Scheduling methods and systems for rapid testing of batch satellites

Abstract

This invention provides a scheduling method and system for rapid batch testing of satellites, relating to the field of scheduling technology for rapid batch testing of satellites. The method includes establishing a discrete event simulation model based on test items; constructing a bi-objective mixed-integer programming model; solving the bi-objective mixed-integer programming model using a constraint-satisfaction-based heuristic scheduling algorithm; and constructing a conflict topology graph among test items based on resource occupation conflict relationships for each test workstation, and using a greedy selection algorithm for parallel scheduling of test items. The bi-objective mixed-integer programming model can simultaneously consider production efficiency and resource utilization balance. The constraint-satisfaction-based heuristic scheduling algorithm, through enumeration and Pareto non-dominated sorting to select the optimal solution, has a simple structure, few parameters, and fast computation speed. The use of conflict topology graph modeling and concurrent batch partitioning through a greedy selection algorithm or graph coloring algorithm significantly improves the computational efficiency of the underlying scheduling.

Description

Technical Field

[0001] This invention relates to the field of scheduling technology for rapid testing of batch satellites, specifically to scheduling methods and systems for rapid testing of batch satellites. Background Technology

[0002] With the trend of mass production and rapid deployment of satellite constellations, the AIT cycle of a single satellite has been compressed from months to weeks or even days. However, the scheduling complexity of parallel testing of multiple satellites has increased significantly. Existing satellite test scheduling methods typically employ fixed pipeline sequential operations or simple priority queuing strategies, which have two prominent problems: First, at the macro level, there is a lack of effective modeling for the collaborative scheduling of multiple satellites, multiple workstations, and multiple test stages, making it difficult to balance the conflict between the total completion cycle and resource load. This results in some test workstations being overloaded while others are idle, leading to low overall test efficiency. Second, at the micro level, the conflicts arising from shared resources such as test instruments, cables, and operators within the same test workstation are ignored. Existing methods often use serial execution or simple mutex lock mechanisms, resulting in low resource utilization and insufficient parallelism. In the existing technology, the document with publication number CN112257909A proposes a method for optimizing the scheduling strategy of batch satellites in the test island pipeline mode by explicitly using an adaptive genetic algorithm. However, this method only uses simple node completion variables and reference value threshold judgments, does not distinguish between guaranteed and exclusive occupancy types, does not establish an event-driven simulation loop mechanism, and only takes the shortest total time as the single objective. It uses a genetic algorithm without involving a Pareto screening mechanism, only traversing all possible schemes or using a genetic algorithm for optimization, does not involve conflict topology graphs and hybrid scheduling switching mechanisms, and does not involve any dynamic parameter adaptive adjustment mechanism based on actual execution feedback. Therefore, there is an urgent need for a scheduling method and system for rapid testing of batch satellites.

[0003] The information disclosed in the background section is only intended to enhance the understanding of the background of this disclosure, and therefore may include information that does not constitute prior art known to those skilled in the art. Summary of the Invention

[0004] The purpose of this invention is to provide a scheduling method and system for rapid testing of batch satellites, in order to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides the following technical solution: The scheduling method for rapid testing of batch satellites includes the following steps: S1: For each satellite in the batch of satellites, the entire testing process of the satellite is broken down into multiple test stations connected in sequence according to function. Each test station contains several test items, and a discrete event simulation model is established based on the test items. S2: Using the discrete event simulation model as the operational support, with the shortest total completion cycle of batch satellite parallel operations as the first optimization objective and the minimization of resource load balance among each test station segment as the second optimization objective, a dual-objective mixed integer programming model is constructed; the dual-objective mixed integer programming model is solved using a constraint-based heuristic scheduling algorithm, and a scheduling scheme for the macro-strategy layer is selected through Pareto non-dominated sorting; S3: After completing the scheduling scheme of the macro strategy layer, for each test station segment, based on the corresponding test items, construct a conflict topology graph between test items based on the resource occupation conflict relationship. According to the conflict topology graph, the greedy selection algorithm is used to schedule the test items in parallel. When the test items in any test station segment exceed the preset threshold, switch to the graph coloring algorithm for parallel scheduling. S4: Based on the parallel scheduling results, optimize the process with the goal of minimizing the completion time of all test items within each test station segment, and generate a parallel scheduling strategy for the micro-execution layer. After completing the parallel scheduling strategy for the micro-execution layer, collect feedback data on the actual execution of each test station segment. The macro-strategy layer dynamically updates the allocation priority and test duration parameters of the test station segments based on the feedback data. Repeat the above update process until the preset maximum number of updates is reached, and output the final scheduling strategy.

[0006] Furthermore, in the discrete event simulation model, for each test station segment, resource status variables and project status variables are set based on the test project. The resource status variables are used to characterize the idle state, guaranteed occupancy state, and exclusive occupancy state of resources. The project status variables are used to characterize the unstarted state, in-process state, and completed state of the test project. The executable conditions of the test project are determined according to the guaranteed occupancy requirement or exclusive occupancy requirement of the test project for resources. Test projects that do not meet the executable conditions are put into the waiting queue. The start events of the test items are added to the event queue according to the preset scheduling order, and the simulation clock is advanced according to the event occurrence time. When a test project meets the executable conditions, it occupies the corresponding resources and generates a completion event. When the test project is completed, it releases the corresponding resources and triggers the execution of test projects in the waiting queue that meet the executable conditions. Once all test items within a test station segment are completed and related resources are released, the completion time of that test station segment is output, and the corresponding satellite is allowed to enter the next test station segment.

[0007] Furthermore, a heuristic scheduling algorithm based on constraint satisfaction is used to solve the bi-objective mixed integer programming model, specifically as follows: The bi-objective mixed integer programming model takes the shortest total completion cycle of batch satellite parallel operations as the primary optimization objective and the optimal resource load balance between test workstations as the secondary optimization objective. The total completion cycle is characterized by the maximum completion time among all satellite completion times, and the resource load balance is characterized by the deviation of the utilization rate of each test station segment under each test phase from the average utilization rate. The smaller the deviation, the more balanced the resource load among the test station segments. The utilization rate of the test station segment is determined based on the actual usage time of the corresponding test station segment and the total completion cycle; and the test duration of each satellite in each test stage is a preset fixed duration, and each test stage is executed sequentially according to a predetermined order, with test station segments with the same function in the same test stage being used in parallel.

[0008] Furthermore, the candidate scheduling strategies under the corresponding rule combinations are obtained, specifically: When solving the bi-objective mixed integer programming model using a constraint-satisfaction-based heuristic scheduling algorithm, satellite sorting rules and test station allocation rules are pre-set, and the satellite sorting rules and test station allocation rules are combined to form multiple rule combinations. For each rule combination, the satellites are processed sequentially according to the order of each test phase of the satellite AIT integrated test. In the current test phase, the satellites to be scheduled are sorted according to the corresponding satellite sorting rules, and the sorted satellites are sequentially assigned to the parallel test workstations with the same function in the current test phase according to the corresponding test workstation allocation rules. The satellite sorting rules include sorting according to preset priority and planned launch time, and the test workstation allocation rules include prioritizing the earliest available test workstation. The start time of each satellite in the current testing phase is determined by the completion time of its previous testing phase and the available time of the assigned testing station segment. After the allocation is completed, the available time of the corresponding testing station segment is updated. After all testing phases are completed, candidate scheduling strategies under the corresponding rule combinations are obtained.

[0009] Furthermore, a scheduling scheme at the macro-strategy level is obtained, specifically: each candidate scheduling strategy is evaluated to obtain the corresponding total completion period and resource load balance of the test station segment. The total completion period is characterized by the maximum completion time among all satellite completion times, and the resource load balance is characterized by the deviation of the utilization rate of each test station segment from the average utilization rate. Pareto non-dominance screening is performed based on the total completion time and resource load balance of each candidate scheduling strategy. The candidate scheduling strategy that is not dominated by other candidate scheduling strategies in both total completion time and resource load balance is selected as the scheduling scheme of the macro strategy layer.

[0010] Furthermore, the established conflict topology graph is as follows: Conflict topology graph between test items established in the micro-execution layer ,in To test the nodes of the project, For conflict edges, edge Indicates test items and There are resource conflicts, so parallel execution is not possible; for any two different test projects and Iterate through all resources and check for conflicts according to the following conflict determination rules: If two test items occupy the same resource, there will be a conflict if at least one of them is an exclusive occupy; there will be no conflict only if both are guaranteed occupy. The data on resource occupancy types comes from a predefined test project-resource occupancy relationship matrix; If any resource is used to determine the test item and If a conflict exists, then it is in the conflict topology graph. Add an undirected edge To the edge of the market In the process, if there are conflicts on multiple resources, only one edge is added, and no duplicate edges are added. After completing the evaluation of all test items, output a complete conflict diagram. The nodes represent test projects, and the edges represent resource conflicts between two projects that cannot be executed in parallel.

