A multi-aircraft test flight subject planning method and system considering meteorological window constraints

Abstract

This invention discloses a method and system for planning multi-aircraft flight test subjects considering meteorological window constraints, belonging to the field of aircraft flight test management technology. The method includes: constructing a multi-aircraft flight test mission planning model incorporating immediate logic constraints, resource exclusivity constraints, and meteorological time window constraints; solving the model using an improved gold mining optimization algorithm (I-GRO), representing the joint decision of "mission priority-aircraft allocation" through a two-layer hybrid encoding strategy; introducing a dynamic topology sorting and meteorological window alignment mechanism in the decoding stage to map continuous position vectors into feasible scheduling schemes that satisfy temporal logic; and in the search stage, adaptively adjusting the execution probabilities of migration, gold mining, and cooperation operators through a dynamic strategy selection mechanism, and introducing a linearly decaying elite cooperation and perturbation mechanism. This invention can effectively solve the search stagnation problem under fragmented feasible domains, significantly shorten the total flight test duration, and improve the robustness of the planning scheme.

Description

Technical Field

[0001] This invention belongs to the field of aerospace project management and operations optimization technology, specifically relating to a multi-aircraft flight test subject planning method and system that considers weather window constraints. Background Technology

[0002] Airworthiness certification flight tests for civil aircraft are a massive and complex systems engineering project, characterized by long cycles, high risks, and strong coupling. Existing flight test planning methods mostly employ metaheuristic algorithms such as genetic algorithms and particle swarm optimization. For example, one type of existing genetic algorithm is based on serial cutting and mutation, which attempts to increase population diversity and avoid getting trapped in local optima by randomly cutting and recombinating task chromosomes. However, this method has significant drawbacks when facing complex weather window constraints: First, the existence of weather windows makes the feasible solution space highly fragmented, with effective time periods being split into discrete "islands" by ineffective weather periods. Serial cutting and mutation is essentially a random sequence recombination operation, lacking the ability to perceive the time dimension. In a fragmented space, random cutting easily disrupts the established "task-window" matching relationship in parent individuals, causing many offspring solutions to become infeasible due to violations of weather time constraints, forcing the algorithm to expend significant computational resources on ineffective trial and error repair. Second, this type of method mainly focuses on solving the problem of balanced resource allocation and cannot actively align weather windows during the search process. When multiple subjects are constrained by overlapping weather windows, simple sequence variation struggles to find the optimal permutation that satisfies all temporal logics within extremely narrow feasible regions. This limits the convergence accuracy of the algorithm when dealing with strongly coupled and constrained problems, making it difficult to find the global optimal solution. Therefore, there is an urgent need for a planning method that can proactively adapt to fragmented feasible regions and has stronger spatiotemporal constraint handling capabilities. Summary of the Invention

[0003] This invention aims to...

[0004] To achieve the above-mentioned technical objectives and effects, the present invention provides the following technical solution:

[0005] A method for planning multi-aircraft flight test subjects considering weather window constraints includes the following steps:

[0006] S1. Model Construction: Based on the total number of test flight subjects, the duration of each subject, the immediate dependencies, the weather window constraints, and the number of available test aircraft, a test flight subject planning model is established with the goal of minimizing the total test flight time.

[0007] S2. Population Initialization and Encoding: Set algorithm parameters, initialize the population position of the improved gold mining optimization algorithm, and use a two-layer hybrid encoding for each individual, including a task priority vector and a resource allocation vector;

[0008] S3. Iterative optimization: The population is iteratively updated based on the improved gold mining optimization algorithm. The iterative process includes migration, gold mining and cooperation operations by a dynamic strategy selection mechanism, and introduces an elite-guided cooperation and perturbation mechanism.

[0009] S4. Topology Decoding and Fitness Evaluation: In each iteration, a decoding rule based on dynamic topology sorting is adopted, and the individual position vectors are mapped to the flight test scheduling scheme. Weather window alignment is performed during the decoding process, and the total flight test duration is used as the fitness value.

[0010] S5. Result Generation: When the termination condition is met, output the flight test subject planning scheme with the optimal fitness value.

