Abnormal flight route and crew recovery method, device, equipment and medium

Abstract

The present application relates to a method, device, equipment and medium for recovering abnormal flight routes and crew members, wherein the method includes: obtaining abnormal flight routes and crew member information, and encoding to obtain a coding scheme; initializing the population to generate a first population and a second population, and generating an initial fitness value for each individual in the population; using differential evolution operators and genetic operators to move the positions of all individuals in the two populations in a random proportion to obtain a first basic population and a second basic population; sorting all individuals according to a preset sorting method, and screening out feasible solutions in the sorted first basic population and the second basic population; if the first termination condition is not met, using a strategy network model to generate a target action based on the current environmental state characteristics, re-performing dual population iterative processing, and obtaining a target feasible solution, a target partition and the number of individuals in each target partition. The present application achieves efficient integrated recovery of flight routes and crew members.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence technology, and in particular to a method, device, equipment and medium for recovering abnormal flight routes and crew members. Background Art

[0002] With the development of science and technology and the deepening of globalization, market competition is becoming increasingly fierce. Major airlines are constantly innovating and improving service quality to attract more passengers. At the same time, international routes are constantly expanding and flight frequencies are increasing, providing more options for people's cross-border travel. As important resources in flight planning, the reasonable layout and scheduling requirements of routes and crew members can not only meet the travel needs of passengers, but also avoid the waste of resources such as routes and crew members, so as to achieve the optimal resource allocation of airlines and minimize operating costs. Therefore, when an abnormal flight occurs, how to quickly and effectively provide a flight route and crew scheduling recovery plan is of great significance to airlines.

[0003] The existing airlines’ scheduling of abnormal flights is to adjust and recover them manually. However, manual adjustment relies too much on the experience of the dispatcher, which may lead to suboptimal or unrealistic recovery plans due to subjective judgment. Moreover, in the recovery problem involving both route and crew integration, since there are multiple recovery objects and goals involved, there is a coupling relationship between variables, and the problem is very complex, making it difficult for dispatchers to make accurate decisions in a timely manner. Summary of the invention

[0004] The purpose of the embodiments of the present application is to propose a method, device, equipment and medium for recovering abnormal flight routes and crew members, so as to improve the efficiency and accuracy of abnormal flight routes and crew scheduling recovery, and realize efficient integrated recovery of flight routes and crew members.

[0005] In order to solve the above technical problems, the embodiment of the present application provides an abnormal flight route and crew recovery method, including:

[0006] Acquire abnormal flight routes and crew information, and use a matrix hybrid coding method to encode the abnormal flight routes and crew information to obtain a coding scheme;

[0007] Performing random population initialization processing based on the encoding scheme to generate a first population and a second population, and generating an initial fitness value of each individual in the first population and the second population;

[0008] Entering the environmental change stage, using differential evolution operators and genetic operators to perform position movement operations on all individuals in the first population and the second population in a random proportion to obtain a first basic population and a second basic population;

[0009] Sorting all individuals in the first basic population and the second basic population respectively according to a preset sorting method to obtain a sorted first basic population and a sorted second basic population;

[0010] Screening out feasible solutions from the sorted first basic population and the sorted second basic population, and calculating partitions and the number of individuals in each partition based on the feasible solutions;

[0011] If the first termination condition is not met, the current environment state characteristics are obtained, and a pre-trained policy network model is used to generate a target action based on the current environment state characteristics, wherein the target action is the probability of the dual population implementing the differential evolution operator and the genetic operator in the next iteration;

[0012] Based on the target action, the environment change stage is re-entered to perform dual-population iterative processing until the first termination condition is met, thereby obtaining a target feasible solution, a target partition, and the number of individuals in each target partition.

[0013] In order to solve the above technical problems, the embodiment of the present application provides an abnormal flight route and crew recovery device, including:

[0014] A coding unit, used for obtaining abnormal flight routes and crew information, and encoding the abnormal flight routes and crew information using a matrix hybrid coding method to obtain a coding scheme;

[0015] an initialization unit, configured to perform random population initialization processing based on the coding scheme, generate a first population and a second population, and generate an initial fitness value of each individual in the first population and the second population;

[0016] A position moving unit, used to enter the environmental change stage, use a differential evolution operator and a genetic operator to perform a position moving operation on all individuals in the first population and the second population in a random proportion to obtain a first basic population and a second basic population;

[0017] A sorting unit, configured to sort all individuals in the first basic population and the second basic population respectively according to a preset sorting method to obtain a sorted first basic population and a sorted second basic population;

[0018] A feasible solution screening unit, used to screen out feasible solutions in the sorted first basic population and the sorted second basic population, and calculate partitions and the number of individuals in each partition based on the feasible solutions;

[0019] A target action generation unit, configured to obtain the current environment state characteristics if the first termination condition is not met, and generate a target action based on the current environment state characteristics using a pre-trained policy network model, wherein the target action is the probability of the dual population implementing the differential evolution operator and the genetic operator at the next iteration;

[0020] A target solution generating unit is used to re-enter the environmental change phase based on the target action to perform dual-population iterative processing until the first termination condition is met, thereby obtaining a target feasible solution, a target partition, and the number of individuals in each of the target partitions.