[0011] Furthermore, a greedy selection algorithm is adopted, specifically: Using the constructed conflict topology graph Parallel scheduling of test items: Sort all test items in descending order of execution time, and then try to compare each test item with the test items in the current parallel set in turn. If the test item has no edges with any test items in the set in the conflict graph, add it to the current parallel set; otherwise, skip it. The test items in the current parallel set are executed simultaneously. Repeat this process to process the remaining test items until all test items have been scheduled. When the number of test items exceeds a preset threshold, the graph coloring algorithm is switched for parallel scheduling, specifically as follows: Conflict topology diagram Vertex coloring is performed, and adjacent nodes connected by edges cannot be colored with the same color. Each color represents a time period for parallel execution. The minimum number of colors is the minimum number of parallel batches required. Items of the same color are executed simultaneously. Based on the scheduling results, calculate the total completion time of all test items within the test station segment. For the greedy parallel scheduling algorithm, the total completion time is equal to the sum of the durations of the longest test items in each parallel batch; for the graph coloring algorithm, the total completion time is equal to the sum of the durations of the longest test items in each batch corresponding to each color. Output the start and end times of each test item as the specific execution plan for the test items within the test station segment.

[0012] Furthermore, the target test station segment is determined according to the updated allocation priority score. Specifically, after the micro execution layer completes the parallel scheduling of test items within each test station segment, it collects the actual completion time, fixed test duration, actual resource utilization rate, and resource conflict frequency of the corresponding test station segment, and reports the feedback data along with the test station segment number to the macro strategy layer. The macro strategy layer dynamically updates the allocation priority of the test station segment based on the actual resource utilization rate, sets an allocation priority score for each test station segment, and reduces the allocation priority score of the test station segment when the actual resource utilization rate of the test station segment is lower than the preset utilization rate threshold twice in a row, and restores the allocation priority score of the test station segment to the initial value when it is higher than the utilization rate threshold twice in a row. In the subsequent allocation of test sites for satellites, the target test site will be determined according to the updated allocation priority score.

[0013] Furthermore, the macro-strategy layer's scheduling scheme and the micro-execution layer's parallel scheduling strategy are output. Specifically, the macro-strategy layer dynamically updates the fixed test duration of the test station segment based on the resource conflict frequency and the actual completion time. When the resource conflict frequency of the test station segment exceeds the preset conflict threshold twice in a row, a time buffer factor is introduced for the test station segment, and the fixed test duration is updated based on the time buffer factor. The time buffer factor increases in a preset step size and does not exceed a preset upper limit. When the proportion of test station segments whose actual completion time exceeds the fixed test duration multiple times reaches a preset threshold, the fixed test duration is updated based on the actual average completion time of the test station segment. After each satellite is scheduled, the macro strategy layer updates the allocation priority and fixed test duration of the test workstation based on the latest feedback data, and outputs the scheduling scheme of the macro strategy layer and the parallel scheduling strategy of the micro execution layer after reaching the preset maximum number of updates.

[0014] The present invention also provides a scheduling system for rapid testing of batch satellites, the scheduling system for rapid testing of batch satellites being used to execute the above-described scheduling method for rapid testing of batch satellites, comprising: Data acquisition module: For each satellite in the batch, the entire testing process of the satellite is broken down into multiple sequentially connected test stations according to function. Each test station contains several test items, and a discrete event simulation model is established based on the test items. The macro-scheduling calculation module is used to construct a dual-objective mixed integer programming model based on the discrete event simulation model, with the shortest total completion cycle of batch satellite parallel operations as the first optimization objective and the minimization of resource load balance among each test station segment as the second optimization objective. The dual-objective mixed integer programming model is solved by a heuristic scheduling algorithm based on constraint satisfaction, and the scheduling scheme of the macro-strategy layer is selected by Pareto non-dominated sorting. Local scheduling calculation module: After completing the scheduling scheme of the macro strategy layer, for each test station segment, based on the corresponding test items, a conflict topology graph between test items based on the resource occupation conflict relationship is constructed. According to the conflict topology graph, a greedy selection algorithm is used to schedule the test items in parallel. When the test items in any test station segment exceed the preset threshold, the graph coloring algorithm is switched to perform parallel scheduling. The strategy output module is used to optimize the parallel scheduling based on the results of the parallel scheduling, with the goal of minimizing the completion time of all test items in each test station segment, and generate a parallel scheduling strategy at the micro-execution layer. After completing the parallel scheduling strategy at the micro-execution layer, feedback data of the actual execution of each test station segment is collected. The macro-strategy layer dynamically updates the allocation priority and test duration parameters of the test station segment based on the feedback data. The above update process is repeated until the preset maximum number of updates is reached, and the final scheduling strategy is output.

[0015] Compared with the prior art, the beneficial effects of the present invention are: This solution breaks down the entire satellite testing process into sequentially connected test workstations and establishes a dual-objective mixed-integer programming model at the macro-strategy level and a conflict topology graph model at the micro-execution level, forming a two-layer collaborative scheduling architecture of "macro-scheduling + micro-resolution". It optimizes pipeline scheduling among multiple satellites with the goal of minimizing the total completion cycle and resource load balancing, and optimizes the parallel execution order within each workstation with the goal of minimizing the completion time of test items, effectively solving the problem of the separation between macro-scheduling and micro-scheduling in existing methods. At the macro-strategy level, a constraint-based heuristic scheduling algorithm and Pareto non-dominated sorting are used to generate multiple Pareto-optimal satellite-workstation allocation schemes for decision-makers to choose from. At the micro-execution level, greedy selection and graph coloring algorithms are dynamically switched according to the number of test items, balancing scheduling efficiency and conflict resolution quality, significantly improving the parallel utilization rate of resources within the test workstations. Attached Figure Description

[0016] Figure 1 This is a schematic diagram of the overall method flow of the present invention; Figure 2 A graph showing the relationship between the updated fixed test duration and the number of resource conflicts; Figure 3 This is a schematic diagram of the overall system structure of the present invention. Detailed Implementation

[0017] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to specific embodiments.

[0018] It should be noted that, unless otherwise defined, the technical or scientific terms used in this invention should have the ordinary meaning understood by one of ordinary skill in the art to which this invention pertains. The terms "first," "second," and similar terms used in this invention do not indicate any order, quantity, or importance, but are merely used to distinguish different components. Terms such as "comprising" or "including" mean that the element or object preceding the word encompasses the elements or objects listed following the word and their equivalents, without excluding other elements or objects. Terms such as "connected" or "linked" are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect. Terms such as "upper," "lower," "left," and "right" are used only to indicate relative positional relationships; when the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0019] Example: Please see Figures 1-2 The present invention provides a technical solution: The scheduling method for rapid testing of batch satellites includes the following steps: S1: For each satellite in the batch of satellites, the entire testing process of the satellite is broken down into multiple test stations connected in sequence according to function. Each test station contains several test items, and a discrete event simulation model is established based on the test items. In this embodiment, the established discrete event simulation model is specifically as follows: In the discrete event simulation model, for each test station segment, resource status variables and project status variables are set based on the test project. The resource status variables are used to characterize the idle state, guaranteed occupancy state, and exclusive occupancy state of resources. The project status variables are used to characterize the unstarted state, in-process state, and completed state of the test project. The executable conditions of the test project are determined according to the guaranteed occupancy or exclusive occupancy requirements of the test project for resources. Test projects that do not meet the executable conditions are put into the waiting queue. In the above process, a shared count variable is set for each resource status variable. This variable is only valid when the resource status variable is 1 and is used to record the number of test projects currently using the resource in a guaranteed occupancy mode. When the shared count variable decreases from 1 to 0, the resource status variable is synchronously restored from 1 to 0. The reason for distinguishing the three values ​​of the "resource status variable" is that different test projects use the same type of resource in different ways in actual satellite testing: multiple test projects can share resources in the "guaranteed occupation" mode at the same time, such as multiple devices reading the status of the same power supply system at the same time, but only one test project can exclusively occupy the resource in the "exclusive occupation" mode, such as a dedicated measuring instrument can only serve one test project at a time; by setting a "shared count variable" for each "resource status variable" and recording the number of test projects currently using the resource in the guaranteed occupation mode when its value is 1, the shareable usage mode of the resource can be accurately managed.

[0020] The conditions for execution are as follows: a test project can only start execution when none of the resources required by the test project are exclusively occupied by other test projects; if any resource is currently exclusively occupied by other test projects, or if the resource is currently guaranteed but this test project needs to exclusively occupy it, the conditions for execution are not met, and the test project enters the waiting queue. In the above process, the judgment logic of the "executability condition" stipulates that "a test project can only start execution when every resource required by the test project is not exclusively occupied by other test projects." It also explicitly states that the condition is not met when "the resource is currently under guaranteed occupation but this test project needs exclusive occupation." The core reason for this design is that exclusive occupation is exclusive; once a resource is exclusively occupied by a test project, all other test projects must wait, regardless of their demand for that resource. Guaranteed occupation, on the other hand, allows multiple projects to share the resource, but when a project requiring exclusive occupation arrives, even if the resource is currently under guaranteed occupation (i.e., multiple projects are sharing it), that project must enter the "waiting queue" and cannot forcibly preempt it. This design ensures that exclusive projects have complete control over the resource while allowing multiple guaranteed projects to execute concurrently, thus achieving a balance between resource security and parallel efficiency.