[0011] Furthermore, the dual-layer hybrid encoding in step S2 specifically involves:

[0012] Individual X i It consists of two N-dimensional real vectors, denoted as X. i =[X seq X alloc ];

[0013] X seq =[x1,x2,...,x N [ ] represents the task priority vector, with element values This represents the relative weight by which the j-th subject is prioritized for scheduling;

[0014] X alloc =[y1,y2,...,y N [] represents the resource allocation vector, with element values... The subject j is mapped to a specific testing machine number by rounding down.

[0015] Furthermore, the calculation formula for the dynamic strategy selection mechanism described in step S3 is as follows:

[0016] The probability selection formula for the collaboration phase is:

[0017]

[0018] in, Indicates the probability of collaboration; i represents the current iteration number; Indicates the maximum number of iterations;

[0019] The probability selection formula during the gold panning stage is:

[0020]

[0021] in, Indicates the probability of finding gold;

[0022] The probabilistic selection formula during the migration phase is:

[0023]

[0024] in, This represents the migration probability.

[0025] Furthermore, in step S3, the elite-guided collaboration mechanism introduces the current globally optimal individual for guidance during position updates, and uses a linearly decaying difference variable to control the intensity of guidance.

[0026] Furthermore, the decoding rule based on dynamic topological sorting in step S4 is specifically as follows:

[0027] Step S4.1: Initialize the in-degree table for all subjects, and add all subjects with an in-degree of 0 to the candidate set. Initialize the in-degree table for all subjects, and add all subjects with an in-degree of 0 to the candidate set. ;

[0028] Step S4.2: Based on the task priority vector X seq The values ​​in the candidate set Choose a subject with a certain probability;

[0029] Step S4.3: Select the subject S sel Add to the final execution sequence and from Remove from the middle, and simultaneously iterate through all S. sel Subsequent subjects are decremented by 1 in degree, and subsequent subjects with an in degree of 0 are added to the list. ;

[0030] Step S4.4: Repeat steps 4.2 to 4.3 until all subjects have been added to the sequence. This allows us to obtain a valid sequence of tasks that satisfies the preconditions.

[0031] Furthermore, the meteorological window alignment operation described in step S4 specifically involves: if subject j is affected by the meteorological window [a j , b j Constraints, and the earliest start time ES calculated based on the predecessor relation. j Earlier than the window start time a j Then the actual start time S of the subject will be forced to be changed. j Revised to a j If the revised completion time S j + d j Later than the window closing time b jIf so, the subject will be postponed to the next feasible weather cycle or a penalty will be imposed on the solution.

[0032] On the other hand, the present invention also provides a multi-aircraft flight test subject planning system that considers weather window constraints for implementing the above method, comprising:

[0033] (1) Data acquisition module: used to input test flight subject parameters, logical dependency diagrams and meteorological window data;

[0034] (2) Algorithm processing module: Built-in improved gold mining optimization algorithm, used to perform the above population initialization, dynamic strategy search and elite guidance cooperation and perturbation operations;

[0035] (3) Decoding and scheduling module: used to perform dynamic topology decoding and meteorological window alignment, and to convert the vectors generated by the algorithm into a scheduling scheme in the form of a Gantt chart;

[0036] (4) Scheme output module: used to output the final test flight subject allocation sequence and the minimized total duration.

[0037] Compared with the prior art, the beneficial effects of the present invention are:

[0038] 1. This invention overcomes the problem of low search efficiency or even difficulty in finding the optimal solution under strong constraints in existing serial cutting and mutation algorithms. Unlike existing technologies that passively search for feasible solutions through random cutting and mutation, this invention introduces a dynamic topological sorting and meteorological window alignment mechanism in the decoding stage. This mechanism enables the algorithm to actively "absorb" the task time into the nearest effective meteorological window when generating each scheduling scheme, fundamentally ensuring the feasibility of candidate solutions in the time dimension and avoiding the feasibility destruction problem caused by random mutation.

[0039] 2. Superior optimization capability in fragmented space. Addressing the problem of feasible region fragmentation caused by weather windows, this invention employs an improved gold mining optimization algorithm (I-GRO), utilizing an elite-guided collaborative mechanism to replace blind random crossover and mutation. By dynamically adjusting migration and gold mining strategies, the algorithm can establish search bridges between different "feasible region islands," effectively solving the problems of population degradation and premature convergence caused by the large number of infeasible solutions generated by genetic algorithms under weather constraints. Thus, it can obtain a globally optimal solution with a shorter total project duration under the same computational cost. Attached Figure Description

[0040] To more clearly illustrate the technical solution of the present invention, the accompanying drawings used in the description of the specific embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0041] This accompanying drawing is a schematic diagram of functional and connection relationships, and is for illustrative purposes only, not restrictive. The shape, relative proportions, and arrangement of components can be adjusted for engineering purposes without departing from the claims.