[0021] In order to solve the above technical problems, a technical solution adopted by the present invention is: to provide a computer device, including one or more processors; a memory for storing one or more programs, so that the one or more processors can implement any one of the above-mentioned abnormal flight route and crew recovery methods.

[0022] In order to solve the above technical problems, a technical solution adopted by the present invention is: a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, any one of the above-mentioned abnormal flight route and crew recovery methods is implemented.

[0023] The embodiment of the present invention provides a method, device, equipment and medium for recovering abnormal flight routes and crew members. The method includes: obtaining abnormal flight routes and crew information, and encoding the abnormal flight routes and crew information using a matrix hybrid coding method to obtain a coding scheme; performing random population initialization processing based on the coding scheme to generate a first population and a second population, and generating an initial fitness value for each individual in the first population and the second population; entering the environmental change stage, using differential evolution operators and genetic operators to perform position movement operations on all individuals in the first population and the second population in a random proportion to obtain a first basic population and a second basic population; sorting all individuals in the first basic population and the second basic population according to a preset sorting method, and obtaining the sorted individuals. The first basic population and the sorted second basic population; screen out feasible solutions from the sorted first basic population and the sorted second basic population, and calculate the partitions and the number of individuals in each partition based on the feasible solutions; if the first termination condition is not met, obtain the current environmental state characteristics, and use the pre-trained strategy network model to generate the target action based on the current environmental state characteristics, wherein the target action is the probability of the dual population implementing the differential evolution operator and the genetic operator in the next iteration; re-enter the environmental change stage based on the target action to perform iterative processing of the dual population until the first termination condition is met, and obtain the target feasible solution, target partition and the number of individuals in each target partition. The embodiment of the present invention utilizes deep reinforcement learning methods and heuristic methods, and can automatically generate integrated recovery plans for flight routes and crew members without manual participation, thereby improving the efficiency and accuracy of normal flight routes and crew scheduling and recovery, and is conducive to achieving efficient integrated recovery of flight routes and crew members. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] In order to more clearly illustrate the scheme in the present application, a brief introduction is given below to the drawings required for use in the description of the embodiments of the present application. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0025] Figure 1 It is a flowchart for implementing the abnormal flight route and crew recovery method provided in the embodiment of the present application;

[0026] Figure 2 This is a schematic diagram of an example of a matrix hybrid coding method provided in an embodiment of the present application;

[0027] Figure 3 It is a flowchart for implementing a sub-process in the abnormal flight route and crew recovery method provided in an embodiment of the present application;

[0028] Figure 4 It is a flowchart for implementing a sub-process in the abnormal flight route and crew recovery method provided in an embodiment of the present application;

[0029] Figure 5 It is a flowchart for implementing a sub-process in the abnormal flight route and crew recovery method provided in an embodiment of the present application;

[0030] Figure 6 It is a flowchart for implementing a sub-process in the abnormal flight route and crew recovery method provided in an embodiment of the present application;

[0031] Figure 7 It is a flowchart for implementing a sub-process in the abnormal flight route and crew recovery method provided in an embodiment of the present application;

[0032] Figure 8 It is a flowchart for implementing a sub-process in the abnormal flight route and crew recovery method provided in an embodiment of the present application;

[0033] Fig. 9 It is a schematic diagram of an abnormal flight route and crew recovery device provided by an embodiment of the present application;

[0034] Fig.10 It is a schematic diagram of a computer device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0035] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by technicians in the technical field of the present application; the terms used in the specification of the application herein are only for the purpose of describing specific embodiments and are not intended to limit the present application; the terms "including" and "having" and any variations thereof in the specification and claims of the present application and the above-mentioned drawings are intended to cover non-exclusive inclusions. The terms "first", "second", etc. in the specification and claims of the present application or the above-mentioned drawings are used to distinguish different objects, not to describe a specific order.

[0036] Reference to "embodiments" herein means that a particular feature, structure, or characteristic described in conjunction with the embodiments may be included in at least one embodiment of the present application. The appearance of the phrase in various locations in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0037] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the accompanying drawings.

[0038] The present invention is described in detail below with reference to the accompanying drawings and embodiments.

[0039] It should be noted that the abnormal flight route and crew recovery method provided in the embodiment of the present application is generally executed by a server, and accordingly, the abnormal flight route and crew recovery device is generally configured in the server.

[0040] See also Figure 1 Figure 2 , Figure 1 A specific implementation method of an abnormal flight route and crew recovery method is shown. Figure 2 It is a schematic diagram of an example of a matrix hybrid encoding method provided in an embodiment of the present application.

[0041] It should be noted that if there are substantially the same results, the method of the present invention is not limited to Figure 1 The process sequence shown is limited to the following steps:

[0042] S1: Obtaining abnormal flight routes and crew information, and encoding the abnormal flight routes and crew information using a matrix hybrid coding method to obtain a coding scheme.

[0043] Specifically, the abnormal flight route information includes flight number, take-off and landing airports, planned take-off time, actual take-off time, delay time or cancellation reason, etc. Crew information: including the names, positions, license numbers, etc. of the captain, co-pilot, and flight attendant. Since this application involves integer and discrete data, a matrix hybrid coding method is selected for encoding. A matrix hybrid coding method is used to map abnormal flight routes and crew information, optimization parameters, and constraints to specific rows and columns of the matrix to obtain a coding scheme.

