Aircraft fleet dynamic recovery method, system, equipment and medium

Abstract

The invention discloses an aircraft fleet dynamic recovery method, system and device and a medium, and relates to the field of air traffic control, and the method comprises the steps: obtaining the state data of each aircraft in an aircraft fleet at a current decision-making stage; constructing a state space of the current decision-making stage based on the state data; inputting the state space into a strategy network to obtain the action probability of each airplane in the current decision-making stage; determining an action mask vector of each airplane in the current decision-making stage based on the residual fuel quantity and the numerical value of the recovery state; optimizing the action probability based on the action mask vector, and normalizing the optimized action probability to obtain the normalized action probability of each airplane in the current decision-making stage; and recovering the airplane with the highest action probability and the minimum residual fuel quantity after normalization. The real-time performance, accuracy and efficiency of aircraft recovery can be improved.

Description

Technical Field

[0001] The present application relates to the field of aviation traffic control, and in particular to a method, system, equipment and medium for dynamic recovery of a fleet of aircraft. Background Art

[0002] In modern aviation management, aircraft recovery sequencing is a key factor in determining recovery efficiency. In some cases, aircraft recovery sequencing relies primarily on predefined priority rules, such as fuel urgency or mission level, while others employ heuristic algorithms for static scheduling.

[0003] However, in the complex recovery environment with high real-time requirements in the open sea, the above methods have significant defects: (1) Insufficient dynamic response: unable to handle dynamic events such as changes in fuel consumption and sudden mechanical failures in real time; (2) Rigid constraint processing: when abnormal conditions such as aircraft go-around and mechanical failure occur, manual intervention is required to adjust the recovery queue; (3) The optimization target is relatively single: only the fuel urgency or mission level is determined as the optimization target, which makes it difficult to balance multiple targets such as recovery success rate, mission priority and resource utilization.

[0004] Although existing reinforcement learning methods have the potential for dynamic decision-making, they face two major bottlenecks in the design of the state-action space. On the one hand, the state representation fails to integrate the coupling effects between mechanical integrity and real-time fuel consumption, resulting in an inaccurate and comprehensive portrayal of the aircraft state. On the other hand, the action space lacks an embedded processing mechanism for physical constraints (such as cooling ban rules and emergency channel restrictions), making it difficult for the decision-making process to conform to actual physical conditions. Summary of the Invention

[0005] The purpose of this application is to provide a method, system and equipment for dynamic recovery of aircraft fleets, which can improve the real-time, accuracy and efficiency of aircraft recovery.

[0006] To achieve the above objectives, this application provides the following solutions: In a first aspect, the present application provides a method for dynamic recovery of an aircraft fleet, comprising: obtaining status data of each aircraft in an aircraft fleet at a current decision stage; constructing a state space for the current decision stage based on the status data; the state space comprising values ​​of remaining fuel, mechanical integrity, task priority, recovery success rate, and recovery status; inputting the state space into a strategy network to obtain the action probability of each aircraft at the current decision stage; the strategy network is trained using a PPO algorithm; determining an action mask vector for each aircraft at the current decision stage based on the values ​​of the remaining fuel and recovery status; optimizing the action probability based on the action mask vector, and normalizing the optimized action probability to obtain the normalized action probability of each aircraft at the current decision stage; and recovering the aircraft with the highest normalized action probability and the smallest remaining fuel.

[0007] In the second aspect, the present application provides a dynamic recovery system for an aircraft fleet, comprising: an acquisition module for acquiring status data of each aircraft in an aircraft fleet in a current decision-making stage; a construction module for constructing a state space for the current decision-making stage based on the status data; the state space includes numerical values ​​of remaining fuel, mechanical integrity, task priority, recovery success rate and recovery status; a strategy module for inputting the state space into a strategy network to obtain the action probability of each aircraft in the current decision-making stage; the strategy network is trained using a PPO algorithm; a determination module for determining the action mask vector of each aircraft in the current decision-making stage based on the numerical values ​​of the remaining fuel and the recovery status; an optimization and normalization module for optimizing the action probability based on the action mask vector, and normalizing the optimized action probability to obtain the normalized action probability of each aircraft in the current decision-making stage; a recovery module for recovering the aircraft with the highest normalized action probability and the smallest remaining fuel.

[0008] In a third aspect, the present application provides a computer device comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the above-described method for dynamic recovery of an aircraft fleet.