[0042] Figure 1 This is a network topology diagram of the dependency relationships of 80 test flight subjects constructed in this embodiment of the invention;

[0043] Figure 2 This is a fusion visualization diagram under the baseline constraint scenario in an embodiment of the present invention;

[0044] Figure 3 This is a comparison chart of the convergence curves of the method (I-GRO) of this invention and the comparison algorithm under the benchmark constraint scenario in this embodiment of the invention;

[0045] Figure 4 This is a Gantt chart of flight test subject allocation and ranking generated based on the method of this invention under the baseline constraint scenario in this embodiment of the invention;

[0046] Figure 5 This is a fusion visualization diagram under complex meteorological constraints in an embodiment of the present invention;

[0047] Figure 6 This is a comparison of the convergence curves of the method (I-GRO) of this invention and the comparison algorithm under complex meteorological constraints in an embodiment of this invention.

[0048] Figure 7 This is a Gantt chart showing the allocation and ranking of flight test subjects generated based on the method of this invention under complex meteorological constraints in an embodiment of this invention.

[0049] Figure 8 This is a flowchart of the algorithm in an embodiment of the present invention. Detailed Implementation

[0050] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0051] The purpose of this invention is to provide a multi-aircraft flight test subject planning method and system that takes into account weather window constraints. By improving the gold mining optimization algorithm (I-GRO), it solves the defects of existing technology in low search efficiency and premature convergence when dealing with strongly coupled and fragmented feasible region problems.

[0052] This embodiment provides a test flight subject planning method based on an improved gold mining optimization algorithm. Its overall process principle is as follows: Figure 8 As shown, the steps include:

[0053] Based on the total number of test flight subjects, the number of test aircraft, and the execution parameters of the test aircraft, an initial set of test flight subject planning schemes is constructed.

[0054] Initialize the population to form different flight test subject planning schemes, and determine parameters such as population size, operator parameters, and maximum number of iterations.

[0055] This algorithm solves the problem of flight test subject optimization based on a two-segment hybrid coding strategy, and achieves efficient scheduling by coordinating the optimization of task timing and equipment resources.

[0056] In the task sorting dimension (real number encoding segment), continuous random values ​​(domain [0,1]) are used to drive the generation of topology sequence: based on the in-degree priority queue selection mechanism of task dependency network, the subject to be executed is dynamically selected, and the pre-constraints between subjects are updated in real time to ensure that the pre-constraints of subjects are met (e.g., subject 9 needs to be executed after subjects 5 and 8 are completed).

[0057] In the equipment allocation dimension (integer encoding segment), the subject-testing machine relationship is directly mapped through discrete integer values ​​(value range [1,M]). Combined with the equipment-driven scheduling algorithm, the earliest start time (max(predecessor completion time, testing machine idle time)) is calculated, the equipment timeline status is dynamically maintained, and the single testing machine task exclusive constraint is strictly enforced.

[0058] The dual-segment hybrid coding is coupled with a topology sorting decoder and a timetable scheduler: real-number coded segments are topologically sorted to generate a feasible sequence of tasks, while integer coded segments determine the equipment allocation scheme. Both are then input into the scheduler to generate a complete solution. This mechanism drives the evolutionary direction through a negative maximum completion time fitness function. While ensuring all engineering constraints (subject position dependency, equipment resource mutual exclusion) are met, the optimization objective is to minimize the total flight test duration, providing an efficient solution space representation framework for flight test task planning.

[0059] In this embodiment, the test flight subjects are represented as follows: ,in, Indicates the first One test flight subject, Indicates completion of the first The time required for each test flight course Indicates completion of the first The set of prerequisites required for each test flight course. Indicates the first The execution status of each test flight subject.

