constants, represents a target heading requirement, represents a target position requirement.

[0124] After completing the position assignment, the heading requirement and the target position requirement that have been constrained in the foregoing can be used to output the control quantity by setting constants and , and the three-degree-of-freedom dynamics and kinematics model of the warplane in the foregoing can be combined to control the two warplanes of the own side to reach the specified position, thereby realizing the cooperative position occupation and collision avoidance of the two warplanes of the own side in the typical borrowed bomb scenario.

[0125] Based on the same inventive concept, the embodiment of the present application further provides an optimization system for implementing the optimization method for cooperative position occupation of the two warplanes of the own side in the typical borrowed bomb scenario. The implementation scheme for solving the problem provided by the system is similar to the implementation scheme recorded in the foregoing method, and therefore the specific limitations in the following optimization system embodiment for cooperative position occupation of the two warplanes of the own side in the typical borrowed bomb scenario can be referred to the limitations of the optimization method for cooperative position occupation of the two warplanes of the own side in the typical borrowed bomb scenario in the foregoing, which will not be repeated here.

[0126] In an exemplary embodiment, as shown in Figure 8 , an optimization system for cooperative position occupation of the two warplanes of the own side in the typical borrowed bomb scenario is provided, comprising the following functional modules.

[0127] A scene construction and role task definition module is configured to construct a borrowed missile air combat scene and define a dual-aircraft role task, the dual-aircraft including a detection and guidance aircraft and a weapon launching aircraft, the detection and guidance aircraft being tasked to keep the enemy aircraft within a detectable range, and the weapon launching aircraft being tasked to safely reach a specified position and state within a specified time to complete missile launching.

[0128] A model and function construction module is configured to construct a three-degree-of-freedom dynamic and kinematic model of a combat aircraft, a weapon engagement zone and non-escapable zone model, and an over-the-horizon air combat situation reward and penalty situation function based on the borrowed missile air combat scene and the dual-aircraft role task, wherein the three-degree-of-freedom dynamic and kinematic model of the combat aircraft includes a particle kinematic equation and a dynamic equation, state variables of the particle kinematic equation and the dynamic equation including position, velocity component and track angle parameters in an inertial coordinate system, control variables including tangential overload, normal overload and roll angle; the weapon engagement zone and non-escapable zone model is a simulation model based on a weapon engagement zone and non-escapable zone; and the over-the-horizon air combat situation reward and penalty situation function includes an off-axis launch angle reward and a relative position reward based on a missile attack zone, and the weight of the off-axis launch angle reward and the relative position reward based on the missile attack zone is balanced in the form of an adjustment coefficient.

[0129] An enemy state prediction module is configured to perform multi-step track prediction of an enemy aircraft using an LSTM model based on the three-degree-of-freedom dynamic and kinematic model of the combat aircraft, the weapon engagement zone and non-escapable zone model, and the over-the-horizon air combat situation reward and penalty situation function, to determine enemy state information at each time step, the enemy state information including position information and velocity information of the enemy aircraft.

[0130] An optimal occupation route solving module is configured to solve an optimal occupation route using an improved particle swarm algorithm based on the enemy state information at each time step.

[0131] A track planning module is configured to control the dual-aircraft to reach a specified position using an improved track planning algorithm based on the optimal occupation route.

[0132] The technical features of the above embodiments can be combined in any manner. To make the description concise, not all possible combinations of the technical features in the above embodiments are described, but as long as the combinations of the technical features do not contradict, they should be considered as within the scope of the present disclosure.

[0133] The principles and implementation manners of the present application are described by applying specific examples in the embodiments above, and the above embodiment descriptions are only used for helping to understand the method of the present application and its core idea; meanwhile, for the general technical personnel in the art, according to the idea of the present application, the specific implementation manners and application ranges will have changes. In conclusion, the content of the present description should not be understood as the limitation of the present application.

