Low-orbit satellite large-scale earth observation task adaptive scheduling method and system

Abstract

The invention relates to the technical field of satellite observation task scheduling, in particular to a low-orbit satellite large-scale earth observation task self-adaptive scheduling method and system. The method comprises the following steps: generating a task allocation result according to a task candidate resource mapping table, namely generating an initial allocation scheme by using a greedy strategy of double factors of time and load, and establishing an allocation frequency memory matrix; constructing an inverse correlation selection probability model by utilizing the distribution frequency memory matrix, and outputting a task distribution result considering global revenue and load balance; task scheduling optimization is carried out based on a task allocation result, wherein conflict resolution fine adjustment is carried out by utilizing a time elastic adjustment mechanism of conflict perception; inter-satellite link collaborative rescheduling based on task characteristics; and global optimization and real-time response capability are effectively considered. According to the method, a complex joint scheduling problem is decoupled into two sub-problems with clear logic and low coupling degree, and the calculation complexity of single scheduling is greatly reduced.

Description

Technical Field

[0001] This invention relates to the field of satellite observation mission scheduling technology, and in particular to an adaptive scheduling method and system for large-scale low-orbit satellite Earth observation missions. Background Technology

[0002] With the increasing diversification and frequency of Earth observation needs, traditional single-satellite observation systems can no longer meet the requirements of large-scale, all-weather, and full-coverage missions. Earth observation satellites are gradually moving from single-satellite operation to a new stage of constellation-based and clustered collaborative observation. Low-Earth orbit satellite constellations, due to their advantages such as near-Earth operation, wide coverage, short revisit cycles, and flexible deployment, have become the core carrier for building a global observation network.

[0003] However, the development of satellite clusters also places higher demands on system resource management and real-time scheduling. While current research on large-scale Earth observation satellite mission scheduling has made some progress, several key limitations still exist: (1) Scheduling strategies struggle to balance real-time performance with global optimization capabilities. Existing methods often fall into a dilemma of "high computational complexity" and "poor solution quality" when dealing with large-scale task sets (such as hundreds to thousands of tasks). Exact algorithms cannot be applied online due to state space explosion, while heuristic or greedy strategies, although responsive, are prone to getting trapped in local optima, resulting in limited overall system benefits and resource utilization.

[0004] (2) The task conflict handling mechanism is simplistic, resulting in a low task completion rate. When multiple tasks have overlapping execution windows on the same satellite, most scheduling algorithms only consider task value, tending to retain high-value tasks while directly discarding low-value tasks. This "either / or" approach simplifies the scheduling logic but significantly reduces the system's overall ability to meet observation requirements. Especially in task-intensive scenarios, a large number of low- and medium-priority tasks cannot be executed for extended periods, affecting the integrity of service coverage.

[0005] (3) Insufficient consideration of load balancing affects the long-term stability of the system. Some scheduling schemes excessively pursue the maximization of total mission benefits and ignore the balance of storage, computing and communication loads among satellites, resulting in some satellites operating at high loads for a long time, accelerating platform aging, and even causing single points of failure, threatening the sustainable service capability of the entire constellation.

[0006] In summary, existing technologies have significant shortcomings in terms of dynamic adaptability, mission completion assurance, and system robustness. Therefore, there is an urgent need for a novel scheduling method that can deeply integrate task allocation and time scheduling, achieve task-characteristic-based reallocation, and take into account multi-objective collaborative optimization, in order to overcome the performance bottleneck of current low-Earth orbit satellite constellations in scheduling large-scale Earth observation missions. Summary of the Invention

[0007] To address the aforementioned problems, this invention provides an adaptive scheduling method and system for large-scale low-orbit satellite Earth observation missions.

[0008] Firstly, the present invention provides an adaptive scheduling method for large-scale low-orbit satellite Earth observation missions, which employs the following technical solution: An adaptive scheduling method for large-scale low-orbit satellite Earth observation missions includes: Receive observation requests from multi-source heterogeneous users and satellite data; Data preprocessing is performed on the received observation requests from multiple heterogeneous users. The preprocessed data is used to generate a task candidate resource mapping table; The task allocation results are generated based on the task candidate resource mapping table, including generating an initial allocation scheme and establishing an allocation frequency memory matrix using a greedy strategy with time and load as the two factors; using the allocation frequency memory matrix to construct an inverse correlation selection probability model, and outputting task allocation results that take into account both global benefits and load balancing. Task scheduling optimization is performed based on task allocation results, including fine-tuning for conflict resolution using a conflict-aware time elastic adjustment mechanism; and inter-satellite link collaborative rescheduling based on task characteristics. The scheduling instructions are encoded based on the optimized task allocation results.