[0044] S2: Performing random population initialization processing based on the encoding scheme to generate a first population and a second population, and generating an initial fitness value of each individual in the first population and the second population.

[0045] The embodiment of the present application is a multi-objective dual-population heuristic algorithm based on deep reinforcement learning, so it is necessary to randomly initialize the population to generate the first population and the second population, and then calculate the initial fitness value of each individual in the first population and the second population.

[0046] S3: Entering the environmental change stage, using differential evolution operators and genetic operators to perform position movement operations on all individuals in the first population and the second population in random proportions to obtain a first basic population and a second basic population.

[0047] Specifically, the initial goal of the dual population algorithm is to hope that the unconstrained subpopulation can guide the constrained subpopulation to cross the infeasible region and get a higher quality solution after getting closer to the Pareto frontier. Based on this, there are two requirements for the individual changes of the two populations, namely exploration and solution. Therefore, after the population is initialized, the environmental change stage is entered. The differential evolution operator and the genetic operator are used to perform position movement operations on all individuals in the two populations in a random proportion to generate the first basic population and the second basic population. The position movement operation here can be to modify the representation of the individual in the solution space (such as genetic coding) to explore new solution areas or optimize the current solution.

[0048] Among them, the operators in the differential evolution algorithm (DE) mainly include mutation operators, crossover operators and selection operators. These operators together constitute the core of the differential evolution algorithm, through which cooperation and competition between individuals are realized, thereby driving the population to evolve towards a better solution. Genetic operators (Genetic Operators) are the basic operations used in genetic algorithms (GA) to simulate the exchange and mutation of genetic information in the process of biological evolution. These operators select individuals with higher fitness from the population by simulating processes such as natural selection, crossover (hybridization) and mutation, and generate new individuals through genetic operations.

[0049] See also Figure 3 , Figure 3 A specific implementation of step S3 is shown, which is described in detail as follows:

[0050] S31: Entering the environmental change stage, determining the algorithms used by the first population and the second population according to the random ratio, using the population processed by the differential evolution operator as the first target population, and using the population processed by the genetic operator as the second target population.

[0051] S32: determining a parent individual from the first target population based on the initial fitness value, and performing crossover and mutation processing based on the parent individual to generate a differential population.

[0052] See also Figure 4 , Figure 4 A specific implementation of step S32 is shown, which is described in detail as follows:

[0053] S321: Selecting the individual with the highest initial fitness value from the first target population as the parent individual.

[0054] S322: Randomly perform a crossover operation on the two parent individuals to generate a first crossover individual.

[0055] S323: performing mutation processing on the first crossover individual to obtain a first mutant individual;

[0056] S324: Compare the first mutant individual with the originally corresponding first population or second population to select individuals in the first population or the second population that exceed a first preset fitness value to form a new differential population.

[0057] S325: If the second termination condition is not met, return to step S321 and execute until the second termination condition is met to obtain the differential population.

[0058] Specifically, an individual with the highest initial fitness value is selected from the first target population as a parent individual to generate the next generation population. Then two parent individuals are randomly selected, and a crossover operation is performed with a certain probability to generate a first crossover individual; then the first crossover individual is mutated with a certain probability to obtain a first mutant individual, and randomness is introduced during the mutation operation to prevent falling into a local optimal solution. Since the first population or the second population may select the differential evolution operator for processing, the first mutant individual is compared with its original population (which may be the first population or the second population), and the individuals in the first population or the second population that exceed the first preset fitness value are selected to form a new differential population. Determine whether the second termination condition is met. If not, return to execute step S321 until the second termination condition is met to obtain a differential population.

[0059] The first preset fitness value is set according to actual conditions and is not limited here. The second termination condition may be a preset number of iterations, a fitness threshold or a convergence condition.

[0060] S33: Randomly generate an initial population based on all individuals in the second target population, and select non-repetitive individuals from the initial population for mutation and crossover processing to generate a genetic population.

[0061] See also Figure 5 , Figure 5 A specific implementation of step S33 is shown, which is described in detail as follows:

[0062] S331: Randomly generate an initial population based on individuals in the second target population.

[0063] S332: Select three non-repeating individuals from the initial population, and perform differential vector calculation based on the non-repeating individuals to mutate the non-repeating individuals to obtain second mutated individuals.

[0064] S333: Perform a crossover operation on the second variant individual and the non-repetitive individual corresponding to the second variant individual to obtain a second crossover individual.

[0065] S334: Select the second crossover individuals that exceed the second preset fitness value as part of the individuals in the initial population in the next iteration.

[0066] S335: If the third termination condition is not met, return to step S332 until the third termination condition is met to obtain the genetic population.

[0067] Specifically, an initial population is randomly generated based on the individuals in the second target population, and then three non-repeating individuals (a, b, c) are randomly selected from the initial population; the difference vector is calculated as follows: V i =a+F(bc), where F is the scaling factor. Perform mutation processing according to the differential vector to obtain a second mutant individual; then perform a crossover operation on the second mutant individual and the non-repeating individual corresponding to the second mutant individual to obtain a second crossover individual. Determine whether the third termination condition is met. If not, return to step S332 until the third termination condition is met to obtain a genetic population.

[0068] The second preset fitness value is set according to actual conditions and is not limited here. The third termination condition may be a preset number of iterations, a fitness threshold or a convergence condition.

[0069] S34: Generate the first basic population and the second basic population based on the differential population and the genetic population.