[0009] In a fourth aspect, the present application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of any one of the above-mentioned methods for dynamic recovery of an aircraft fleet.

[0010] According to the specific embodiments provided in this application, this application has the following technical effects.

[0011] (1) The action probability of each aircraft in the current decision stage is optimized and normalized by the action mask vector of each aircraft in the current decision stage. The normalized action probability of each aircraft in the current decision stage can be obtained in real time, realizing dynamic optimization and improving the real-time performance of aircraft recovery.

[0012] (2) The state space is represented by the numerical values ​​of the remaining fuel, mechanical integrity, mission priority, recovery success rate and recovery status, which can fully consider various abnormal conditions of the aircraft. Dynamic adjustment can be achieved through the strategy network, solving the problem of requiring manual intervention and adjustment. At the same time, the action mask vector of each aircraft in the current decision-making stage is determined by the numerical values ​​of the remaining fuel and recovery status, and the action probability generated by the state space is optimized by the action mask vector. The optimization of the action probability generated by the state space by the numerical values ​​of the remaining fuel and recovery status is realized, solving the singleness of optimization only by the fuel urgency or mission level.

[0013] (3) The action probability of each aircraft in the current decision stage is optimized and normalized through the action mask vector of each aircraft in the current decision stage, so that the normalized action probability is more accurate, thereby making the subsequent screened actions more accurate and improving the accuracy of aircraft recovery.

[0014] (4) The action probability of each aircraft in the current decision stage is optimized and normalized through the action mask vector of each aircraft in the current decision stage. The action probability of 0 can be masked by the mask vector, and after normalization and screening, the actions in the current decision stage can be quickly screened out, thereby improving the efficiency of aircraft recovery. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] In order to more clearly illustrate the technical solutions in the embodiments of the present application or related technologies, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0016] Figure 1 A flowchart of a method for dynamic recovery of a fleet of aircraft provided in one embodiment of the present application is provided.

[0017] Figure 2 A schematic structural diagram of a dynamic recovery system for a fleet of aircraft provided in one embodiment of the present application.

[0018] Figure 3 A schematic diagram of the structure of a computer device provided in one embodiment of the present application. DETAILED DESCRIPTION

[0019] The following will be combined with the drawings in the embodiments of this application to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the embodiments described are only part of the embodiments of this application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

[0020] In order to make the above-mentioned purposes, features and advantages of the present application more obvious and easy to understand, the present application is further described in detail below with reference to the accompanying drawings and specific implementation methods.

[0021] In an exemplary embodiment, Figure 1 As shown, a method for dynamic recovery of an aircraft fleet is provided. The method is executed by a computer device, and specifically can be executed by a computer device such as a terminal or a server alone, or can be executed by a terminal and a server together. In the embodiment of the present application, the method is applied to a server as an example for explanation, including the following steps S1 to S6.

[0022] Step S1: Obtain the status data of each aircraft in the aircraft fleet at the current decision stage.

[0023] Furthermore, the status data includes: initial remaining fuel, initial mechanical integrity, initial mission priority and flight status.

[0024] Step S2: Construct a state space for the current decision-making stage based on the state data; the state space includes the values ​​of remaining fuel, mechanical integrity, mission priority, recovery success rate, and recovery status.

[0025] Specifically, the state space of the current decision stage ,in, The remaining fuel quantity is set; is the mechanical integrity set; is the task priority set; is the recovery success rate set; is a numerical set of recovery status. Among them, the remaining fuel amount set Including the remaining fuel volume and mechanical integrity of all aircraft Includes mechanical integrity of all aircraft, mission priority set Including the mission priority and recovery success rate of all aircraft Includes the recovery success rate of all aircraft and the numerical value set of recovery status Includes the recovery status of all aircraft.

[0026] Furthermore, step S2 specifically includes steps S21 to S26.

[0027] Step S21: Determine the value of the recovery status based on the flight status.

[0028] Furthermore, the flight status includes: recovered status and non-recovered status; the non-recovered status includes: air standby status, go-around status and mechanical failure status.

[0029] Step S21 specifically includes: when the flight state is the recovered state, determining the value of the recovery state to be 1; when the flight state is the airborne standby state, determining the value of the recovery state to be 0; when the flight state is the go-around state, determining the value of the recovery state to be -1; when the flight state is the mechanical failure state, determining the value of the recovery state to be -2, and so on. The value of the recovery status of each aircraft .