[0009] Secondly, an adaptive scheduling system for large-scale low-orbit satellite Earth observation missions includes: The data acquisition module is configured to receive observation requests from multi-source heterogeneous users and satellite data; The preprocessing module is configured to preprocess the received multi-source heterogeneous user observation requests. The mapping module is configured to process the preprocessed data to generate a task candidate resource mapping table. The allocation module is configured to generate task allocation results based on the task candidate resource mapping table, including generating an initial allocation scheme and establishing an allocation frequency memory matrix using a greedy strategy with time and load as the two factors; using the allocation frequency memory matrix to construct an inverse correlation selection probability model and output a task allocation result that takes into account both global benefits and load balancing. The optimization module is configured to optimize task scheduling based on task allocation results, including fine-tuning conflict resolution using a conflict-aware time elastic adjustment mechanism and coordinating rescheduling of inter-satellite links based on task characteristics. The encoding module is configured to encode scheduling instructions based on the optimized task allocation results.

[0010] Thirdly, the present invention provides a computer-readable storage medium storing a plurality of instructions adapted for loading and execution by a processor of a terminal device of the aforementioned adaptive scheduling method for large-scale low-orbit satellite Earth observation missions.

[0011] Fourthly, the present invention provides a terminal device, including a processor and a computer-readable storage medium, wherein the processor is used to implement various instructions; the computer-readable storage medium is used to store multiple instructions, the instructions being adapted to be loaded and executed by the processor to provide an adaptive scheduling method for large-scale low-orbit satellite Earth observation missions.

[0012] In summary, the present invention has the following beneficial technical effects: (1) Effectively balances global optimization and real-time response capabilities. This invention decouples the complex joint scheduling problem into two logically clear and loosely coupled subproblems, significantly reducing the computational complexity of a single scheduling operation. The first stage uses an improved tabu search strategy based on the allocation frequency matrix to ensure search diversity while avoiding getting trapped in local optima. The second stage adopts a heuristic time scheduling mechanism, which can complete window allocation within milliseconds to seconds, meeting the online real-time scheduling requirements of low-Earth orbit satellite constellations for highly dynamic tasks.

[0013] (2) Balancing the protection of high-value tasks with the recovery of low-value tasks to maximize scheduling efficiency. When conflicts are irreconcilable, this invention constructs a priority ranking based on task benefit density and observation duration, prioritizing the retention of high-benefit-density or long-duration tasks, and incorporating more flexible short-duration tasks into the global rescheduling queue and redistributing them in their original visible satellite set. This ensures the execution of critical tasks while maximizing the recovery of tasks that might otherwise be discarded, thereby improving overall scheduling robustness and service coverage.

[0014] (3) It has good scalability and engineering applicability. The two-stage architecture proposed in this invention has modular characteristics, and the algorithms of each stage can be flexibly adjusted according to the actual constellation size, task density or hardware computing power. At the same time, all constraints, such as visibility window and memory capacity, are modeled based on real satellite orbital parameters, ensuring the feasibility and effectiveness of the scheme in actual engineering deployment. Attached Figure Description

[0015] Figure 1 This is a schematic diagram illustrating the operation of the Earth observation satellite system described in this embodiment of the invention; Figure 2 This is a schematic diagram of the process of generating candidate satellites for a mission as described in an embodiment of the present invention; Figure 3 This is a model architecture diagram of the frequency-guided task allocation strategy used to determine the task allocation scheme in the embodiments of the present invention; Figure 4This is a flowchart illustrating the frequency-guided task allocation strategy used to determine the task allocation scheme in an embodiment of the present invention. Figure 5 This is a comparison chart showing the total benefits of the method described in this embodiment of the invention and existing technical methods under different task scales. Figure 6 This is a comparison chart showing the task completion rate of the method described in this embodiment of the invention and existing technical methods under different task scales. Detailed Implementation

[0016] The present invention will be further described in detail below with reference to the accompanying drawings.