Claims

1. An optimized method for coordinated positioning of two friendly aircraft in a typical bomb-borrowing scenario, characterized in that, The optimization method for our dual-aircraft coordinated positioning in the typical bomb-borrowing scenario includes: Construct a missile-borrowing air combat scenario and define the dual-aircraft roles and tasks; the dual-aircraft include a detection and guidance aircraft and a weapon launcher. The task of the detection and guidance aircraft is to keep the enemy within a detectable range at all times, and the task of the weapon launcher is to safely reach the designated position and status within a specified time to complete the missile launch. Based on the aforementioned missile-assisted air combat scenario and the dual-aircraft role mission, a three-degree-of-freedom dynamics and kinematics model of the fighter jet, a weapon engagement zone and no-escape zone model, and a beyond-visual-range (BVR) air combat situation reward and punishment function are constructed respectively. The three-degree-of-freedom dynamics and kinematics model of the fighter jet includes particle kinematic equations and dynamic equations. The state variables of the particle kinematic equations and dynamic equations include position, velocity components, and track angle parameters in the inertial coordinate system. The control variables include tangential overload, normal overload, and roll angle. The weapon engagement zone and no-escape zone model is a simulation model based on the weapon engagement zone and no-escape zone. The BVR air combat situation reward and punishment function includes an off-axis launch angle reward and a relative position reward based on the missile attack zone, and the weights of the off-axis launch angle reward and the relative position reward based on the missile attack zone are balanced by adjustment coefficients. Based on the aforementioned three-degree-of-freedom dynamics and kinematics model of the fighter jet, the aforementioned weapon engagement zone and no-escape zone model, and the aforementioned beyond-visual-range air combat situation reward and punishment function, an LSTM model is used to predict the enemy's multi-step trajectory and determine the enemy's state information at each time step; the enemy's state information includes the enemy's position information and velocity information. Based on the enemy's state information at each time step, the optimal occupancy route is obtained by using an improved particle swarm optimization algorithm. Based on the optimal positioning route, an improved trajectory planning algorithm is used to control the two aircraft to reach the designated position.

2. The optimized method for our dual-aircraft coordinated positioning in a typical bomb-borrowing scenario as described in claim 1, characterized in that, The expression for the three-degree-of-freedom dynamics and kinematics model of the fighter jet is as follows: ； in, This indicates the position of the aircraft in the inertial coordinate system, where the aircraft includes detection and guidance systems and weapon launchers; This represents the components of the aircraft's velocity along the three coordinate axes of the inertial coordinate system; These represent the track inclination angle, track deviation angle, and roll angle, respectively. and These represent tangential overload along the velocity direction and normal overload perpendicular to the velocity direction, respectively. Indicates the speed of the aircraft; Represents gravitational acceleration; These represent the first derivatives of the aircraft's speed, track inclination angle, and track deviation angle with respect to time, respectively.

3. The optimized method for our dual-aircraft coordinated positioning in a typical bomb-borrowing scenario as described in claim 2, characterized in that, The constraint condition for the control quantity is: ； in, , These represent the minimum values ​​of tangential overload and normal overload, respectively. , These represent the maximum values ​​of tangential overload and normal overload, respectively. , These represent the maximum and minimum roll angles, respectively.

4. The optimized method for our dual-aircraft coordinated positioning in a typical bomb-borrowing scenario as described in claim 1, characterized in that, The expression for the beyond-visual-range air combat situation reward and punishment function is as follows: ； ； ； in, The reward and punishment situation function represents the situational awareness of beyond-visual-range air combat. Off-axis launch angle, For the maximum off-axis launch angle, This indicates the bonus for the off-axis launch angle. This indicates a reward based on the relative position of the missile's attack zone. , They are respectively , Adjustment coefficient, Indicates the distance between the weapon launcher and the weapon engagement zone. Indicates the distance from which one cannot escape. Indicates the attack range.

5. The optimized method for our dual-aircraft coordinated positioning in a typical bomb-borrowing scenario as described in claim 1, characterized in that, In the weapon engagement zone and no escape zone model, the weapon engagement zone is a spatial region centered on the enemy. The weapon engagement zone satisfies the rule that if the enemy moves in a straight line at a constant speed after the attacker launches a missile, the missile can hit the target. The boundary of the weapon engagement zone is calculated using a pattern search algorithm, and the range of the weapon engagement zone changes dynamically with the enemy's azimuth and entry angle. The no-escape zone is a sub-region of the weapon engagement zone. The no-escape zone satisfies the rule that the attacking side can destroy the missile regardless of the enemy's maneuvering evasion after launching it. The no-escape zone is the area in which the enemy cannot escape the missile's kill envelope due to energy constraints or time constraints.

6. The optimized method for our dual-aircraft coordinated positioning in a typical bomb-borrowing scenario as described in claim 1, characterized in that, Based on the enemy state information at each time step, an improved particle swarm optimization algorithm is used to solve for the optimal positioning route, specifically including: Based on the enemy's state information at each time step, a multi-constraint optimization problem is constructed. The constraints of the multi-constraint optimization problem include: the enemy is always in the front hemisphere region of our dual aircraft; our detection and guidance aircraft are always observing the enemy; our weapon launchers do not remain in the enemy's weapon engagement zone for extended periods and do not enter the enemy's no-escape zone. The objective function of the multi-constraint optimization problem is to maximize the sum of the differences between the reward and the penalty at each time step. Based on the aforementioned multi-constraint optimization problem, an improved particle swarm optimization algorithm is adopted to discretize the continuous problem by using a space grid. Particles are defined as pairs of position and orientation matrices, and a three-subgroup cooperative mechanism is constructed to iteratively update the individual optimal solution and the group optimal solution to obtain the optimal placement route.