[0030] Specifically, when the flight state is the go-around state for the first time, Aircraft cooling timer , updated every time after a decision stage ,when The flight state is automatically switched to the airborne standby state, thereby determining that the recovery state value is 0.

[0031] Step S22: normalize the initial remaining fuel amount to obtain the remaining fuel amount.

[0032] Furthermore, the calculation formula for the remaining fuel amount is as follows.

[0033] .

[0034] in, For the Aircraft in the decision phase The remaining fuel amount; For the Aircraft in the decision phase Initial remaining fuel quantity; is the number of aircraft in the fleet; For the Aircraft in the decision phase The minimum value of the initial remaining fuel quantity; For the Aircraft in the decision phase The maximum initial remaining fuel amount.

[0035] Specifically, For the Aircraft in the decision phase The initial remaining fuel quantity (i.e. the actual remaining fuel quantity (unit: tons)) is calculated, and the fuel consumption update formula is: ,in, For the Aircraft in the decision phase The initial remaining fuel volume, No. Aircraft in the decision phase The initial remaining fuel volume, is the consumption rate, is the adjacent decision time step, Second, Tons / second is the typical oil consumption rate.

[0036] Step S23: normalizing the initial mechanical integrity to obtain the mechanical integrity.

[0037] Furthermore, the calculation formula of mechanical integrity is as follows.

[0038] .

[0039] in, For the The mechanical integrity of the aircraft; For the The initial mechanical integrity of the aircraft; For maximum mechanical integrity, .

[0040] Step S24: normalize the initial task priority to obtain the task priority.

[0041] Furthermore, the calculation formula for the normalized initial task priority is as follows.

[0042] .

[0043] in, For the The normalized initial mission priority of each aircraft; For the Initial mission priority for each aircraft; is the maximum task priority, .

[0044] Furthermore, the normalized initial task priority is optimized by the value of the recycling state to obtain the task priority.

[0045] Specifically, the optimization expression of task priority is as follows.

[0046] in, For the The mission priority of each aircraft, is a dynamic adjustment factor, when hour, ,when hour, , other cases .

[0047] Step S25: Calculate the recovery success rate based on the remaining fuel amount and the mechanical integrity.

[0048] Furthermore, the calculation formula for the recovery success rate is as follows.

[0049] .

[0050] in, For the The recovery success rate of each aircraft; For the lowest recovery success rate, .

[0051] Step S26: Constructing the state space of the current decision stage based on the values ​​of the remaining fuel amount, mechanical integrity, task priority, recovery success rate and recovery status.

[0052] Step S3: Input the state space into the policy network to obtain the action probability of each aircraft in the current decision stage; the policy network is trained using the PPO algorithm.

[0053] Specifically, the network structure of the policy network is a 3-layer fully connected network with 128 neurons in each layer, the activation function is ReLU, the output layer is Softmax, and the hyperparameters include the discount factor 、Objective function clip range , value function coefficient , entropy regularization coefficient , learning rate .

[0054] The policy network uses sample state space and sample actions for training. The generation process of the sample state space and sample actions is as follows.

[0055] (1) The hyperparameter is set to: discount factor , target function clip range , the value function coefficient , entropy regularization coefficient , learning rate .

[0056] (2) Initialize the environment and set .

[0057] (3) Obtain the sample state space of the current decision stage.

[0058] (4) Input the sample state space of the current decision stage into the policy network and output the sample action probability of each aircraft in the current decision stage.

[0059] (5) The sample action probability is optimized by the action mask vector, and the optimized sample action probability is normalized to obtain the normalized sample action probability of each aircraft in the current decision stage. The aircraft with the highest normalized sample action probability among the normalized sample action probabilities of each aircraft in the current decision stage is selected. If the normalized sample action probabilities are equal, the action selected is the aircraft with the highest normalized sample action probability and the lowest remaining fuel.

[0060] (6) The aircraft selected in (5) is used as a sample action and the sample action is executed to achieve the recovery of the aircraft.

[0061] (7) After execution, the environment interacts to obtain the sample state space for the next decision stage.

[0062] Among them, the acquisition process of the flight state in the sample state space of the next decision stage: recovery success judgment, the judgment condition is: uniform random number Is it less than or equal to The recovery success rate of each aircraft ,in, , for Uniform random numbers in the interval, For the The recovery success rate of an aircraft. If (i.e. ), then determine the flight status of the aircraft, if not (i.e. ). At the same time, the aircraft in the missed approach state is updated, and the recovery state value is determined to be 0 when re-entering the queue.