[0021] The start events of the test items are added to the event queue according to the preset scheduling order, and the simulation clock is advanced according to the event occurrence time. When a test project meets the executable conditions, it occupies the corresponding resources and generates a completion event. When the test project is completed, it releases the corresponding resources and triggers the execution of test projects in the waiting queue that meet the executable conditions. Once all test items within a test station segment are completed and related resources are released, the completion time of that test station segment is output, and the corresponding satellite is allowed to enter the next test station segment.

[0022] In the above process, all state variables are set to zero, the start events of all test items are placed into the event queue according to the preset scheduling order, and the simulation clock is set to zero; the following steps are executed repeatedly until the event queue is empty and all test items are marked as completed: Take the earliest occurring event from the head of the event queue and jump the simulation clock to the time when that event occurred; If the event is a start event, read the current state variable, check the executable conditions, and if the executable conditions are met, perform the corresponding update based on the resource usage type of the current test project: If it is an exclusive occupation, then change the corresponding resource status variable to an exclusive occupation status; If it is a guaranteed occupancy, then increment the shared count variable of the corresponding resource status by 1; Then calculate the completion time, generate the corresponding completion event, and put it back into the event queue; if the conditions are not met, put the test item into the waiting queue. If the event is a completion event: change the status variable of the exclusively occupied resource back to the idle state, or decrement the shared count variable of the guaranteed resource status variable by 1: if it is zero, change the resource status variable back to the idle state; if it is still greater than 0, keep the resource status variable at 1, immediately traverse the waiting queue, remove all test items that meet the executable conditions from the waiting queue, generate a new start event and put it back into the event queue, and set the time to the current clock. When the above loop ends, the current simulation clock is output as the total completion time. When all test items in the test station segment are completed, all resource status variables are 0, and all shared count variables are 0, the satellite corresponding to the test station segment is allowed to enter the next test station segment.

[0023] The entire simulation loop employs an event-driven mechanism, jumping directly to the next event's occurrence time instead of progressing step-by-step with fixed time increments. This is because the execution time of test items is typically much longer than the time granularity of computer simulation, and the event-driven approach significantly reduces unnecessary state checks. For a "start event," the model determines if the current resource state meets the "executable conditions." If it does, it immediately updates the resource state, either changing exclusive resources to exclusively occupied or incrementing the "shared count variable" of guaranteed resources and generating a corresponding "completion event" to be returned to the event queue. If not, the test item is placed in the "waiting queue." For a "completion event," the model releases the corresponding resource, either changing the state variable of exclusively occupied resources back to idle or decrementing the "shared count variable" of guaranteed resources, reverting it to idle when it reaches zero. Then, it immediately iterates through the "waiting queue," removing all test items that currently meet the executable conditions and generating a new "start event," with the time set to the current clock. This design ensures that once resources are released, items in the waiting queue are activated immediately, avoiding idle waiting.

[0024] When the above loop ends, the current simulation clock is output as the total completion time. When all test items in the test station segment are completed, all resource status variables are 0, and all shared count variables are 0, the satellite corresponding to the test station segment is allowed to enter the next test station segment.

[0025] In the above process, the satellite is only allowed to move to the next test station when "all test items within the test station segment have been completed, all resource status variables are 0, and all shared count variables are 0". This condition is set because: non-zero resource status variables or non-zero shared count variables mean that there are still test items being executed or resources still being occupied. Allowing the satellite to move to the next test station segment in this case would lead to resource crosstalk and status confusion between different test stations. Only by ensuring that all resources in the current test station segment have been completely released and all test items have been completed can the satellite safely move to the next test station segment.

[0026] S2: Using the discrete event simulation model as the operational support, with the shortest total completion cycle of batch satellite parallel operations as the first optimization objective and the minimization of resource load balance among each test station segment as the second optimization objective, a dual-objective mixed integer programming model is constructed; the dual-objective mixed integer programming model is solved using a constraint-based heuristic scheduling algorithm, and a scheduling scheme for the macro-strategy layer is selected through Pareto non-dominated sorting; In this embodiment, the dual-objective mixed integer programming model takes the shortest total completion cycle of batch satellite parallel operations as the primary optimization objective and the optimal resource load balance between test workstations as the secondary optimization objective. The total completion cycle is characterized by the maximum completion time among all satellite completion times, and the resource load balance is characterized by the deviation of the utilization rate of each test station segment under each test phase from the average utilization rate. The smaller the deviation, the more balanced the resource load among the test station segments. The utilization rate of the test station segment is determined based on the actual usage time of the corresponding test station segment and the total completion cycle; and the test duration of each satellite in each test stage is a preset fixed duration, and each test stage is executed sequentially according to a predetermined order, with test station segments with the same function in the same test stage being used in parallel.

[0027] The formula upon which the above process is based is: in, Indicates the first The first test phase The utilization rate of each test station segment is obtained by dividing the actual usage time of the test station segment by the total completion cycle. This represents the average utilization rate of all test workstations, i.e., all The arithmetic mean; Indicates the first The completion time of testing for each satellite; This indicates the total number of testing phases. Represented as the first Number of test workstations in each test phase; Indicates the satellite number index; Indicates the total completion period; Indicates the resource load balance between test workstations; Indicates the number of test station segment indices; Indicates the number of indexes during the testing phase; In the above process, the testing duration of each satellite in each testing phase is fixed, that is, fixed testing duration. The testing order of the satellites is arranged and does not overlap. Each satellite uses each testing station independently and without interference. Testing stations with the same function in the same testing phase are used in parallel.

[0028] In the above process, the core purpose of designing the dual-objective mixed-integer programming model is to simultaneously optimize two conflicting objectives at the macro-scheduling level of batch satellite parallel testing. The first objective... To minimize the "total completion time," which is the longest test completion time among all satellites, directly reflects the overall testing efficiency of the batch of satellites and is a core constraint on production cycle time and delivery schedule. The second objective... To minimize the "resource load balancing between test workstations", i.e., the utilization rate of each "test workstation". With average utilization rate The sum of the variances between them serves to avoid the imbalance where some workstations are overloaded while others are idle. Uneven resource load can prolong the overall completion cycle of bottleneck workstations and waste equipment resources. The reason for placing both objectives in the same model and employing a primary / secondary hierarchical optimization logic is that, in satellite mass production testing scenarios, the overall completion cycle is usually the decision-maker's top priority, while resource load balancing is a secondary objective that further optimizes the overall completion cycle while ensuring it is as short as possible. In the definition of the objective function, the utilization rate of the "test station section" The method used is "actual usage time divided by total completion time" rather than divided by the available time of that workstation segment itself, because "total completion time" is... It is a unified time benchmark for all testing stations. Specifically, for the first... The first "test phase" The actual usage time of each "test station segment" refers to the total time occupied by that station segment, which is the sum of the testing time of multiple satellites in that station segment, divided by the overall "total completion cycle". , obtained This reflects the workload of the workstation segment throughout the entire testing cycle. If the total available time of each workstation segment is used as the denominator, the utilization rates of different workstation segments will become incomparable due to differences in configuration quantity and available time, making it impossible to accurately measure whether the load is balanced. Therefore, a unified approach using... As the denominator, all At the same scale, That is, the arithmetic mean of the utilization rates of all workstations, thus The variance calculation can accurately reflect the load differences between each workstation section; The model establishment process needs to satisfy several constraints, which are derived from the actual engineering rules of satellite AIT testing. "The testing duration for each satellite in each testing phase is fixed, i.e., fixed testing duration" means that the testing time for the same satellite in different test workstations within the same testing phase is a known constant, unaffected by scheduling schemes. This simplifies the problem's complexity, as the scheduling of specific test items within a workstation is handled at the micro-execution level, and the macro-strategy level only needs to focus on workstation-level scheduling. "The satellite testing sequence is arranged and does not overlap" ensures that each satellite can only be assigned to one "test workstation" in the same testing phase, and testing of different satellites in the same workstation cannot be performed simultaneously. That is, when any two satellites are assigned to the same workstation, their testing time intervals are not... The ability to overlap aligns with the exclusivity constraints of physical resources, namely test equipment, operators, and cable connections. "Each satellite's use of each test station segment is independent and does not interfere with each other" means that different satellites can simultaneously and independently execute on parallel "test stations" with the same function without waiting for each other, demonstrating the advantages of parallel pipelines. "Parallel use of test stations with the same function in the same test phase" is a core feature of rapid batch satellite testing. This means that each "test phase" is configured with multiple "test stations" with the same function that can be run in parallel. Multiple satellites can be simultaneously assigned to different parallel test stations to execute the same stage's test tasks, thereby significantly shortening the overall test cycle for batch satellites.

[0029] The candidate scheduling strategies under the corresponding rule combinations are obtained as follows: when solving the bi-objective mixed integer programming model using a heuristic scheduling algorithm based on constraint satisfaction, satellite sorting rules and test station allocation rules are pre-set, and the satellite sorting rules and test station allocation rules are combined to form multiple rule combinations. For each rule combination, the satellites are processed sequentially according to the order of each test phase of the satellite AIT integrated test. In the current test phase, the satellites to be scheduled are sorted according to the corresponding satellite sorting rules, and the sorted satellites are sequentially assigned to the parallel test workstations with the same function in the current test phase according to the corresponding test workstation allocation rules. The satellite sorting rules include sorting according to preset priority and planned launch time, and the test workstation allocation rules include prioritizing the earliest available test workstation. The start time of each satellite in the current test phase is determined by the completion time of its previous test phase and the available time of the assigned test workstation segment. After the allocation is completed, the available time of the corresponding test workstation segment is updated. After all test phases are processed, candidate scheduling strategies under the corresponding rule combinations are obtained.