[0070] Specifically, the first population is processed by a differential evolution operator or a genetic operator as the first basic population, and the second population is processed by a genetic operator or a genetic operator as the second basic population.

[0071] S4: sorting all individuals in the first basic population and the second basic population respectively according to a preset sorting method to obtain a sorted first basic population and a sorted second basic population.

[0072] See also Figure 6 , Figure 6 A specific implementation of step S4 is shown, which is described in detail as follows:

[0073] S41: Sorting all individuals in the first basic population in a constrained sorting manner to obtain the sorted first basic population.

[0074] S42: Sorting all individuals in the second basic population according to to obtain the sorted second basic population.

[0075] Specifically, the preset sorting method includes a constrained sorting method and an unconstrained sorting method. Among them, the constrained sorting method is: when two solutions are both infeasible solutions, the solution with a smaller constraint violation degree ranks higher; when one of the two solutions is a feasible solution and the other is an infeasible solution, the feasible solution ranks higher; when both solutions are feasible solutions, the solution with a smaller objective function value ranks higher. The constrained sorting method is to sort individuals based on the objective function value of the optimization target.

[0076] S5: Screen out feasible solutions from the sorted first basic population and the sorted second basic population, and calculate partitions and the number of individuals in each partition based on the feasible solutions.

[0077] See also Figure 7 , Figure 7 A specific implementation of step S5 is shown, which is described in detail as follows:

[0078] S51: Screening out feasible solutions from the sorted first basic population and the sorted second basic population, and merging the feasible solutions to obtain the feasible solution set.

[0079] S52: Calculate the partition based on the individual ordering and constraint violation degree in the feasible solution set.

[0080] S53: Calculate the number of individuals in each of the partitions.

[0081] Specifically, individuals that meet all constraints, i.e., feasible solutions, are selected from the sorted first basic population and the second basic population, respectively. This involves checking the constraints of each individual in the middle population to ensure that they do not violate any given constraints. The feasible solutions selected from the two populations are merged into a new set, called the feasible solution set. This set will be used for subsequent partitioning and individual number calculations. The feasible solution set is divided into multiple partitions based on the feasible solution set and the degree of constraint violation. The purpose of partitioning is to group similar individuals together so that subsequent operations (such as selection, crossover, and mutation) can be performed more efficiently. After the partitioning is completed, the number of individuals contained in each partition is counted, which helps to understand the scale and diversity of each partition.

[0082] S6: If the first termination condition is not met, the current environment state characteristics are obtained, and a pre-trained policy network model is used to generate a target action based on the current environment state characteristics.

[0083] The target action is the probability of the dual population implementing the differential evolution operator and the genetic operator in the next iteration. The first termination condition may be a preset number of iterations, a fitness threshold or a convergence condition.

[0084] Specifically, the embodiment of the present application uses a deep learning algorithm to construct a policy network model, and generates a target action based on the current environmental state characteristics through the policy network model.

[0085] The five-tuple required for deep reinforcement learning: environment, reward, state, action, and transition probability can be expressed in mathematical symbols as (E, R, S, A, P). The specific process of each part is:

[0086] Environment. For the parameter control problem of the dual population algorithm, it is expected to learn the optimal controller from the optimization experience obtained when optimizing a set of problems. Therefore, the environment consists of a set of optimization problems (called training functions). They are used to evaluate the performance of the controller during the learning process. It should be noted that these training functions have common characteristics, and good parameter controllers can be obtained by learning these characteristics.

[0087] State. The state is used to reflect the iterative progress of the current individual and population, and provide decision information for the iteration of deep reinforcement learning. That is, the design of the state needs to reflect the characteristics of the individual and the group as well as the status of the Pareto frontier. In the embodiment of this application, the state design is as follows: 1. Generational distance (GD): 2. Inverted generational distance (IGD): 3. The progress of the current iteration: NFE / MAXNFE, where NFE is the number of fitness value evaluations and MAXNFE is the maximum number of fitness value evaluations set; 4. The proportion of non-dominated solutions in the population: n nondominated / S; 5. The distance correlation between the fitness of the ith individual and the other three randomly selected Pareto solutions (when the number of non-dominated solutions is less than three, randomly replicate until the condition is met); 6. The ratio of the number of chemotaxis, reproduction, elimination and diffusion to the total number. Among them, when p = 2, d i represents the Euclidean distance between the i-th solution obtained and the nearest reference point, and Represents the distance from the Pareto solution set to the nearest reference point in the population.

[0088] Action. In a Markov decision process (MDP), given a state S t , the agent can be defined as a probability distribution p(A t |S t ;θ) selects (samples) an action from the strategy π, where θ represents the parameters of the strategy. t Defined as the control parameter,

[0089] Reward: For optimization problems, it is necessary to guide the agent to provide good actions for the dual population algorithm, so as to generate a better Pareto frontier in the iterative process. The design of the reward needs to be based on the state update of the Pareto frontier.

[0090] See also Figure 8 , Figure 8 A specific implementation method before step S6 is shown, which is described in detail as follows:

[0091] S61: Setting initial parameters of the dual population algorithm and initializing the policy network model to obtain an initial policy network model.

[0092] S62: In each time step, obtain the current state, and sample an action from the initial strategy network model according to the current state.