[0063] (8) After determining whether the sample action is successful, the environment returns a reward based on the judgment result. .

[0064] After all aircraft are recovered, trajectory samples (sample state space, sample actions, and sample environment return rewards) are collected. T is the total number of decision stages. The training cycle is performed independently, and then the parameters are updated as follows, and updated 500 times.

[0065] Based on the above trajectory samples, the parameters of the policy network and the value learning network are updated respectively. The parameter update formula of the policy network is as follows.

[0066] .

[0067] .

[0068] in, are the parameters of the policy network, For expectations, For the strategy network, is a sample action, is the sample state space, are the initial parameters of the policy network, Therefore The strategic network and The strategic network in Select action in status The action probability ratio, is the advantage function, As the shear function, The clipping function is limited to , is the parameter of the shear function, usually 0.2.

[0069] The parameter update formula of the value learning network is as follows.

[0070] .

[0071] in, are the parameters of the value learning network, For the value network, is the discount factor power, For the stage The sample environment returns rewards, is the stage index.

[0072] During the initial training, the parameters of the policy network are the same as the initial parameters of the policy network. In the subsequent training process, every 10 training times, the initial parameters of the policy network are updated to the parameters of the 10th policy network to achieve parameter update.

[0073] Step S4: Based on the values ​​of the remaining fuel amount and the recovery status, the action mask vector of each aircraft in the current decision stage is determined.

[0074] Furthermore, step S4 specifically includes: when the value of the recovery status is 1 or -1, determining that the action mask vector of the aircraft in the current decision stage is 0; when the value of the recovery status is -2 or the remaining fuel amount is less than the remaining fuel amount threshold, determining that the action mask vector of the aircraft in the current decision stage is 2; when the value of the recovery status is 0 or the remaining fuel amount is greater than or equal to the remaining fuel amount threshold, determining that the action mask vector of each aircraft in the current decision stage is 1.

[0075] Specifically, the current decision-making stage Mask vector of the aircraft .

[0076] Step S5: Optimize the action probability based on the action mask vector, and normalize the optimized action probability to obtain the normalized action probability of each aircraft in the current decision stage.

[0077] Furthermore, step S5 specifically includes: if the action mask vector is 0, optimizing the action probability to 0; if the action mask vector is 1, keeping the action probability unchanged; if the action mask vector is 2, optimizing the action probability to 1.

[0078] Furthermore, the calculation formula for the normalized action probability of each aircraft in the current decision stage is as follows.

[0079] .

[0080] in, For the The normalized action probability of each aircraft; For the The motion mask vector of each aircraft; For the The optimized action probability of each aircraft; For the The optimized action probability of each aircraft; For the Action mask vector of the aircraft.

[0081] Step S6: Recover the aircraft with the highest normalized action probability and the smallest remaining fuel.

[0082] Specifically, , ,in, is a discrete action space, is the total number of aircraft in the fleet that are not recovered, , is the aircraft with the highest normalized action probability and the smallest remaining fuel.

[0083] The beneficial effects of the dynamic recovery method for aircraft fleet proposed in this application are mainly reflected in the following aspects.

[0084] (1) This application comprehensively considers various situations of the aircraft from the five-dimensional state space of normalized remaining fuel, normalized mechanical integrity, normalized mission priority, recovery success rate and state labeling, and realizes dynamic adjustment through the coordination of strategy network and action mask vector, thereby achieving dynamic optimization and improving the real-time performance of aircraft recovery.

[0085] (2) This application innovatively introduces an action mask mechanism to dynamically shield invalid actions, adjusts the probability of an aircraft in a fault state or with urgent remaining fuel to 1, and can quickly screen out the actions in the current decision-making stage, thereby improving the efficiency of aircraft recovery; at the same time, the normalized action probability obtained is made more accurate, thereby making the subsequent screened actions more accurate and improving the accuracy of aircraft recovery.

[0086] Based on the same inventive concept, embodiments of the present application also provide a system for dynamic recovery of a fleet of aircraft. The implementation solutions provided by this system are similar to those described in the aforementioned methods. Therefore, the specific limitations of one or more embodiments of the dynamic recovery system for a fleet of aircraft provided below can be found in the aforementioned limitations of the dynamic recovery method for a fleet of aircraft, and are not further elaborated here.