[0060] To verify the safety and performance of an aircraft in extreme environments, special tests such as high-temperature flight tests, low-temperature flight tests, sandstorm flight tests, and high-altitude flight tests are required. These missions are highly dependent on weather conditions and can often only be conducted during a limited number of months each year. This constraint is as follows:

[0061]

[0062]

[0063] In the formula: The start time of the subject. For its set of prerequisite subjects, C j The completion time for test flight subject j. The fixed time consumption for subject j, The earliest start time of the weather window. This is the latest closing time of the weather window.

[0064] Calculate the time required to complete each test flight course planning scheme in the initial test flight course planning scheme set.

[0065] The fitness function is used to calculate the time required for planning each flight test subject, which is equivalent to evaluating the fitness of each individual in the population.

[0066] The fitness function is used to measure the quality of an individual in a genetic algorithm. Its purpose is to quantify the fitness of an individual, enabling migration, optimization, and cooperation among individuals during the evolutionary process, ultimately finding the optimal solution to the problem. The form of the fitness function varies depending on the nature of the problem and the optimization objective.

[0067] The objective of the test flight course planning problem in this embodiment is to determine the allocation scheme with the shortest total test flight course time. Therefore, the total time is used as the standard for judging the quality of individual test flight courses. It is a function that judges the time required for the allocation of test flight subjects, and its expression is:

[0068]

[0069] In the formula, for The indicated time required to complete the flight test subject allocation plan is as follows: This represents the result of the fitness function calculation, i.e., the test flight duration.

[0070] Based on the time required for each test flight subject planning scheme, optimize the current set of test flight subject planning schemes and obtain a new set of test flight subject planning schemes.

[0071] In this embodiment, the method for optimizing the current flight test subject planning scheme set includes: sequentially performing migration operation, collaboration operation, and gold mining operation on the current flight test subject planning scheme set, with the specific formulas as follows:

[0072] The probability selection formula for the collaboration phase is:

[0073]

[0074] in, Indicates the probability of collaboration; i represents the current iteration number; Indicates the maximum number of iterations;

[0075] The probability selection formula during the gold panning stage is:

[0076] ,

[0077] in, Indicates the probability of finding gold;

[0078] The probabilistic selection formula during the migration phase is:

[0079] ,

[0080] in, This represents the migration probability.

[0081] Elite preservation is one of the core mechanisms in metaheuristic algorithms, aiming to retain historically optimal solutions to prevent the loss of high-quality solutions. This paper adopts the following improvement strategy:

[0082] Elite-guided collaboration mechanism: After position update, the current global best individual is introduced for guidance, and a linearly decaying difference variable α is used to enhance the algorithm's ability to converge to a high-quality solution region, as shown in the following formula:

[0083]

[0084] In the formula: The best individual in the current iteration. For randomly selected individuals, α is a linearly decaying difference variable, the value of which is determined by the following formula:

[0085]

[0086] The preferred range of values ​​in the formula is: .

[0087] Elite group construction: After each iteration, select the top performers based on their fitness values. The best individuals constitute the elite group. The specific formula is as follows:

[0088]

[0089]

[0090] Elite perturbation mechanism: To prevent elite individuals from causing premature population convergence, Gaussian random noise is added to each dimension of each elite individual, as shown in the following formula:

[0091]

[0092] In the formula: Gaussian noise with zero mean The intensity of the disturbance is controlled and decreases with the number of iterations.

[0093] An elite-oriented gold-mining strategy: This changes the original gold-mining operation from "moving towards random gold miners" to "moving towards the best gold miner of our time," thus giving the search process a clear direction.

[0094]

[0095] In the formula: Indicates the first The position of the best gold prospector in the generation.

[0096] This strategy ensures convergence by retaining elites, and then balances development and exploration by combining elite perturbation and guided search, thereby accelerating convergence while maintaining global search capability.

[0097] This improvement enables the algorithm to perceive that "although the current solution is still infeasible, the degree of violation is decreasing," thereby guiding the population across infeasible regions through evolutionary gradients and eventually sliding into the space of valid solutions.

[0098] Sequence generation based on dynamic topological sorting:

[0099] Initialization: Construct an in-degree table for all subjects, and add all subjects with an in-degree of 0 (i.e., subjects with no prerequisites or whose prerequisites have been completed) to the candidate set. .

[0100] Probabilistic selection: In the k-th selection, based on X seq The value in determines the value from. Which subject to select? To enhance search diversity, the index idx of the selected subject is determined using the following mapping formula, as shown in the formula:

[0101]

[0102] in x is the size of the current candidate set. k This represents the value of the corresponding dimension in the position vector.