[0030] The preset priority is a ranking weight parameter pre-determined based on the satellite's engineering mission attributes and delivery constraints. It is used to make ranking decisions for different satellites in the heuristic scheduling algorithm at the macro-strategy level. Specifically, the "preset priority" can be determined based on one or more of the following factors: the planned launch time of the satellite (the earlier the launch time, the higher the priority); the urgency of the satellite's delivery in the batch (urgent tasks are given higher priority); the complexity of the satellite testing process (satellites with complex processes are prioritized to avoid becoming bottlenecks); and the scarcity of satellite resources (satellites requiring special resources are prioritized to match the available resource window). The above-described process of solving the bi-objective mixed-integer programming model using a constraint-satisfaction-based heuristic scheduling algorithm is designed to generate candidate scheduling schemes at the macro-level strategy layer while ensuring computational efficiency. Since the bi-objective mixed-integer programming model is an NP-hard problem, when the number of satellites... Total number of testing phases and the number of test stations at each stage When the size is large, precise solutions will face a combinatorial explosion problem, making it impossible to obtain the optimal solution within an engineering-acceptable timeframe. Therefore, this invention employs a heuristic scheduling algorithm, which compresses the infinite solution space into a small number of computable candidate scheduling strategies by enumerating a finite number of engineering-meaning combinations of "sorting rules" and "allocation rules." The "sorting rules" include "highest preset priority" and "earliest planned launch time priority," which directly correspond to the actual business logic in satellite testing, such as prioritizing urgent tasks and satellites with earlier delivery nodes. The "allocation rule" is set to "prioritize the test station segment with the earliest available unit," reflecting the idea of ​​load balancing and avoiding queuing backlog caused by consecutively allocating multiple satellites to the same test station segment. During the generation of candidate scheduling strategies, the algorithm processes them sequentially according to the order of each test phase of the satellite AIT integrated test. This is because the satellite testing process has a natural serial dependency, meaning that a satellite can only be tested after completing the first phase. Only after passing the stage test can one proceed to the next stage. Phase. For the current testing phase, the satellites to be scheduled are first sorted according to the "sorting rules" in the current rule combination. The sorted order determines the priority order in which satellites obtain resources in this phase. Then, "test workstations" are allocated to each satellite in the sorted order, following the "allocation rules" in the current rule combination. The start time of a satellite in this phase is the greater of the "completion time of the previous phase" and the "available time of the selected test workstation," reflecting two constraints: first, the satellite must complete all tests in the previous phase before starting this phase; second, the satellite must wait for the selected test workstation to become available before it can begin using it. Within the same testing phase, multiple "test workstations" can be used in parallel, and different satellites can be simultaneously assigned to different test workstations for independent execution. This is the core parallel mechanism for rapid batch testing of satellites. For example, if the first... The phase is configured with three parallel test workstations with identical functions, allowing up to three satellites to be tested in parallel during this phase, which significantly shortens the overall completion cycle of the entire batch of satellites. Each time a combination of rules is completed—that is, a pairing of a "sorting rule" and an "allocation rule"—a candidate scheduling strategy is obtained.

[0031] The scheduling scheme obtained at the macro-level strategy layer is as follows: Each candidate scheduling strategy is evaluated to obtain the corresponding total completion time and resource load balance of the test station segment. The total completion time is characterized by the maximum completion time among all satellite completion times, and the resource load balance is characterized by the deviation of the utilization rate of each test station segment from the average utilization rate. Pareto non-dominance screening is performed based on the total completion time and resource load balance of each candidate scheduling strategy. The candidate scheduling strategy that is not dominated by other candidate scheduling strategies in both total completion time and resource load balance is selected as the scheduling scheme of the macro strategy layer.

[0032] In the above process, for each candidate scheduling strategy, the maximum completion time among all satellites is calculated. Simultaneously, the utilization rate of each unit is obtained by dividing the actual usage time of each test station segment by the total completion cycle. Then, the load balancing degree of the utilization of all units is calculated to obtain Each candidate scheduling strategy corresponds to a set value; In the above process, Pareto filtering is performed on the candidate scheduling strategies. Specifically, all candidate scheduling strategies are treated as points in a two-dimensional target space, with the horizontal coordinate being... The vertical axis is For any two candidate scheduling policies, if candidate scheduling policy A's Candidate scheduling strategy B And candidate scheduling strategy A Candidate scheduling strategy B If at least one of the two strict inequalities holds, then candidate scheduling strategy A dominates candidate scheduling strategy B. After traversing all candidate scheduling strategies, the candidate scheduling strategy that is not dominated by any other candidate scheduling strategy is the candidate scheduling strategy on the Pareto front, which is marked as the Pareto better strategy and serves as the scheduling scheme at the macro-strategy level.

[0033] The core objective of Pareto screening of candidate scheduling strategies in a multi-objective optimization scenario is to identify the non-dominated, better strategy from a finite number of candidate scheduling strategies. Since the macro-level strategy layer of this invention simultaneously optimizes two objectives—total completion time—[further details are needed]. Resource load balancing between test workstations Typically, there is no candidate scheduling strategy that makes and It is superior to all other strategies, so the Pareto non-dominated sorting method is needed to screen out candidate scheduling strategies on the Pareto front. By traversing all candidate scheduling policies, those policies that are not dominated by any other candidate scheduling policies are the candidate scheduling policies on the Pareto front. These Pareto front policies have the following property: they cannot deteriorate without being dominated by any other candidate scheduling policy. Further improvements under the premise It is also impossible to prevent the situation from worsening. Further improvements under the premise In other words, these strategies constitute the optimal trade-off boundary between the two optimization objectives. Labeling these strategies as Pareto-optimal strategies and outputting them as scheduling schemes at the macro-level strategy layer has engineering significance: decision-makers can flexibly select a scheduling scheme from the Pareto front based on the current task's focus. For example, when satellite delivery nodes are urgent, a scheduling scheme from the Pareto front can be selected. Smaller options should prioritize minimizing the overall completion time; when testing resources are limited and need to be used evenly, the Pareto front can be selected. Smaller options prioritize minimizing resource load balancing. This Pareto selection method does not introduce any subjective weighting coefficients, maintaining the objectivity of the decision-making process. At the same time, it avoids the tedious process of repeatedly adjusting weighting coefficients when transforming a bi-objective problem into a single-objective problem through weighted summation.

[0034] S3: After completing the scheduling scheme of the macro strategy layer, for each test station segment, based on the corresponding test items, construct a conflict topology graph between test items based on the resource occupation conflict relationship. According to the conflict topology graph, the greedy selection algorithm is used to schedule the test items in parallel. When the test items in any test station segment exceed the preset threshold, switch to the graph coloring algorithm for parallel scheduling. In this embodiment, the established conflict topology graph is specifically as follows: Conflict topology diagram between test items established in the micro-execution layer ,in For testing project nodes, For conflict edges, edge Indicates test items and There are resource conflicts, so parallel execution is not possible; for any two different test projects and Iterate through all resources and check for conflicts according to the following conflict determination rules: If two test projects occupy the same resource, there will be a conflict if at least one of them is an exclusive occupy; there will be no conflict only if both are guaranteed occupy. The data on resource occupancy types comes from a predefined test project-resource occupancy relationship matrix; If any resource is used to determine the test item and If a conflict exists, then in the conflict topology graph Add an undirected edge To the edge of the market In the process, if there are conflicts on multiple resources, only one edge is added, and no duplicate edges are added. After completing the evaluation of all test items, output a complete conflict diagram. The nodes represent test projects, and the edges represent resource conflicts between two projects that cannot be executed in parallel.