[0087] In an exemplary embodiment, Figure 2 As shown, a dynamic recovery system for an aircraft fleet is provided, including: an acquisition module for acquiring status data of each aircraft in the aircraft fleet at the current decision stage; a construction module for constructing a state space for the current decision stage based on the status data; the state space includes values ​​of remaining fuel, mechanical integrity, task priority, recovery success rate, and recovery status; a strategy module for inputting the state space into a strategy network to obtain the action probability of each aircraft at the current decision stage; the strategy network is trained using a PPO algorithm; a determination module for determining the action mask vector of each aircraft at the current decision stage based on the values ​​of the remaining fuel and the recovery status; an optimization and normalization module for optimizing the action probability based on the action mask vector and normalizing the optimized action probability to obtain the normalized action probability of each aircraft at the current decision stage; and a recovery module for recovering the aircraft with the highest normalized action probability and the smallest remaining fuel.

[0088] In an exemplary embodiment, a computer device is provided. The computer device may be a server or a terminal. The internal structure diagram thereof may be as follows: Figure 3As shown. The computer device includes a processor, a memory, an input / output interface (I / O), and a communication interface. The processor, memory, and I / O interface are connected via a system bus, and the communication interface is connected to the system bus via the I / O interface. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for the operation of the operating system and computer program in the non-volatile storage medium. The database of the computer device is used to store status data of each aircraft in the aircraft fleet at the current decision stage. The I / O interface of the computer device is used to exchange information between the processor and external devices. The communication interface of the computer device is used to communicate with an external terminal via a network connection. When the computer program is executed by the processor, a method for dynamic recovery of an aircraft fleet is implemented.

[0089] Those skilled in the art will understand that Figure 3 The structure shown in the figure is merely a block diagram of a portion of the structure related to the solution of the present application and does not constitute a limitation on the computer device to which the solution of the present application is applied. A specific computer device may include more or fewer components than shown in the figure, or combine certain components, or have a different component arrangement. In an exemplary embodiment, a computer device is provided, including a memory and a processor. The memory stores a computer program, and the processor implements the steps of the above-mentioned method embodiments when executing the computer program.

[0090] In an exemplary embodiment, a computer-readable storage medium is provided, storing a computer program. When the computer program is executed by a processor, the steps in the above-mentioned method embodiments are implemented.

[0091] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant regulations.

[0092] Those skilled in the art will appreciate that all or part of the processes in the above-mentioned embodiments can be implemented by instructing the relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the above-mentioned embodiments. In particular, any reference to memory, database, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM may be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM).

[0093] The databases involved in the various embodiments provided herein may include at least one of a relational database and a non-relational database. Non-relational databases may include, but are not limited to, distributed databases based on blockchains. The processors involved in the various embodiments provided herein may include, but are not limited to, general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic units, data processing logic units based on quantum computing, and the like.

[0094] The technical features of the above embodiments can be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0095] This document uses specific examples to illustrate the principles and implementation methods of this application. The description of the above examples is only intended to help understand the method and core concept of this application. At the same time, for those skilled in the art, based on the concept of this application, there may be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting this application.

Claims

1. A method for dynamic recovery of an aircraft fleet, characterized in that: The aircraft fleet dynamic recovery method comprises: Obtain status data of each aircraft in the aircraft fleet at the current decision stage; Constructing a state space for the current decision stage based on the state data; the state space includes values ​​for remaining fuel, mechanical integrity, mission priority, recovery success rate, and recovery status; Inputting the state space into the policy network to obtain the action probability of each aircraft in the current decision stage; the policy network is trained using the PPO algorithm; Determining an action mask vector for each aircraft in a current decision phase based on the remaining fuel amount and the recovery status; Optimizing the action probability based on the action mask vector and normalizing the optimized action probability to obtain the normalized action probability of each aircraft in the current decision stage; The aircraft with the highest normalized action probability and the smallest remaining fuel amount is recovered.

2. The method for dynamic recovery of a fleet of aircraft according to claim 1, characterized in that: The status data includes: initial remaining fuel quantity, initial mechanical integrity, initial mission priority and flight status; Constructing a state space of the current decision stage based on the state data specifically includes: determining a value of a recovery status based on the flight status; Normalizing the initial remaining fuel amount to obtain the remaining fuel amount; Normalizing the initial mechanical integrity to obtain a mechanical integrity; Normalizing the initial task priority to obtain a normalized initial task priority; The normalized initial task priority is optimized by the value of the recycling state to obtain the task priority; calculating a recovery success rate based on the remaining fuel amount and the mechanical integrity; A state space of a current decision stage is constructed based on the values ​​of the remaining fuel amount, the mechanical integrity, the task priority, the recovery success rate, and the recovery status.