[0093] S63: Generate new state features based on the action through the dual population algorithm and calculate the reward.

[0094] S64: Update the parameters of the initial policy network model based on the new state features and the reward, and iteratively train the initial policy network model to obtain the pre-trained policy network model.

[0095] Specifically, before step S6, the embodiment of the present application also needs to use a deep reinforcement learning algorithm to train the policy network model. The dual population algorithm refers to an algorithm that uses differential evolution operators and genetic operators to process two populations in the embodiment of the present application. The initial parameter setting includes defining the size of the two populations, crossover rate, mutation rate, number of iterations and other algorithm-related parameters. The action generation process includes: inputting the current state into the policy network; outputting the probability distribution of the action through the policy network; and randomly selecting an action according to the probability distribution. After executing the selected action, the environment will transfer to a new state. In the dual population algorithm, the characteristics of the new state are used for interaction or evaluation between the two populations. The reward value is calculated based on the new state (or the difference between the new and old states), which is a quantitative indicator for evaluating the quality of the action. The reward value and the new state characteristics are used to update the parameters of the policy network to optimize the network's ability to make better decisions in the future. Repeat the above process until a predetermined number of iterations is reached or other stop conditions are met to obtain a pre-trained policy network model.

[0096] S7: re-entering the environmental change phase based on the target action to perform dual-population iterative processing until the first termination condition is met, thereby obtaining a target feasible solution, a target partition, and the number of individuals in each target partition.

[0097] Specifically, in the embodiment of the present application, it is necessary to return to execute steps S3-S7 according to the target action until the first termination condition is met, and the target feasible solution, target partition and the number of individuals in each target partition are obtained. Since the target action is the probability of implementing the differential evolution operator and the genetic operator in the dual population at the next iteration, when iterating again in the environmental change stage, it is necessary to change the above step S3 to enter the environmental change stage, and use the differential evolution operator and the genetic operator to perform position movement operations on all individuals in the first population and the second population based on the target action to obtain the first basic population and the second basic population, and the other steps are the same.

[0098] Each of these objectives can be solved as an integrated recovery plan for flight routes and crew members.

[0099] In an embodiment of the present application, abnormal flight routes and crew information are obtained, and matrix hybrid coding is used to encode the abnormal flight routes and crew information to obtain a coding scheme; based on the coding scheme, random population initialization processing is performed to generate a first population and a second population, and an initial fitness value of each individual in the first population and the second population is generated; entering the environmental change stage, a differential evolution operator and a genetic operator are used to perform a position movement operation on all individuals in the first population and the second population in a random proportion to obtain a first basic population and a second basic population; all individuals in the first basic population and the second basic population are sorted according to a preset sorting method to obtain sorted individuals. The first basic population and the sorted second basic population; the feasible solutions in the sorted first basic population and the sorted second basic population are screened out, and the partitions and the number of individuals in each partition are calculated based on the feasible solutions; if the first termination condition is not met, the current environmental state characteristics are obtained, and the pre-trained strategy network model is used to generate the target action based on the current environmental state characteristics, wherein the target action is the probability of the dual population implementing the differential evolution operator and the genetic operator at the next iteration; based on the target action, the environmental change stage is re-entered to perform iterative processing of the dual population until the first termination condition is met, and the target feasible solution, target partition and the number of individuals in each target partition are obtained. The embodiment of the present invention utilizes deep reinforcement learning methods and heuristic methods, which can automatically generate integrated recovery plans for flight routes and crew members without manual participation, improve the efficiency and accuracy of normal flight routes and crew scheduling and recovery, and are conducive to achieving efficient integrated recovery of flight routes and crew members. The embodiment of the present application can obtain a set of excellent integrated recovery plans for flight routes and crew members in a relatively short time, and can obtain better solutions than other latest multi-objective multi-population heuristic methods. The embodiments of the present application incorporate deep reinforcement learning methods for optimization, which improves the algorithm's ability to solve large-scale complex optimization problems and adapt to the needs of practical problems.

[0100] Please refer to Fig. 9 , as a response to the above Figure 1 The present application provides an embodiment of an abnormal flight route and crew recovery device, and the device embodiment is similar to Figure 1 Corresponding to the method embodiment shown, the device can be specifically applied to various electronic devices.

[0101] like Fig. 9 As shown, the abnormal flight route and crew recovery device of this embodiment includes: an encoding unit 81, an initialization unit 82, a position movement unit 83, a sorting unit 84, a feasible solution screening unit 85, a target action generation unit 86 and a target solution generation unit 87, wherein:

[0102] The encoding unit 81 is used to obtain the abnormal flight route and crew information, and encode the abnormal flight route and crew information using a matrix hybrid encoding method to obtain a coding scheme;

[0103] An initialization unit 82, configured to perform random population initialization processing based on the coding scheme, generate a first population and a second population, and generate an initial fitness value of each individual in the first population and the second population;

[0104] A position moving unit 83 is used to enter the environmental change stage, use a differential evolution operator and a genetic operator to perform a position moving operation on all individuals in the first population and the second population in a random proportion to obtain a first basic population and a second basic population;

[0105] A sorting unit 84 is used to sort all individuals in the first basic population and the second basic population according to a preset sorting method to obtain a sorted first basic population and a sorted second basic population;

[0106] A feasible solution screening unit 85 is used to screen out feasible solutions in the sorted first basic population and the sorted second basic population, and calculate partitions and the number of individuals in each partition based on the feasible solutions;

[0107] A target action generating unit 86 is used to obtain the current environment state characteristics if the first termination condition is not met, and use the pre-trained policy network model to generate a target action based on the current environment state characteristics, wherein the target action is the probability of the dual population implementing the differential evolution operator and the genetic operator in the next iteration;

[0108] The target solution generating unit 87 is used to re-enter the environment change stage based on the target action to perform dual-population iterative processing until the first termination condition is met, thereby obtaining a target feasible solution, a target partition and the number of individuals in each target partition.