[0017] Example 1 Reference Figure 1 An adaptive scheduling method for large-scale low-orbit satellite Earth observation missions according to this embodiment includes: S1 Data Acquisition and Multidimensional Spatiotemporal Constraint Modeling S1.1 Users send observation mission requests to the satellite ground station via a client. To accommodate the diverse needs of different industries (such as disaster emergency response, agricultural monitoring, and urban planning), mission requests are encapsulated in the JSON standard format to form heterogeneous mission data packets. These data packets not only contain basic information such as target latitude and longitude and observation time windows, but also extend the definition to include imaging mode requirements (such as pushbroom, staring, and multi-point target), resolution thresholds, and spectral band selections (such as panchromatic, multispectral, and hyperspectral).

[0018] S1.2 Obtain high-precision satellite orbit and status parameters First, the real-time two-line orbital elements (TLE) of the satellite constellation are obtained through ground control stations or ephemeris servers. TLE data includes key Keplerian orbital elements such as the satellite's mean perihelion, orbital inclination, right ascension of the ascending node, and eccentricity. For each satellite in the constellation set S... (j=1,…,M), this invention establishes a high-dimensional state vector containing its static and dynamic characteristics. In addition to the basic position and velocity, the following physical properties that limit its working ability need to be precisely defined. These properties directly determine the boundaries of the constraints: Maximum storage capacity The unit is gigabit (Gbits), representing the upper limit of storage for a satellite-borne solid-state recorder. This parameter characterizes the upper limit of physical storage for a high-capacity satellite-borne solid-state recorder. In actual modeling, this parameter is not only limited by the hardware nominal value, but also needs to deduct the reserved space occupied by system logs, telemetry data, and bad blocks, representing only the net space available for storing payload data.

[0019] Maximum energy threshold The unit is ampere-hours (Ah), representing the upper limit of the satellite's battery pack energy. This parameter characterizes the upper limit of the battery pack energy in the satellite's power subsystem. This indicator is crucial for the satellite's operational capability in the Earth's shadow region, determining the maximum duration for which the satellite can sustain continuous payload operation when solar energy is unavailable.

[0020] Data transmission link rate The unit is megabits per second (Mbps), representing the average data transmission rate between the satellite and the ground. This parameter characterizes the average effective rate at which the satellite data transmission subsystem transmits data back to the ground station via X-band or Ka-band transmitters. This rate is typically limited by the modulation and demodulation method and channel coding efficiency, and is a key factor in calculating data transmission delay.

[0021] Sensor field of view The unit is degrees (°). This parameter defines the geometric coverage capability of the satellite-borne optical camera or synthetic aperture radar payload, specifically the semi-cone angle between the sensor's optical axis and the edge of the field of view. Combined with the satellite's orbital altitude, this parameter determines the swath width and ground coverage area of ​​the satellite's Earth observations.

[0022] S1.3 Obtain and parse Earth observation mission requests For each user observation task request received by the ground station, t i Data cleaning is performed on (i=1,…,N). First, syntax integrity is checked to ensure that key fields are not missing and the format conforms to the interface specification. The specific verification steps are as follows: First, verify that the metadata strictly includes essential fields such as the latitude and longitude of the mission location and the start and end times of the observation mission. Any missing field will trigger an abnormal interruption. Then, rigorously check the data types of each field to ensure that latitude and longitude are double-precision floating-point numbers and that the timestamp conforms to the ISO 8601 standard string format. Finally, impose physical constraints on the numerical range, such as latitude... ,longitude And perform timing logic checks.

[0023] Secondly, perform geofencing detection to automatically filter out tasks located in non-observation areas or invalid coordinate points. The specific detection steps are as follows: The system pre-configures a database of effective observation areas consisting of a series of closed polygon vertex sets. For each task target point, the topological relationship between the point and the polygon is determined: a mathematical ray is emitted from the task location P in any fixed direction, and the total number of intersections between this ray and the boundary line segment of the fence polygon is calculated. According to the parity rule, if If the coordinates are odd, the point is determined to be inside the observation area; if the coordinates are even, it is determined to be outside. This detection mechanism can quickly eliminate invalid coordinates located in non-sovereign areas, high seas, or sensitive restricted areas, ensuring the legitimacy of the scheduling targets.