[0035] In the aforementioned process, the core purpose of constructing the conflict topology graph is to abstract the resource conflict relationships between multiple test items within the test station section into a graph model, providing input for subsequent greedy selection algorithms or graph coloring algorithms. Unlike the traditional approach of simply treating all resource conflicts as mutually exclusive, this invention refines the determination of conflicts based on different resource occupancy types. Specifically, a "test item-resource occupancy relationship matrix" is predefined, which clarifies the occupancy type of each test item for each type of resource. The distinction between "guaranteed occupancy" and "exclusive occupancy" is introduced because, in actual satellite testing, different test items use the same type of resource in fundamentally different ways: multiple test items can simultaneously share resources using a "guaranteed occupancy" method (e.g., multiple devices simultaneously reading telemetry data from a power supply device), but only one test item can exclusively occupy resources using a "exclusive occupancy" method (e.g., a dedicated spectrum analyzer can only serve one test item at a time). The determination of the "Test Item-Resource Occupation Relationship Matrix" needs to be predefined based on the technical requirements and resource usage characteristics of each test item within the test station segment. Specifically, for each test station segment, firstly, all resource types involved in the station segment are identified, including test instruments and equipment such as spectrum analyzers, signal generators, multimeters, etc.; power supply equipment such as energy security for power supply and distribution systems; measurement and control subsystem equipment such as measurement and control communication link security; data management subsystem equipment such as ground information flow analysis equipment; supporting test cables; and operators, etc. Then, the usage of each type of resource by each test item within the test station segment during execution is analyzed item by item. Based on the resource usage, the occupancy type is divided into three categories: if a test item does not need to use a certain type of resource at all, it is marked as "not occupied"; if a test item only needs a certain type of resource for operational security, such as power supply security or communication link security, and does not directly participate in functional testing, it is marked as "security-type occupancy"; and if a test item needs a certain type of resource to directly participate in functional testing, it is marked as "exclusive occupancy". The above correspondence is usually sorted out and calibrated item by item by the process engineer or test designer of satellite AIT integrated testing according to the test details, equipment operation manual and resource allocation plan of each test item. Finally, a two-dimensional matrix is ​​formed with test items as rows and resource types as columns. Each element in the matrix takes the value of 0 (no occupation), 1 (guaranteed occupation) or 2 (exclusive occupation). This matrix is ​​predefined and stored before scheduling begins. Conflict topology graphs possess the following important properties: if two test items are adjacent (connected by an edge) in the conflict graph, they cannot be executed simultaneously in the same parallel batch; conversely, if two test items are not adjacent, there is no resource conflict between them, and theoretically, they can be executed in parallel. By transforming the resource conflict problem into a graph model, the originally complex resource conflict detection problem is simplified to an independent set or graph coloring problem in graph theory, laying the mathematical model foundation for subsequent parallel scheduling using greedy selection algorithms or graph coloring algorithms.

[0036] A greedy selection algorithm is used, specifically: Using the constructed conflict topology graph Parallel scheduling of test items: Sort all test items in descending order of execution time, and then try to compare each test item with the test items in the current parallel set in turn. If the test item has no edges with any test items in the set in the conflict graph, add it to the current parallel set; otherwise, skip it. The test items in the current parallel set are executed simultaneously. Repeat this process to process the remaining test items until all test items have been scheduled. When the number of test items exceeds a preset threshold, the graph coloring algorithm is switched for parallel scheduling, specifically as follows: Conflict topology diagram Vertex coloring is performed, and adjacent nodes connected by edges cannot be colored with the same color. Each color represents a time period for parallel execution. The minimum number of colors is the minimum number of parallel batches required. Items of the same color are executed simultaneously. Based on the scheduling results, calculate the total completion time of all test items within the test station segment. For the greedy parallel scheduling algorithm, the total completion time is equal to the sum of the durations of the longest test items in each parallel batch; for the graph coloring algorithm, the total completion time is equal to the sum of the durations of the longest test items in each batch corresponding to each color. Output the start and end times of each test item as the specific execution plan for the test items within the test station segment.

[0037] In the above process, the hybrid parallel scheduling strategy combining the greedy selection algorithm and the graph coloring algorithm aims to efficiently solve the parallel scheduling scheme of test items within the test station section at the micro-execution level. Since the number of test items within the test station section is typically on the order of tens, the conflict topology graph... The problem of minimizing the total completion time is essentially a weighted graph coloring or parallel scheduling problem, which is an NP-hard problem and cannot be solved exactly in polynomial time. A greedy selection algorithm is used as the default scheduling method, and the graph coloring algorithm is switched when the number of test items exceeds a preset threshold to balance computational efficiency and scheduling quality. The advantages of this hybrid scheduling strategy are as follows: when the number of test items is small, the greedy selection algorithm quickly obtains an approximate optimal solution with extremely low computational overhead; when the number of test items exceeds a preset threshold, the graph coloring algorithm can perform parallel batch partitioning from a global perspective, typically achieving better scheduling results and shorter total completion time than the greedy selection algorithm. The dynamic switching between the two ensures a balance between scheduling quality and computational efficiency at different problem scales. Compared to directly using the graph coloring algorithm for all cases, the hybrid strategy avoids computational waste caused by excessive algorithm complexity when the problem scale is small; compared to always using the greedy selection algorithm, the hybrid strategy provides a mechanism to switch to a better algorithm when the problem scale is large, thereby improving the overall engineering applicability and robustness of the micro-execution layer scheduling. The time complexity of the greedy selection algorithm is O(n). ,in To determine the number of test items, the time complexity of graph coloring algorithms, such as greedy coloring or branch and bound, is typically O(n log n). Even exponentially, when the number of test items is small, the greedy selection algorithm can obtain a near-optimal solution in a very short time, with the deviation from the optimal solution within an acceptable range for engineering. As the number of test items increases, the approximate performance of the greedy selection algorithm may decrease, while the graph coloring algorithm, because it can resolve conflicts and partition batches in a global perspective, can usually obtain fewer parallel batches and a shorter total completion time. Therefore, the "preset threshold value" can be calibrated based on historical test data or offline simulation experiments: in a typical test station section, enumerate different numbers of test items. The scheduling result deviation rates of the greedy selection algorithm and the graph coloring algorithm, as well as their computational time ratios, are compared. The minimum value is selected when the scheduling result deviation rate exceeds the upper limit of the allowable deviation (e.g., 5%) and the computational time ratio is still within the upper limit of the allowable computational time. As the "preset threshold value"; if offline calibration conditions are not available, the "preset threshold value" can also be set to 20 based on engineering experience. That is, when the number of test items in the test station section does not exceed 20, the greedy selection algorithm is used, and when it exceeds 20, the graph coloring algorithm is switched.

[0038] S4: Based on the parallel scheduling results, optimize the process with the goal of minimizing the completion time of all test items within each test station segment, and generate a parallel scheduling strategy for the micro-execution layer. After completing the parallel scheduling strategy for the micro-execution layer, collect feedback data on the actual execution of each test station segment. The macro-strategy layer dynamically updates the allocation priority and test duration parameters of the test station segments based on the feedback data. Repeat the above update process until the preset maximum number of updates is reached, and output the final scheduling strategy.

[0039] The target test station segment is determined according to the updated allocation priority score, specifically as follows: After completing the parallel scheduling of test items in each test station segment, the micro-execution layer collects the actual completion time, fixed test duration, actual resource utilization rate and resource conflict frequency of the corresponding test station segment, and reports the feedback data along with the test station segment number to the macro-strategy layer. The macro strategy layer dynamically updates the allocation priority of the test station segment based on the actual resource utilization rate, sets an allocation priority score for each test station segment, and reduces the allocation priority score of the test station segment when the actual resource utilization rate of the test station segment is lower than the preset utilization rate threshold twice in a row, and restores the allocation priority score of the test station segment to the initial value when it is higher than the utilization rate threshold twice in a row. In the subsequent allocation of test sites for satellites, the target test site will be determined according to the updated allocation priority score.

[0040] In the above process, "actual resource utilization rate" refers to the proportion of time that various resources, such as test instruments, power supply equipment, measurement and control links, and operators, are actually occupied during the test execution after the parallel scheduling of all test items within a test station segment is completed at the micro-execution layer. For exclusive occupied resources, "actual resource utilization rate" equals the cumulative time the resource is in an "exclusive occupied state" divided by the total completion cycle; for guaranteed occupied resources, "actual resource utilization rate" equals the sum of the occupation times of all test items using the resource in a guaranteed manner, i.e., when multiple items share the same resource, the occupation time of the resource will be repeatedly accumulated and divided by the total completion cycle. The magnitude of this parameter directly reflects the resource busyness of the test station segment: a higher "actual resource utilization rate" indicates that the resources of the station segment are fully utilized, but too high a rate may lead to increased resource conflicts; a lower "actual resource utilization rate" indicates that there are idle resources in the station segment, and the macro-strategy layer should reduce its allocation priority to balance the load. The determination of the "preset utilization threshold" needs to be based on the resource utilization efficiency requirements of the test station segment and the adjustment sensitivity of the load balancing at the macro-strategy layer. The utilization threshold is a dimensionless threshold between 0 and 1, representing the utilization rate of a single "test station segment". A value below this threshold indicates that the workstation is relatively idle, while a value above this threshold indicates it is in a normal or busy state. This threshold can be set based on the theoretical utilization rate under ideal load balancing conditions: in a perfectly balanced ideal situation, the utilization rate of each workstation should equal the average utilization rate of all workstations. Therefore, the "preset utilization threshold" can be set to A certain ratio value, for example, set to or This ensures that priority reduction is only triggered when the utilization rate of a workstation is significantly below average, avoiding frequent priority adjustments due to normal fluctuations. When it is impossible to calculate in advance... At that time, the "preset utilization threshold" can be set to a fixed value based on engineering experience, such as 0.3 or 0.4. That is, when the utilization rate of the "test station section" is lower than 30% or 40%, it is considered to be in an idle state, and its allocation priority should be reduced to guide subsequent satellites to other station sections. The output shows the scheduling scheme at the macro-level strategy layer and the parallel scheduling strategy at the micro-level execution layer, specifically: The macro strategy layer dynamically updates the fixed test duration of the test station segment based on the resource conflict frequency and the actual completion time. When the resource conflict frequency of the test station segment exceeds the preset conflict threshold twice in a row, a time buffer factor is introduced for the test station segment, and the fixed test duration is updated based on the time buffer factor. The time buffer factor increases in a preset step size and does not exceed a preset upper limit. When the proportion of test station segments whose actual completion time exceeds the fixed test duration multiple times reaches a preset threshold, the fixed test duration is updated based on the actual average completion time of the test station segment. After each satellite is scheduled, the macro strategy layer updates the allocation priority and fixed test duration of the test workstation based on the latest feedback data, and outputs the scheduling scheme of the macro strategy layer and the parallel scheduling strategy of the micro execution layer after reaching the preset maximum number of updates.