[0109] Further, the position moving unit 83 includes:

[0110] A population confirmation unit, used for entering the environmental change stage, determining the algorithms used by the first population and the second population according to the random ratio, using the population processed by the differential evolution operator as the first target population, and using the population processed by the genetic operator as the second target population;

[0111] A differential population generating unit, configured to determine a parent individual from the first target population based on the initial fitness value, and perform crossover and mutation processing based on the parent individual to generate a differential population;

[0112] A genetic population generation unit, used to randomly generate an initial population based on all individuals in the second target population, and select non-repeating individuals from the initial population for mutation and crossover processing to generate a genetic population;

[0113] A basic population confirmation unit is used to generate the first basic population and the second basic population based on the differential population and the genetic population.

[0114] Furthermore, the differential population generation unit includes:

[0115] A parent individual selection unit, used to select the individual with the highest initial fitness value from the first target population as the parent individual;

[0116] A first crossover operation unit, used for randomly performing a crossover operation on two parent individuals to generate a first crossover individual;

[0117] A first variation processing unit, configured to perform variation processing on the first crossover individual to obtain a first variation individual;

[0118] A comparison unit, used for comparing the first variant individual with the originally corresponding first population or second population, so as to select individuals in the first population or the second population that exceed a first preset fitness value to form a new differential population;

[0119] The first iteration unit is used to return to execute the parent individual selection unit if the second termination condition is not met, until the second termination condition is met to obtain the differential population.

[0120] Furthermore, the genetic population generation unit includes:

[0121] An initial population generating unit, configured to randomly generate an initial population based on individuals in the second target population;

[0122] A second variation processing unit is used to select three non-repeating individuals from the initial population, and perform differential vector calculation based on the non-repeating individuals to perform variation processing on the non-repeating individuals to obtain second variation individuals;

[0123] A second crossover operation unit, configured to perform a crossover operation on the second variant individual and a non-repeating individual corresponding to the second variant individual to obtain a second crossover individual;

[0124] A crossover individual selection unit, used for selecting second crossover individuals exceeding a second preset fitness value as part of the individuals in the initial population in the next iteration;

[0125] The second iteration processing unit is used to return to execute the second variation processing unit if the third termination condition is not met, until the third termination condition is met to obtain the genetic population.

[0126] Further, the sorting unit 84 includes:

[0127] A first sorting unit is configured to sort all individuals in the first basic population in a constrained sorting manner to obtain the sorted first basic population, wherein the constrained sorting manner is as follows: when two solutions are both infeasible solutions, the solution with a smaller constraint violation degree is ranked higher; when one of the two solutions is a feasible solution and the other is an infeasible solution, the feasible solution is ranked higher; when both solutions are feasible solutions, the solution with a smaller objective function value is ranked higher;

[0128] The second sorting unit is used to sort all individuals in the second basic population in an unconstrained sorting manner to obtain the sorted second basic population.

[0129] Furthermore, the target action generating unit 86 further includes:

[0130] A model initialization unit is used to set the initial parameters of the dual population algorithm and initialize the policy network model to obtain an initial policy network model;

[0131] An action generation unit, used for obtaining a current state in each time step, and sampling an action from the initial strategy network model according to the current state;

[0132] A reward calculation unit, used to generate new state features based on the action through the dual population algorithm and calculate a reward;

[0133] A model training unit is used to update the parameters of the initial policy network model based on the new state features and the reward, and iteratively train the initial policy network model to obtain the pre-trained policy network model.

[0134] Further, the feasible solution screening unit 85 includes:

[0135] A feasible solution set generating unit, used for screening out feasible solutions in the sorted first basic population and the sorted second basic population, and merging the feasible solutions to obtain the feasible solution set;

[0136] A partition calculation unit, configured to calculate the partition based on the individual rankings and constraint violation degrees in the feasible solution set;

[0137] The individual quantity calculation unit is used to calculate the individual quantity of each partition.

[0138] To solve the above technical problems, the present application also provides a computer device. Fig.10 , Fig.10 This is a basic structural block diagram of the computer device in this embodiment.

[0139] The computer device 9 includes a memory 91, a processor 92, and a network interface 93 that are interconnected and communicated through a system bus. It should be noted that the figure only shows a computer device 9 having three components: a memory 91, a processor 92, and a network interface 93, but it should be understood that it is not required to implement all the components shown, and more or fewer components can be implemented instead. Among them, those skilled in the art can understand that the computer device here is a device that can automatically perform numerical calculations and / or information processing according to pre-set or stored instructions, and its hardware includes but is not limited to microprocessors, application specific integrated circuits (ASIC), programmable gate arrays (FPGA), digital signal processors (DSP), embedded devices, etc.