7. The optimized method for our dual-aircraft coordinated positioning in a typical bomb-borrowing scenario as described in claim 6, characterized in that, The expression for the multi-constraint optimization problem is: ； in, This means maximizing the sum of the differences between the reward and penalty at each moment. This indicates the maximum time required to solve the problem. This indicates that the enemy is always in the front hemisphere area of ​​our two aircraft. This indicates that our detection and guidance systems are constantly observing the enemy. This indicates that our weapon launchers will not remain in the enemy's weapon engagement zone for extended periods and will not enter the enemy's no-escape zone.

8. The optimized method for our dual-aircraft coordinated positioning in a typical bomb-borrowing scenario as described in claim 1, characterized in that, Based on the optimal positioning route, an improved trajectory planning algorithm is used to control the two aircraft to reach the designated position, specifically including: Based on the optimal positioning route, a relative position description matrix is ​​constructed, and a mapping function is used to discretize the continuous relative distance between the two machines into direction labels to generate the current formation matrix and the target formation matrix. Construct a similarity matrix based on the current formation matrix and the target formation matrix; A position allocation optimization model is constructed based on the similarity matrix, and the optimal fighter-target position mapping relationship is obtained by solving the optimization model; the optimal fighter-target position mapping relationship is the fighter-target position mapping relationship with the highest matching degree. Based on the optimal fighter-target position mapping relationship, and combined with the tangential overload, normal overload, and roll angle output by the three-degree-of-freedom dynamics and kinematics model of the fighter jet as the control variables, the two aircraft are controlled to reach the designated position, thereby achieving time coordination and collision avoidance between the two aircraft.

9. The optimized method for our dual-aircraft coordinated positioning in a typical bomb-borrowing scenario as described in claim 8, characterized in that, The expression for the mapping function is: ； in, Represents a mapping function. Indicates the front-to-back or left-to-right distance between the two machines; The expression for the location allocation optimization model is: ； ； in, and Representing the target location respectively and target location , Indicates the target location and target location Similarity between them For selection function, This indicates that a location has been assigned to each detection guidance device or weapon launcher. This means that a detection guidance device or weapon launcher is assigned to each location.

10. An optimized system for coordinated positioning of two friendly aircraft in a typical bomb-borrowing scenario, characterized in that: The optimized system for our dual-aircraft coordinated positioning in the typical ammunition borrowing scenario applies the optimized method for our dual-aircraft coordinated positioning in the typical ammunition borrowing scenario as described in any one of claims 1-9. The optimized system for our dual-aircraft coordinated positioning in the typical ammunition borrowing scenario includes: The scenario construction and role task definition module is used to construct the air combat scenario of borrowing missiles and define the dual-aircraft role tasks; the dual-aircraft includes a detection and guidance aircraft and a weapon launcher. The task of the detection and guidance aircraft is to keep the enemy within the detection range at all times, and the task of the weapon launcher is to safely reach the designated position and status within a specified time to complete the missile launch. The model and function construction module is used to construct, based on the missile-borrowing air combat scenario and the dual-aircraft role mission, a three-degree-of-freedom dynamics and kinematics model of the fighter jet, a weapon engagement zone and no-escape zone model, and a beyond-visual-range air combat situation reward and punishment function. The three-degree-of-freedom dynamics and kinematics model of the fighter jet includes a particle kinematic equation and a dynamic equation. The state variables of the particle kinematic equation and the dynamic equation include position, velocity components, and track angle parameters in the inertial coordinate system. The control variables include tangential overload, normal overload, and roll angle. The weapon engagement zone and no-escape zone model is a simulation model based on the weapon engagement zone and no-escape zone. The beyond-visual-range air combat situation reward and punishment function includes an off-axis launch angle reward and a relative position reward based on the missile attack zone, and the weights of the off-axis launch angle reward and the relative position reward based on the missile attack zone are balanced by adjustment coefficients. The enemy state prediction module is used to predict the enemy's multi-step trajectory using an LSTM model based on the three-degree-of-freedom dynamics and kinematics model of the fighter jet, the weapon engagement zone and no escape zone model, and the beyond-visual-range air combat situation reward and punishment function, and to determine the enemy state information at each time step; the enemy state information includes the enemy's position information and velocity information. The optimal positioning route solution module is used to solve for the optimal positioning route based on the enemy's state information at each time step using an improved particle swarm optimization algorithm. The trajectory planning module is used to control the two aircraft to reach the designated position based on the optimal positioning route and an improved trajectory planning algorithm.

Images