[0041] The formula used in the above process is: in, in, Indicates the first The first test phase The test station section is in the first The buffer factor during secondary resource conflicts, and ; Indicates a preset fixed step size; Indicates the upper limit of the buffer factor; Indicates the first The first test phase The first test station section experienced the [number]th [event / incident]. The updated fixed test duration in the event of a resource conflict; Indicates the first The first test phase Fixed test duration for each test station segment; Indicates the first The first test phase The test station section is in the first Buffer factor during secondary resource conflicts; In the above embodiments, 20 sets of data on the number of resource conflicts and the corresponding updated fixed test duration are provided to reflect the change of the updated fixed test duration as the number of resource conflicts changes, as shown in Table 1: Table 1: Relationship between the number of resource conflicts and the corresponding updated fixed test duration The table above shows that when the number of resource conflicts... When the buffer factor increases from 0 to 10, The updated fixed test duration increases linearly from 1.0 to a maximum of 2.0. The buffer factor increases proportionally from 100 minutes to 200 minutes, reflecting a linear amplification relationship in a product form; when the number of resource conflicts exceeds 10, the buffer factor reaches its upper limit. The updated fixed test duration has stabilized at 200 minutes and will not increase further, reflecting... The constraint effect of a function on its upper bound.

[0042] In the formula, the dependent variable Indicates the first The first test phase The first test station section experienced the [number]th [event / incident]. The updated fixed test duration after a series of consecutive resource conflicts exceed the threshold. The magnitude of this value directly determines the length of the time window allocated to this workstation segment by the macro-strategy layer in subsequent scheduling. The larger the value, the more lenient the testing time budget for that workstation segment, allowing for more resource buffering at the micro-execution level; conversely, a smaller value indicates a more flexible testing time budget. The smaller the value, the tighter the scheduling time for that workstation segment, and the higher the requirement for the parallel scheduling efficiency of the micro-execution layer. Influence The independent variables include the original fixed test duration. Buffer factor of the previous state Fixed step size and the upper limit of the buffer factor From a physical perspective: when a test station experiences high resource conflict frequency twice consecutively, it indicates that the resource competition among test items within that station has exceeded the capacity of the micro-scheduling algorithm. Therefore, it is necessary to increase the time budget at a macro-level. The duration should be increased incrementally based on the original duration. The formula uses a recursive cumulative form. This is because of the cumulative effect of resource conflicts—the more frequent the conflicts, the greater the time compensation required; at the same time, the introduction of... Functions and Upper Limits The constraint is that the time magnification factor cannot increase indefinitely; it must be limited by the physical time limits of the satellite testing process, such as the maximum continuous operating time of equipment and personnel shift cycles, to avoid excessive magnification leading to an overly conservative scheduling scheme. Furthermore, and The relationship is directly proportional; that is, the longer the original test duration of a workstation, the greater the absolute time loss caused by the conflict, and the greater the time compensation required. The formula uses a product form. Non-additive form The rationale is that the time loss caused by resource conflicts is usually positively correlated with the execution time of the test project itself—the longer the test time, the more significant the blocking effect of the waiting queue after a conflict occurs, and the required time compensation ratio should be relatively stable rather than an absolute increment. (Recursive formula) This reflects a gradual mechanism where the buffer factor increases linearly with the number of conflicts, avoiding scheduling oscillations caused by large jumps in a single conflict; and The initial value setting ensures that, in the absence of resource conflicts, the updated fixed test duration is equal to the original fixed test duration, without affecting normal scheduling; The determination of the "preset conflict threshold" needs to be based on the acceptability of resource conflicts within the test workstation and the fault tolerance capability of the micro-execution layer scheduling algorithm. The conflict threshold is a dimensionless frequency threshold, representing the upper limit of the allowed number of resource conflicts occurring within a unit of test time or a single test workstation. This threshold can be calibrated based on historical operating data of typical test workstations: statistically analyze the average resource conflict frequency of the workstation under the condition of no scheduling anomalies, and multiply it by a safety factor, such as 1.5 to 2.0, to obtain the conflict threshold; alternatively, it can be set directly as a fixed value based on engineering experience, such as setting the conflict threshold to 3 times / workstation or 5 times / workstation, that is, if the number of resource conflicts exceeds this value during the completion of all test items in each test workstation, it is considered abnormal; when historical data is lacking and cannot be estimated in advance, the conflict threshold can be set to 10% to 20% of the total number of test items, because the more test items there are and the more intensive the resource reuse, the higher the upper limit of the allowed number of conflicts can be. The requirement that "the resource conflict frequency of a test station segment exceeds the preset conflict threshold twice consecutively" triggers the operation of increasing the buffer factor. This design aims to prevent the buffer factor from being mistakenly amplified due to single, occasional conflict fluctuations, thus ensuring the stability of scheduling parameters. Resource conflict frequency may be affected by various accidental factors, such as the special resource requirements of a satellite's test project or temporary performance fluctuations in a batch of equipment. A single instance exceeding the conflict threshold is insufficient to prove a systemic resource contention problem in that test station segment. Only when the reported conflict frequency exceeds the conflict threshold twice consecutively does it indicate that the test station segment does indeed have a persistent resource bottleneck under the current scheduling strategy. In this case, increasing the "buffer factor" amplifies the "fixed test duration," thereby allocating a more relaxed time window for that test station segment at the macro-strategy level. This allows subsequent satellites executing on that test station segment to accommodate more resource waiting time, reducing cascading delays caused by resource conflicts. "Preset fixed step size" Determining the buffer factor requires balancing its sensitivity to resource conflicts with the stability of scheduling parameters; if If the value is too large, the "fixed test duration" after a single conflict trigger will be significantly amplified, potentially leading to excessive time window reservations at the macro-strategy level, resulting in idle resources and unnecessary extension of the overall completion cycle; if If the value is too small, multiple conflict triggers are required for the buffer factor to accumulate to a level that effectively alleviates resource contention, resulting in a delayed response and the scheduling scheme failing to adjust in time during periods of continuous conflict. Therefore, the "preset fixed step size" can be calibrated based on the average time loss ratio caused by a single resource conflict in a typical test workstation segment: statistically analyze the historical data to determine the average percentage of actual completion time exceeding the "fixed test duration" when resource conflicts occur, and then calculate this percentage... or As The value can be determined based on engineering experience when historical data is lacking. Set it to between 0.1 and 0.2, meaning that each trigger will increase the "fixed test duration" by 10% to 20%. "Buffer factor upper limit" The determination of the "fixed test duration" needs to prevent it from being infinitely amplified due to repeated resource conflicts, which would lead to an overly conservative scheduling scheme and a total completion cycle that deviates significantly from the actual requirements. This reflects the maximum test duration amplification factor that the macro strategy layer can tolerate for this test station segment due to resource conflicts. Its upper limit is determined by actual engineering constraints: on the one hand, The physical time limit for that "test station segment" in the satellite testing process must not be exceeded, such as the maximum continuous working time of the equipment and the personnel shift cycle; on the other hand, Required and total completion period In line with the optimization objectives, excessively large This could result in the workstation segment being allocated an excessively large time window in the macro-scheduling, potentially becoming a new bottleneck. The "buffer factor upper limit" can be set differently based on the type and importance of the test workstation segment: for bottleneck workstation segments with intense resource competition and difficult to optimize through scheduling, the upper limit can be set... Set to 1.5 to 2.0; for non-bottleneck workstations where resource contention can be effectively mitigated through scheduling, the [performance / resource allocation] can be [adjusted / reduced]. Set to 1.2 to 1.5; when it is impossible to determine in advance, it can be adjusted based on engineering experience. The value is uniformly set to 1.5, meaning that the "fixed test duration" can be amplified to a maximum of 1.5 times the original value.

[0043] When the proportion of test station segments whose actual completion time exceeds the fixed test duration multiple times reaches a preset threshold, the fixed test duration is updated based on the actual average completion time of the test station segment. After each satellite is scheduled, the macro strategy layer updates the allocation priority and fixed test duration of the test workstation based on the latest feedback data, and outputs the scheduling scheme of the macro strategy layer and the parallel scheduling strategy of the micro execution layer after reaching the preset maximum number of updates.