[0140] Computer devices can be computing devices such as desktop computers, notebooks, PDAs, and cloud servers. Computer devices can interact with users through keyboards, mice, remote controls, touch pads, or voice control devices.

[0141] The memory 91 includes at least one type of readable storage medium, and the readable storage medium includes flash memory, hard disk, multimedia card, card-type memory (for example, SD or DX memory, etc.), random access memory (RAM), static random access memory (SRAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), programmable read-only memory (PROM), magnetic memory, disk, optical disk, etc. In some embodiments, the memory 91 can be an internal storage unit of the computer device 9, such as a hard disk or memory of the computer device 9. In other embodiments, the memory 91 can also be an external storage device of the computer device 9, such as a plug-in hard disk equipped on the computer device 9, a smart memory card (SmartMedia Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), etc. Of course, the memory 91 can also include both the internal storage unit of the computer device 9 and its external storage device. In this embodiment, the memory 91 is generally used to store the operating system and various application software installed on the computer device 9, such as the program code of the abnormal flight route and the crew recovery method, etc. In addition, the memory 91 can also be used to temporarily store various types of data that have been output or are to be output.

[0142] In some embodiments, the processor 92 may be a central processing unit (CPU), a controller, a microcontroller, a microprocessor, or other data processing chips. The processor 92 is generally used to control the overall operation of the computer device 9. In this embodiment, the processor 92 is used to run the program code stored in the memory 91 or process data, such as running the program code of the abnormal flight route and crew recovery method described above, so as to implement various embodiments of the abnormal flight route and crew recovery method.

[0143] The network interface 93 may include a wireless network interface or a wired network interface, and the network interface 93 is generally used to establish a communication connection between the computer device 9 and other electronic devices.

[0144] The present application also provides another embodiment, namely, providing a computer-readable storage medium, which stores a computer program, and the computer program can be executed by at least one processor to enable the at least one processor to perform the steps of an abnormal flight route and crew recovery method as described above.

[0145] Through the description of the above implementation methods, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be implemented by means of software plus a necessary general hardware platform, and of course by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present application, or the part that contributes to the prior art, can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, a disk, or an optical disk), and includes a number of instructions for a terminal device (which can be a mobile phone, a computer, a server, an air conditioner, or a network device, etc.) to execute the methods of each embodiment of the present application.

[0146] Obviously, the embodiments described above are only some embodiments of the present application, rather than all embodiments. The preferred embodiments of the present application are given in the accompanying drawings, but they do not limit the patent scope of the present application. The present application can be implemented in many different forms. On the contrary, the purpose of providing these embodiments is to make the understanding of the disclosure of the present application more thorough and comprehensive. Although the present application is described in detail with reference to the aforementioned embodiments, for those skilled in the art, it is still possible to modify the technical solutions recorded in the aforementioned specific implementation methods, or to replace some of the technical features therein with equivalents. Any equivalent structure made using the contents of the specification and drawings of this application, directly or indirectly used in other related technical fields, is similarly within the scope of patent protection of this application.

Claims

1. A method for recovering abnormal flight routes and crew members, characterized in that: include: Acquire abnormal flight routes and crew information, and use a matrix hybrid coding method to encode the abnormal flight routes and crew information to obtain a coding scheme; Performing random population initialization processing based on the encoding scheme to generate a first population and a second population, and generating an initial fitness value of each individual in the first population and the second population; Entering the environmental change stage, using differential evolution operators and genetic operators to perform position movement operations on all individuals in the first population and the second population in a random proportion to obtain a first basic population and a second basic population; Sorting all individuals in the first basic population and the second basic population respectively according to a preset sorting method to obtain a sorted first basic population and a sorted second basic population; Screening out feasible solutions from the sorted first basic population and the sorted second basic population, and calculating partitions and the number of individuals in each partition based on the feasible solutions; If the first termination condition is not met, the current environment state characteristics are obtained, and a pre-trained policy network model is used to generate a target action based on the current environment state characteristics, wherein the target action is the probability of the dual population implementing the differential evolution operator and the genetic operator in the next iteration; Re-entering the environmental change phase based on the target action to perform dual-population iterative processing until the first termination condition is met, and obtaining a target feasible solution, a target partition, and the number of individuals in each target partition; When the environment changes, the differential evolution operator and the genetic operator are used to perform position movement operations on all individuals in the first population and the second population in a random proportion to obtain a first basic population and a second basic population, including: Entering the environmental change stage, determining the algorithms used by the first population and the second population according to the random ratio, using the population processed by the differential evolution operator as the first target population, and using the population processed by the genetic operator as the second target population; Determining a parent individual from the first target population based on the initial fitness value, and performing crossover and mutation processing based on the parent individual to generate a differential population; Randomly generate an initial population based on all individuals in the second target population, and select non-repeating individuals from the initial population for mutation and crossover processing to generate a genetic population; The first basic population and the second basic population are generated based on the differential population and the genetic population.

2. The abnormal flight route and crew recovery method according to claim 1, characterized in that: The step of determining a parent individual from the first target population based on the initial fitness value, and performing crossover and mutation processing based on the parent individual to generate a differential population includes: Selecting the individual with the highest initial fitness value from the first target population as the parent individual; Randomly crossover the two parent individuals to generate a first crossover individual; Performing mutation processing on the first crossover individual to obtain a first mutant individual; Comparing the first variant individual with the original corresponding first population or second population, so as to select individuals in the first population or the second population that exceed a first preset fitness value to form a new differential population; If the second termination condition is not met, the process returns to the step of selecting the individual with the highest initial fitness value from the first target population as the parent individual until the second termination condition is met to obtain the differential population.