[0103] Status update: Selected subject S sel Add to the final execution sequence and from Remove from the middle. Simultaneously, iterate through all S... sel For subsequent subjects, decrement their in-degree by 1. If the in-degree of a subsequent subject becomes 0, add it to the list. .

[0104] Loop: Repeat this process until all subjects have been added to the sequence. This allows us to obtain a valid task sequence that strictly satisfies all preconditions.

[0105] Weather window alignment: When the earliest start time of a subject is calculated (ES) j At that time, check whether it is affected by the weather window [a j , b j Constraints.

[0106] If the earliest start time is ES j Earlier than the window start time a j Then the actual start time S of the subject will be forced to be changed. j Revised to a j .

[0107] If the corrected completion time S j + d j Later than the window closing time b j If so, the subject will be postponed to the next feasible weather cycle or a penalty will be imposed on the solution.

[0108] After all subjects are arranged according to this rule, the end time of the last subject is the total test flight time of the plan.

[0109] This invention innovatively introduces an improved gold mining optimization algorithm into the flight test subject planning problem. It reconstructs the optimization path through a dynamic exploration-linear decay exploration cooperation method, solving the problem that existing metaheuristic algorithms are prone to getting trapped in local optima. By transforming the cooperation of two random individuals into random individuals moving towards the optimal individual, the solution structure becomes more directional, solving the problems of slow iteration speed and low efficiency of traditional gold mining optimization algorithms.

[0110] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0111] 1. Condition construction and parameter settings:

[0112] To verify the performance of the proposed algorithm under conditions of strong pre-dependencies and multi-resource parallelism, this paper constructs a scheduling example containing 80 flight test subjects and four test aircraft. The execution time of subject j is denoted as "day". The start time is denoted as S. j There are strong prerequisite dependencies between subjects, recorded as Pre... j Let S be the set of precedence of subject j; if i∈Pre(j), then S must satisfy... j ≥S i +p i Each testing machine can execute at most one test at a time, and the execution of the test cannot be interrupted.

[0113] Based on the prerequisite set given in Table 1, this paper abstracts the 80 test flight subjects and their sequential dependencies into a directed acyclic graph (DAG). To intuitively present the dependency structure and key chain characteristics, topological sorting is used to hierarchically arrange the nodes: nodes at higher levels correspond to subjects with no (or few) prerequisite constraints and can be triggered earlier; nodes at lower levels are often at the end of multiple dependency chains and are subject to more prerequisite constraints. Figure 1 The topology of this dependency network is illustrated, where circular nodes represent subject numbers and gray lines represent preceding dependencies. It can be seen that the network has a large number of dependency edges and significant cross-layer connections, exhibiting a typical "multi-source convergence—multi-branch diffusion" structure. This means that feasible scheduling often requires simultaneously satisfying the synchronous release conditions of multiple chains, causing the feasible domain to exhibit fragmented characteristics on the time axis, thereby increasing the difficulty of solving feasibility maintenance and critical path compression during the decoding and scheduling phase.

[0114] Table 1. Flight Test Subject Parameters

[0115]

[0116] 2. Implementation effect under baseline constraint scenario:

[0117] This embodiment solves the problem without introducing a weather window. The main challenge in this scenario stems from the dense cross-layer connections of the dependent network: on the one hand, critical path compression requires the algorithm to quickly identify and reduce waiting on the critical chain; on the other hand, multi-resource parallelism requires coordinated optimization of task allocation and sequence structure to improve overall parallelism and avoid local resource idleness.

[0118] To visually represent the coupling relationship of "dependency-assignment-sequence", this paper introduces a fusion visualization diagram: the node color represents the test machine (AFT-1 to AFT-4) to which the subject is assigned, the gray straight line is the preceding dependency edge, and the colored curve is the execution sequence edge on each test machine.

[0119] Figure 2 It can simultaneously reflect: (1) the order of subjects on the same resource; (2) whether the prerequisite is satisfied; and (3) whether cross-resource dependencies cause long-distance waiting chains. Under the basic constraints, the algorithm can usually adjust the allocation and sequence to make cross-resource dependencies proceed as early as possible, which can significantly reduce critical chain propagation wait and compress C. max The subject allocation sequence diagram under basic constraints is as follows: Figure 2 As shown.