[0044] The process described above, where the macro-strategy layer updates based on feedback data, aims to establish a closed-loop dynamic feedback mechanism from the micro-execution layer to the macro-strategy layer, enabling the scheduling strategy to adaptively adjust based on actual execution. Traditional static scheduling methods typically employ a "one-time planning, fixed execution" model. However, due to factors such as resource conflict fluctuations and uncertainties in test item duration during actual testing, there is often a significant deviation between the theoretical scheduling scheme and the actual execution effect. Therefore, a dynamic update mechanism based on feedback data is designed. After completing the parallel scheduling of test items within each "test workstation segment," the micro-execution layer collects feedback data such as the "actual completion time," "fixed test duration," "actual resource utilization rate," and "resource conflict frequency" for that workstation segment. After standardization, this data is appended with the "test workstation segment number" and reported to the macro-strategy layer. The macro-strategy layer dynamically updates based on this feedback data from two dimensions: priority adjustment and test duration correction. It also avoids erroneous adjustments caused by single, occasional fluctuations through continuous cumulative judgment from multiple satellites, ultimately outputting an optimized scheduling strategy after multiple iterations. Regarding priority adjustment, if a "test workstation segment's actual resource utilization rate" is lower than the preset utilization rate threshold during testing with two consecutive satellites, it indicates that the workstation segment is relatively idle under the current scheduling strategy. Continuing to allocate satellites with a higher priority would exacerbate the macro-level load imbalance. Therefore, the macro-strategy layer sets a "priority score" for this "test workstation segment," with an initial value of 1. When the utilization rate is reported below the threshold twice consecutively, its "priority score" is halved, thereby reducing the competitive priority of this workstation segment in subsequent satellite allocations. When the utilization rate is reported above the threshold twice consecutively, it indicates that the workstation segment's load is normal or excessive, and its "priority score" is restored to the initial value of 1. When allocating "test workstation segments" for subsequent satellites, all workstation segments are sorted from highest to lowest "priority score," and the workstation segment with the highest priority score is directly selected. If multiple workstation segments have the same priority score, the workstation segment with the smallest number is selected. This priority-based allocation strategy is simple to implement and has low computational overhead. It can guide satellites to move from high-load workstations to low-load workstations, thereby promoting continuous improvement in the "resource load balance between test workstations" at the macro level. Regarding test duration correction, this solution designs update mechanisms for two typical scenarios. The first scenario is an excessively high "resource conflict frequency": if the "resource conflict frequency" of a certain "test workstation" exceeds the preset conflict threshold twice consecutively, it indicates that resource competition among test items within that workstation is intense. At the micro-execution layer, a large number of test items enter the waiting queue due to resource waiting, causing the actual completion time to significantly exceed the theoretical "fixed test duration." Therefore, the macro-strategy layer introduces a "buffer factor" for this workstation. Update "Fixed test duration" to ,in The meaning of this recursive formula is: each time two consecutive conflicts exceed the threshold, the "buffer factor" increases by a fixed step size based on the previous one. Until the upper limit is reached. ;initial This indicates the baseline testing duration when no conflict occurs. In this way, workstations with frequent resource conflicts are allocated longer testing duration budgets, thus allowing for more flexible time windows for subsequent satellites in the macro-scheduling process, alleviating resource competition pressure. The second scenario is when the "actual completion time" is consistently too long: If the percentage of "actual completion time" exceeding the "fixed test duration" for a certain "test station segment" reaches a preset threshold multiple times, it indicates that the original "fixed test duration" parameter can no longer reflect the true test time. In this case, the macro strategy layer directly updates the "fixed test duration" of that station segment to the "actual average completion time." These two duration correction mechanisms complement each other: the former is suitable for random delays caused by resource conflicts, using a gradually increasing buffer factor for smooth adjustment; the latter is suitable for systemic delays caused by equipment aging, changes in operating procedures, etc., using a direct replacement with the actual average value for rapid calibration. The aforementioned dynamic update process is executed once after the scheduling of each satellite is completed, until the preset maximum number of updates is reached. As the number of updates increases, the "allocation priority" and "fixed test duration" parameters of each "test station segment" gradually converge to a state that matches the actual test environment. The deviation between the scheduling scheme output by the macro strategy layer and the actual execution effect is significantly reduced. The final output scheduling strategy includes both the "macro strategy layer scheduling scheme," i.e., the satellite-station allocation result, and the "micro execution layer parallel scheduling strategy," i.e., the specific execution plan of the test items within the station segment, forming a closed-loop optimization from macro to micro and from theory to practice. The determination of the "preset maximum number of updates" needs to take into account the balance between the convergence speed of the macro-strategy layer scheduling scheme and the actual test batch size. It can be determined proportionally based on the total number of batched satellites, for example, set to 50% or 100% of the total number of batched satellites, to ensure that the update mechanism continues to be effective before most satellites are scheduled. The determination of the "preset threshold" requires a comprehensive balance between the tolerance for deviation between the actual completion time and the "fixed test duration" within the test station segment and the sensitivity of scheduling parameter updates. This "preset threshold" is a dimensionless percentage threshold, indicating that when the relative deviation of the "actual completion time" from the "fixed test duration" reaches this value, the original "fixed test duration" parameter is considered to have significantly deviated from the actual test time and requires calibration and updating. Therefore, the "preset threshold" can be statistically calibrated based on historical operating data of the test station segment: calculate the relative deviation distribution of the "actual completion time" relative to the "fixed test duration" under normal operating conditions, and take a high quantile of the deviation distribution, such as the 80th or 90th quantile, as the threshold, so that only abnormal batches with large deviations trigger updates. When historical data is lacking, the "preset threshold" can be set to 20% to 3% based on engineering experience. 0%, meaning that when the "actual completion time" exceeds the "fixed test duration" by more than 20% for three consecutive times, for example, the original duration parameter is considered no longer applicable, triggering an update to the "actual average completion time." Furthermore, this threshold can complement the determination of "resource conflict frequency": when the resource conflict frequency does not exceed the conflict threshold but the actual completion time remains excessively long, it indicates that the deviation mainly stems from systemic factors such as equipment aging or changes in operating procedures rather than resource competition. In this case, a lower "preset threshold," such as 10% to 15%, can more sensitively capture such deviations. When resource conflicts are frequent, adjustments are prioritized through the "buffer factor" mechanism, and the "preset threshold" can be appropriately relaxed to 25% to 30%.

[0045] Please see Figure 3 The present invention also provides a scheduling system for rapid testing of batch satellites, the scheduling system for rapid testing of batch satellites being used to execute the above-described scheduling method for rapid testing of batch satellites, comprising: Data acquisition module: For each satellite in the batch, the entire testing process of the satellite is broken down into multiple sequentially connected test stations according to function. Each test station contains several test items, and a discrete event simulation model is established based on the test items. The macro-scheduling calculation module is used to construct a dual-objective mixed integer programming model based on the discrete event simulation model, with the shortest total completion cycle of batch satellite parallel operations as the first optimization objective and the minimization of resource load balance among each test station segment as the second optimization objective. The dual-objective mixed integer programming model is solved by a heuristic scheduling algorithm based on constraint satisfaction, and the scheduling scheme of the macro-strategy layer is selected by Pareto non-dominated sorting. Local scheduling calculation module: After completing the scheduling scheme of the macro strategy layer, for each test station segment, based on the corresponding test items, a conflict topology graph between test items based on the resource occupation conflict relationship is constructed. According to the conflict topology graph, a greedy selection algorithm is used to schedule the test items in parallel. When the test items in any test station segment exceed the preset threshold, the graph coloring algorithm is switched to perform parallel scheduling. The strategy output module is used to optimize the parallel scheduling based on the results of the parallel scheduling, with the goal of minimizing the completion time of all test items in each test station segment, and generate a parallel scheduling strategy at the micro-execution layer. After completing the parallel scheduling strategy at the micro-execution layer, feedback data of the actual execution of each test station segment is collected. The macro-strategy layer dynamically updates the allocation priority and test duration parameters of the test station segment based on the feedback data. The above update process is repeated until the preset maximum number of updates is reached, and the final scheduling strategy is output.

[0046] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented in software, the above embodiments can be implemented, in whole or in part, as a computer program product. Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution.

[0047] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment, depending on actual needs.

[0048] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application.

Claims

1. A scheduling method for rapid testing of batch satellites, characterized in that, include: S1: For each satellite in the batch of satellites, the entire testing process of the satellite is broken down into multiple test stations connected in sequence according to function. Each test station contains several test items, and a discrete event simulation model is established based on the test items. S2: Using the discrete event simulation model as the operational support, with the shortest total completion cycle of batch satellite parallel operations as the first optimization objective and the minimization of resource load balance among each test station segment as the second optimization objective, a dual-objective mixed integer programming model is constructed; the dual-objective mixed integer programming model is solved using a constraint-based heuristic scheduling algorithm, and a scheduling scheme for the macro-strategy layer is selected through Pareto non-dominated sorting; S3: After completing the scheduling scheme of the macro strategy layer, for each test station segment, based on the corresponding test items, construct a conflict topology graph between test items based on the resource occupation conflict relationship. According to the conflict topology graph, the greedy selection algorithm is used to schedule the test items in parallel. When the test items in any test station segment exceed the preset threshold, switch to the graph coloring algorithm for parallel scheduling. S4: Based on the parallel scheduling results, optimize the process with the goal of minimizing the completion time of all test items within each test station segment, and generate a parallel scheduling strategy for the micro-execution layer. After completing the parallel scheduling strategy for the micro-execution layer, collect feedback data on the actual execution of each test station segment. The macro-strategy layer dynamically updates the allocation priority and test duration parameters of the test station segments based on the feedback data. Repeat the above update process until the preset maximum number of updates is reached, and output the final scheduling strategy.