3. The abnormal flight route and crew recovery method according to claim 1, characterized in that: The step of randomly generating an initial population based on all individuals in the second target population, and selecting non-repeating individuals from the initial population for mutation and crossover processing to generate a genetic population includes: Randomly generate the initial population based on individuals in the second target population; Selecting three non-repeating individuals from the initial population, and performing differential vector calculation based on the non-repeating individuals to perform mutation processing on the non-repeating individuals to obtain second mutant individuals; Perform a crossover operation on the second variant individual and the non-repeating individual corresponding to the second variant individual to obtain a second crossover individual; Selecting second crossover individuals that exceed a second preset fitness value as part of the individuals in the initial population in the next iteration; If the third termination condition is not met, return to the step of selecting three non-repetitive individuals from the initial population, and performing differential vector calculation based on the non-repetitive individuals to mutate the non-repetitive individuals to obtain second mutated individuals, until the third termination condition is met to obtain the genetic population.

4. The abnormal flight route and crew recovery method according to claim 1, characterized in that: The step of sorting all the individuals in the first basic population and the second basic population respectively according to a preset sorting method to obtain a sorted first basic population and a sorted second basic population comprises: Sorting all individuals in the first basic population in a constrained sorting manner to obtain the sorted first basic population, wherein the constrained sorting manner is: when two solutions are both infeasible solutions, the solution with a smaller constraint violation degree is ranked higher; when one of the two solutions is a feasible solution and the other is an infeasible solution, the feasible solution is ranked higher; when both solutions are feasible solutions, the solution with a smaller objective function value is ranked higher; All individuals in the second basic population are sorted in an unconstrained sorting manner to obtain the sorted second basic population.

5. The abnormal flight route and crew recovery method according to claim 1, characterized in that: If the first termination condition is not met, the current environment state feature is obtained, and before the target action is generated based on the current environment state feature using a pre-trained policy network model, the method further includes: Set the initial parameters of the dual population algorithm and initialize the policy network model to obtain the initial policy network model; In each time step, a current state is obtained, and an action is sampled from the initial policy network model according to the current state; Generate new state features based on the action through the dual population algorithm and calculate the reward; The parameters of the initial policy network model are updated based on the new state features and the reward, and the initial policy network model is iteratively trained to obtain the pre-trained policy network model.

6. The abnormal flight route and crew recovery method according to any one of claims 1 to 5, characterized in that: The step of screening out feasible solutions from the sorted first basic population and the sorted second basic population, and calculating partitions and the number of individuals in each partition based on the feasible solutions, includes: Screening out feasible solutions from the sorted first basic population and the sorted second basic population, and merging the feasible solutions to obtain the feasible solution set; Calculating the partition based on the individual rankings and constraint violation levels in the feasible solution set; The number of individuals in each of the partitions is calculated.

7. An abnormal flight route and crew recovery device, characterized in that: include: A coding unit, used for obtaining abnormal flight routes and crew information, and encoding the abnormal flight routes and crew information using a matrix hybrid coding method to obtain a coding scheme; an initialization unit, configured to perform random population initialization processing based on the coding scheme, generate a first population and a second population, and generate an initial fitness value of each individual in the first population and the second population; A position moving unit, used to enter the environmental change stage, use a differential evolution operator and a genetic operator to perform a position moving operation on all individuals in the first population and the second population in a random proportion to obtain a first basic population and a second basic population; A sorting unit, configured to sort all individuals in the first basic population and the second basic population respectively according to a preset sorting method to obtain a sorted first basic population and a sorted second basic population; A feasible solution screening unit, used to screen out feasible solutions in the sorted first basic population and the sorted second basic population, and calculate partitions and the number of individuals in each partition based on the feasible solutions; A target action generation unit, configured to obtain the current environment state characteristics if the first termination condition is not met, and generate a target action based on the current environment state characteristics using a pre-trained policy network model, wherein the target action is the probability of the dual population implementing the differential evolution operator and the genetic operator at the next iteration; A target solution generating unit, configured to re-enter the environmental change phase based on the target action to perform dual-population iterative processing until the first termination condition is met, thereby obtaining a target feasible solution, a target partition, and the number of individuals in each target partition; The position moving unit comprises: A population confirmation unit, used for entering the environmental change stage, determining the algorithms used by the first population and the second population according to the random ratio, using the population processed by the differential evolution operator as the first target population, and using the population processed by the genetic operator as the second target population; A differential population generating unit, configured to determine a parent individual from the first target population based on the initial fitness value, and perform crossover and mutation processing based on the parent individual to generate a differential population; A genetic population generation unit, used to randomly generate an initial population based on all individuals in the second target population, and select non-repeating individuals from the initial population for mutation and crossover processing to generate a genetic population; A basic population confirmation unit is used to generate the first basic population and the second basic population based on the differential population and the genetic population.

8. A computer device, characterized in that: It comprises a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, it implements the abnormal flight route and crew recovery method as described in any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the abnormal flight route and crew recovery method according to any one of claims 1 to 6 is implemented.

Images