[0120] Figure 3 Convergence curves for four algorithms under the same computational example are presented (the horizontal axis represents the number of iterations, and the vertical axis represents the objective function value per day; a smaller value indicates a shorter maximum completion time and a better scheduling scheme). Overall, all curves show a "step-like" descent, indicating that the algorithm continuously refreshes the current optimal solution during the iteration process; the long plateaus between the steps reflect stage-specific stagnation and the dilemma of local optima.

[0121] In terms of convergence speed, I-GRO achieved significant improvements in the early iterations: the target value rapidly decreased from approximately 560 days to around 520 days in the first few dozen iterations, and continued to improve thereafter, further decreasing to below 500 days in approximately 100–170 iterations, and steadily approaching the final optimum in subsequent iterations. In contrast, although GWO also decreased relatively quickly in the early stages, it rapidly entered a plateau after about 20 iterations and stagnated for a long period, exhibiting obvious premature convergence; both GRO and I-GA continued to improve in the mid-to-late stages, but the magnitude of the decrease and the frequency of improvement were significantly weaker than I-GRO, and they had a longer plateau segment, indicating that their ability to escape local optima was limited.

[0122] In terms of the quality of the final solution, I-GRO achieved the lowest final objective value. This indicates that I-GRO is not only more efficient in the early stages of the search, but also maintains effective improvement capabilities in the mid-to-late stages, balancing global exploration and local development, thus achieving better and more stable scheduling results in the fragmented feasible region caused by strong constraints.

[0123] To visually demonstrate the optimization effect of I-GRO, a task Gantt chart is used to visualize the final allocation scheme. The subject allocation Gantt chart under basic constraints is as follows: Figure 4 As shown.

[0124] 3. Implementation effect under complex weather window constraints:

[0125] Based on 80 subjects and the pre-network, a meteorological window constraint is introduced. The subjects subject to the window are 21, 23, 28, 41, 55, 67, 69, and 70, and their meteorological window data are shown in Table 2. Compared with the basic scenario, the core challenge of the complex constraint is that the earliest start time determined by the dependency relationship often does not match the window, resulting in unavoidable waiting; the misalignment between multiple windows amplifies resource gaps and lengthens the critical path, making the search process more likely to stall near the feasible boundary. Therefore, this example can more fully test the algorithm's "feasibility maintenance capability" and "cross-platform continuous improvement capability" in fragmented feasible domains.

[0126]

[0127] Under complex constraints, fusion visualization is particularly crucial: when a windowed item is located at the convergence of multiple dependency chains, its window alignment alters the feasible starting point of subsequent large chains, thus affecting the sequence structure on multiple test machines. IGRO, by implementing window alignment during the decoding phase and introducing elite collaboration and guidance strategies during the search phase, effectively compresses critical chain waiting while maintaining the feasibility of candidate solutions, thereby reducing and improving the stability of results. A fusion visualization of complex meteorological constraints is shown below. Figure 5 As shown in Figure 5.

[0128] Figure 6 The convergence process of four algorithms under complex constraints is demonstrated. It can be seen that each algorithm shows a rapid decline in the early stages, but the differences gradually widen thereafter: GWO quickly plateaus after early improvements and stagnates at a high level; GRO and I-GA, while continuing to decline slowly in the mid-to-late stages, still experience long periods of stagnation; in contrast, I-GRO not only declines faster in the early stages but also maintains a high improvement frequency in the mid-to-late stages, ultimately achieving the lowest target value. This result is consistent with the conclusions of the basic constraint example above, indicating that even when more calendar classes are introduced and coupling constraints lead to further fragmentation of the feasible region, I-GRO still possesses a stronger ability to escape local optima and a more stable solution quality advantage. The subject allocation Gantt chart under complex constraints is shown below. Figure 7 As shown.

[0129] In the description of this specification, references to terms such as "an embodiment," "example," "specific example," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0130] The preferred embodiments of the present invention disclosed above are merely illustrative of the invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the invention to the specific implementations described. Clearly, many modifications and variations can be made based on the content of this specification. This specification selects and specifically describes these embodiments to better explain the principles and practical applications of the invention, thereby enabling those skilled in the art to better understand and utilize the invention. The invention is limited only by the claims and their full scope and equivalents.