2. The scheduling method for rapid testing of batch satellites according to claim 1, characterized in that, The established discrete event simulation model is as follows: In the discrete event simulation model, for each test station segment, resource status variables and project status variables are set based on the test project. The resource status variables are used to characterize the idle state, guaranteed occupancy state, and exclusive occupancy state of resources. The project status variables are used to characterize the unstarted state, in-process state, and completed state of the test project. The executable conditions of the test project are determined according to the guaranteed occupancy or exclusive occupancy requirements of the test project for resources. Test projects that do not meet the executable conditions are put into the waiting queue. The start events of the test items are added to the event queue according to the preset scheduling order, and the simulation clock is advanced according to the event occurrence time. When a test project meets the executable conditions, it occupies the corresponding resources and generates a completion event. When the test project is completed, it releases the corresponding resources and triggers the execution of test projects in the waiting queue that meet the executable conditions. Once all test items within a test station segment are completed and related resources are released, the completion time of that test station segment is output, and the corresponding satellite is allowed to enter the next test station segment.

3. The scheduling method for rapid testing of batch satellites according to claim 1, characterized in that, The bi-objective mixed-integer programming model is solved using a heuristic scheduling algorithm based on constraint satisfaction, specifically as follows: The bi-objective mixed integer programming model takes the shortest total completion cycle of batch satellite parallel operations as the primary optimization objective and the optimal resource load balance between test workstations as the secondary optimization objective. The total completion cycle is characterized by the maximum completion time among all satellite completion times, and the resource load balance is characterized by the deviation of the utilization rate of each test station segment under each test phase from the average utilization rate. The smaller the deviation, the more balanced the resource load among the test station segments. The utilization rate of the test station segment is determined based on the actual usage time of the corresponding test station segment and the total completion cycle; and the test duration of each satellite in each test stage is a preset fixed duration, and each test stage is executed sequentially according to a predetermined order, with test station segments with the same function in the same test stage being used in parallel.

4. The scheduling method for rapid testing of batch satellites according to claim 1, characterized in that, The candidate scheduling strategies under the corresponding rule combinations are obtained as follows: When solving the bi-objective mixed integer programming model using a constraint-satisfaction-based heuristic scheduling algorithm, satellite sorting rules and test station allocation rules are pre-set, and the satellite sorting rules and test station allocation rules are combined to form multiple rule combinations. For each rule combination, the satellites are processed sequentially according to the order of each test phase of the satellite AIT integrated test. In the current test phase, the satellites to be scheduled are sorted according to the corresponding satellite sorting rules, and the sorted satellites are sequentially assigned to the parallel test workstations with the same function in the current test phase according to the corresponding test workstation allocation rules. The satellite sorting rules include sorting according to preset priority and planned launch time, and the test workstation allocation rules include prioritizing the earliest available test workstation. The start time of each satellite in the current testing phase is determined by the completion time of its previous testing phase and the available time of the assigned testing station segment. After the allocation is completed, the available time of the corresponding testing station segment is updated. After all testing phases are completed, candidate scheduling strategies under the corresponding rule combinations are obtained.

5. The scheduling method for rapid testing of batch satellites according to claim 1, characterized in that, The scheduling scheme obtained at the macro-level strategy layer is as follows: Each candidate scheduling strategy is evaluated to obtain the corresponding total completion time and resource load balance of the test station segment. The total completion time is characterized by the maximum completion time among all satellite completion times, and the resource load balance is characterized by the deviation of the utilization rate of each test station segment from the average utilization rate. Pareto non-dominance screening is performed based on the total completion time and resource load balance of each candidate scheduling strategy. The candidate scheduling strategy that is not dominated by other candidate scheduling strategies in both total completion time and resource load balance is selected as the scheduling scheme of the macro strategy layer.

6. The scheduling method for rapid testing of batch satellites according to claim 1, characterized in that, The established conflict topology graph is as follows: Conflict topology graph between test items established in the micro-execution layer ,in For testing project nodes, For conflict edges, edge Indicates test items and There are resource conflicts, so parallel execution is not possible; for any two different test projects and Iterate through all resources and check for conflicts according to the following conflict determination rules: If two test items occupy the same resource, there will be a conflict if at least one of them is an exclusive occupy; there will be no conflict only if both are guaranteed occupy. The data on resource occupancy types comes from a predefined test project-resource occupancy relationship matrix; If any resource is used to determine the test item and If a conflict exists, then in the conflict topology graph Add an undirected edge To the edge of the market In the process, if there are conflicts on multiple resources, only one edge is added, and no duplicate edges are added. After completing the evaluation of all test items, output a complete conflict diagram. The nodes represent test projects, and the edges represent resource conflicts between two projects that cannot be executed in parallel.

7. The scheduling method for rapid testing of batch satellites according to claim 1, characterized in that, A greedy selection algorithm is used, specifically: Using the constructed conflict topology graph Parallel scheduling of test items: Sort all test items in descending order of execution time, and then try to compare each test item with the test items in the current parallel set in turn. If the test item has no edges with any test items in the set in the conflict graph, add it to the current parallel set; otherwise, skip it. The test items in the current parallel set are executed simultaneously. Repeat this process to process the remaining test items until all test items have been scheduled. When the number of test items exceeds a preset threshold, the graph coloring algorithm is switched for parallel scheduling, specifically as follows: Conflict topology diagram Vertex coloring is performed, and adjacent nodes connected by edges cannot be colored with the same color. Each color represents a time period for parallel execution. The minimum number of colors is the minimum number of parallel batches required. Items of the same color are executed simultaneously. Based on the scheduling results, calculate the total completion time of all test items within the test station segment. For the greedy parallel scheduling algorithm, the total completion time is equal to the sum of the durations of the longest test items in each parallel batch; for the graph coloring algorithm, the total completion time is equal to the sum of the durations of the longest test items in each batch corresponding to each color. Output the start and end times of each test item as the specific execution plan for the test items within the test station segment.

8. The scheduling method for rapid testing of batch satellites according to claim 4, characterized in that, The target test station segment is determined according to the updated allocation priority score, specifically as follows: After completing the parallel scheduling of test items in each test station segment, the micro-execution layer collects the actual completion time, fixed test duration, actual resource utilization rate and resource conflict frequency of the corresponding test station segment, and reports the feedback data along with the test station segment number to the macro-strategy layer. The macro strategy layer dynamically updates the allocation priority of the test station segment based on the actual resource utilization rate, sets an allocation priority score for each test station segment, and reduces the allocation priority score of the test station segment when the actual resource utilization rate of the test station segment is lower than the preset utilization rate threshold twice in a row, and restores the allocation priority score of the test station segment to the initial value when it is higher than the utilization rate threshold twice in a row. In the subsequent allocation of test sites for satellites, the target test site will be determined according to the updated allocation priority score.

9. The scheduling method for rapid testing of batch satellites according to claim 8, characterized in that, The output shows the scheduling scheme at the macro-level strategy layer and the parallel scheduling strategy at the micro-level execution layer, specifically: The macro strategy layer dynamically updates the fixed test duration of the test station segment based on the resource conflict frequency and the actual completion time. When the resource conflict frequency of the test station segment exceeds the preset conflict threshold twice in a row, a time buffer factor is introduced for the test station segment, and the fixed test duration is updated based on the time buffer factor. The time buffer factor increases in a preset step size and does not exceed a preset upper limit. When the proportion of test station segments whose actual completion time exceeds the fixed test duration multiple times reaches a preset threshold, the fixed test duration is updated based on the actual average completion time of the test station segment. After each satellite is scheduled, the macro strategy layer updates the allocation priority and fixed test duration of the test workstation based on the latest feedback data, and outputs the scheduling scheme of the macro strategy layer and the parallel scheduling strategy of the micro execution layer after reaching the preset maximum number of updates.

10. A scheduling system for rapid testing of batch satellites, used to implement the scheduling method according to any one of claims 1-9, characterized in that, include: Data acquisition module: For each satellite in the batch, the entire testing process of the satellite is broken down into multiple sequentially connected test stations according to function. Each test station contains several test items, and a discrete event simulation model is established based on the test items. The macro-scheduling calculation module is used to construct a dual-objective mixed integer programming model based on the discrete event simulation model, with the shortest total completion cycle of batch satellite parallel operations as the first optimization objective and the minimization of resource load balance among each test station segment as the second optimization objective. The dual-objective mixed integer programming model is solved by a heuristic scheduling algorithm based on constraint satisfaction, and the scheduling scheme of the macro-strategy layer is selected by Pareto non-dominated sorting. Local scheduling calculation module: After completing the scheduling scheme of the macro strategy layer, for each test station segment, based on the corresponding test items, a conflict topology graph between test items based on the resource occupation conflict relationship is constructed. According to the conflict topology graph, a greedy selection algorithm is used to schedule the test items in parallel. When the test items in any test station segment exceed the preset threshold, the graph coloring algorithm is switched to perform parallel scheduling. The strategy output module is used to optimize the parallel scheduling based on the results of the parallel scheduling, with the goal of minimizing the completion time of all test items in each test station segment, and generate a parallel scheduling strategy at the micro-execution layer. After completing the parallel scheduling strategy at the micro-execution layer, feedback data of the actual execution of each test station segment is collected. The macro-strategy layer dynamically updates the allocation priority and test duration parameters of the test station segment based on the feedback data. The above update process is repeated until the preset maximum number of updates is reached, and the final scheduling strategy is output.

Images