3. The method for dynamic recovery of a fleet of aircraft according to claim 2, characterized in that: The calculation formula for the remaining fuel amount is: ； in, For the Aircraft in the decision phase The remaining fuel amount; For the Aircraft in the decision phase Initial remaining fuel quantity; is the number of aircraft in the fleet; For the Aircraft in the decision phase The minimum value of the initial remaining fuel quantity; For the Aircraft in the decision phase The maximum value of the initial remaining fuel quantity; The calculation formula of the mechanical integrity is: ； in, For the The mechanical integrity of the aircraft; For the The initial mechanical integrity of the aircraft; For maximum mechanical integrity; The calculation formula of the normalized initial task priority is: ； in, For the The normalized initial mission priority of each aircraft; For the Initial mission priority for each aircraft; is the maximum task priority; The calculation formula of the recovery success rate is: ； in, For the The recovery success rate of each aircraft; The lowest recovery success rate.

4. The method for dynamic recovery of a fleet of aircraft according to claim 2, characterized in that: The flight status includes: recovered status and unrecovered status; the unrecovered status includes: air standby status, go-around status and mechanical failure status; Based on the flight status, determining a value of the recovery status includes: When the flight state is a recovered state, determining the value of the recovered state is 1; When the flight state is an airborne standby state, determining the value of the recovery state is 0; When the flight state is a missed approach state, the value of the recovery state is determined to be -1; When the flight state is a mechanical failure state, the value of the recovery state is determined to be -2.

5. The method for dynamic recovery of a fleet of aircraft according to claim 4, characterized in that: Based on the remaining fuel amount and the recovery status, the action mask vector of each aircraft in the current decision stage is determined, specifically including: When the value of the recovery state is 1 or -1, the action mask vector of the aircraft in the current decision stage is determined to be 0; When the value of the recovery state is -2 or the remaining fuel amount is less than the remaining fuel amount threshold, determining that the action mask vector of the aircraft in the current decision stage is 2; When the value of the recovery state is 0 or the remaining fuel amount is greater than or equal to the remaining fuel amount threshold, the action mask vector of each aircraft in the current decision stage is determined to be 1.

6. The method for dynamic recovery of a fleet of aircraft according to claim 5, characterized in that: The action probability is optimized based on the action mask vector to obtain the optimized action probability of each aircraft in the current decision stage, specifically including: If the action mask vector is 0, the action probability is optimized to 0; If the action mask vector is 1, the action probability remains unchanged; If the action mask vector is 2, the action probability is optimized to 1.

7. The method for dynamic recovery of a fleet of aircraft according to claim 1, characterized in that: The calculation formula for the normalized action probability of each aircraft in the current decision-making stage is: ； in, For the The normalized action probability of each aircraft; For the The motion mask vector of each aircraft; For the The optimized action probability of each aircraft; For the The optimized action probability of each aircraft; For the Action mask vector of the aircraft.

8. An aircraft fleet dynamic recovery system, characterized in that: The aircraft fleet dynamic recovery system includes: An acquisition module is used to obtain the status data of each aircraft in the aircraft fleet at the current decision stage; A construction module is configured to construct a state space of a current decision stage based on the state data; the state space includes values ​​of remaining fuel, mechanical integrity, mission priority, recovery success rate, and recovery status; a strategy module, configured to input the state space into a strategy network to obtain the action probability of each aircraft in the current decision stage; the strategy network is trained using a PPO algorithm; a determination module, configured to determine an action mask vector for each aircraft in a current decision phase based on the values ​​of the remaining fuel amount and the recovery status; an optimization and normalization module, configured to optimize the action probability based on the action mask vector and normalize the optimized action probability to obtain the normalized action probability of each aircraft in the current decision stage; The recovery module is used to recover the aircraft with the highest normalized action probability and the smallest remaining fuel amount.

9. A computer device comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the aircraft fleet dynamic recovery method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method for dynamic recovery of a fleet of aircraft according to any one of claims 1 to 7 is implemented.

Images