Claims

1. A method for planning multi-aircraft flight test subjects considering weather window constraints, characterized in that, Includes the following steps: S1. Model Construction: Based on the total number of test flight subjects, the duration of each subject, the immediate dependencies, the weather window constraints, and the number of available test aircraft, a test flight subject planning model is established with the goal of minimizing the total test flight time. S2. Population Initialization and Encoding: Set algorithm parameters, initialize the population position of the improved gold mining optimization algorithm, and use a two-layer hybrid encoding for each individual, including a task priority vector and a resource allocation vector; S3. Iterative optimization: The population is iteratively updated based on the improved gold mining optimization algorithm. The iterative process includes migration, gold mining and cooperation operations by a dynamic strategy selection mechanism, and introduces an elite-guided cooperation and perturbation mechanism. S4. Topology Decoding and Fitness Evaluation: In each iteration, a decoding rule based on dynamic topology sorting is adopted, and the individual position vectors are mapped to the flight test scheduling scheme. Weather window alignment is performed during the decoding process, and the total flight test duration is used as the fitness value. S5. Result Generation: When the termination condition is met, output the flight test subject planning scheme with the optimal fitness value.

2. The method according to claim 1, characterized in that, The dual-layer hybrid coding in step S2 specifically involves: Individual X i It consists of two N-dimensional real vectors, denoted as X. i =[X seq X alloc ]; X seq =[x1,x2,...,x N [ ] represents the task priority vector, with element values This represents the relative weight by which the j-th subject is prioritized for scheduling; X alloc =[y1,y2,...,y N [] represents the resource allocation vector, with element values... The subject j is mapped to a specific testing machine number by rounding down.

3. The method according to claim 1, characterized in that, The calculation formula for the dynamic strategy selection mechanism mentioned in step S3 is as follows: The probability selection formula for the collaboration phase is: in, Indicates the probability of collaboration; i represents the current iteration number; Indicates the maximum number of iterations; The probability selection formula during the gold panning stage is: in, Indicates the probability of finding gold; The probabilistic selection formula during the migration phase is: in, This represents the migration probability.

4. The method according to claim 1, characterized in that, In step S3, the elite-guided collaboration mechanism introduces the current global best individual for guidance during position updates and uses a linearly decaying difference variable to control the intensity of guidance.

5. The method according to claim 1, characterized in that, The decoding rule based on dynamic topological sorting in step S4 is as follows: Step S4.1: Initialize the in-degree table for all subjects, and add all subjects with an in-degree of 0 to the candidate set. Initialize the in-degree table for all subjects, and add all subjects with an in-degree of 0 to the candidate set. ; Step S4.2: Based on the task priority vector X seq The values ​​in the candidate set Choose a subject with a certain probability; Step S4.3: Select the subject S sel Add to the final execution sequence and from Remove from the middle, and simultaneously iterate through all S. sel Subsequent subjects are decremented by 1 in degree, and subsequent subjects with an in degree of 0 are added to the list. ; Step S4.4: Repeat steps 4.2 to 4.3 until all subjects have been added to the sequence. This allows us to obtain a valid sequence of tasks that satisfies the preconditions.

6. The method according to claim 1, characterized in that, The weather window alignment operation described in step S4 is specifically as follows: if subject j is affected by the weather window [a j , b j Constraints, and the earliest start time ES calculated based on the predecessor relation. j Earlier than the window start time a j Then the actual start time S of the subject will be forced to be changed. j Revised to a j If the revised completion time S j + d j Later than the window closing time b j If so, the subject will be postponed to the next feasible weather cycle or a penalty will be imposed on the solution.

7. A multi-aircraft flight test subject planning system considering weather window constraints for implementing the method of any one of claims 1 to 6, characterized in that, include: (1) Data acquisition module: used to input test flight subject parameters, logical dependency diagrams and meteorological window data; (2) Algorithm processing module: Built-in improved gold mining optimization algorithm, used to perform population initialization, dynamic strategy search and elite guidance cooperation and perturbation operations as described in claims 1-6; (3) Decoding and scheduling module: used to perform dynamic topology decoding and meteorological window alignment, and to convert the vectors generated by the algorithm into a scheduling scheme in the form of a Gantt chart; (4) Scheme output module: used to output the final test flight subject allocation sequence and the minimized total duration.

Images