[0024] Finally, the validated unstructured task requests are transformed into standardized six-tuple vectors that are recognizable within the system. Defined as: , in,( This indicates the geographical latitude and longitude of the ground target point. The value indicates the time window for task observation requested by the user, meaning the task must be completed within this time period. This indicates the duration of continuous observation required for the task, in seconds. This indicates the value of the task.

[0025] In addition, based on the imaging resolution required by the mission Based on the swath width and satellite sensor parameters, calculate the estimated storage resource consumption of this mission. : , in, For ground speed, For width, For pixel bit depth, This represents the compression ratio of the satellite image.

[0026] S1.4 Constructing a set of spatiotemporal visibility windows Computing satellites For the task Geometric visibility. Since the original TLE data is defined in the TME coordinate system, while the observation task is defined in the geographic coordinate system, a complex spatiotemporal reference system transformation is required: Using the SGP4 orbit prediction model, and considering the effects of Earth's non-spherical perturbation, atmospheric drag, and lunar gravitational perturbations, the TLE data is extrapolated to the scheduling time window. At any time t within the interval, we obtain the position vector defined in the true equatorial vernal equinox coordinate system. and velocity vector .

[0027] Then, by rotating using precession, nutation, and polar motion matrices, the satellite's position vector in the J2000 geocentric inertial coordinate system is calculated. and velocity vector To unify the spatiotemporal reference, the effects of precession and nutation of the Earth's rotation axis need to be eliminated. This requires constructing a precession rotation matrix. Nutation rotation matrix Transform the state vector to the J2000 geocentric inertial coordinate system: , , in, Calculated based on the IAU 2006 precession model, including Three Euler angles, Based on Huang Jingzhang's movement and intersecting octave calculate.

[0028] Subsequently, by incorporating Greenwich Mean Time (GMT), the state vector was converted to the WGS-84 Earth-fixed coordinate system, and the Earth-space geometric relationship was calculated: , , in, Used to correct minute deviations between the instantaneous rotation axis and the geographic reference pole. For polar shift parameters. This is the Earth's rotational angular velocity vector, used for the Coriolis term correction of the velocity vector. Describe the Earth's rotation about the Z-axis, the angle of rotation. The formula for calculating Greenwich Mean Time (GMT) is: , in, It is Greenwich Mean Sidereal Time.

[0029] Finally, define the satellite. For the task Visibility discriminant function : , in, For satellite position vectors, The ground target position vector, The elevation angle of the satellite relative to the ground station. This is the minimum observation elevation angle threshold. The discriminant function is constructed based on the geometric meaning of the vector dot product. The left side of the inequality calculates the cosine of the angle between the satellite position vector and the satellite-to-target pointing vector. The inverse cosine function is used. The actual included angle is obtained; if this included angle is less than or equal to the sensor's field of view half-cone angle... Then, geometrically, the target is determined to be within the field of view. Simultaneously, the satellite's elevation angle relative to the ground station must meet the following requirements. It is greater than the minimum threshold to ensure that the view is not obstructed by the curvature of the Earth.

[0030] By integrating the time over the scheduling period and solving the above inequality, the satellite... For the task All physically visible time windows : .

[0031] S1.5 Generate a candidate resource mapping table for tasks. For each task Find the intersection of the user-requested window and the physically visible window, i.e., the effective execution window. ,like Figure 2 As shown.

[0032] , If there exists a constraint that satisfies the duration constraint A continuous period of time, i.e. Make Then the satellite Join the mission candidate satellite set Simultaneously, empty sets are removed, and a preprocessed task resource mapping table is generated.

[0033] S2 Phase 1: Global task allocation guided by the allocation frequency matrix The core of this step is to decouple the complex three-dimensional coupling problem of task, time, and resources. In the task allocation phase, instead of directly determining the task execution time, the first step is to resolve the allocation issue of which satellite the task belongs to. The entire process is as follows: Figure 3 As shown.

[0034] S2.1 Initialize the task allocation scheme To accelerate convergence, a greedy strategy with both time and load factors is used to generate the initial solution. .

[0035] S2.1.1 All task sets T are sorted by arrival time. Sort in ascending order to obtain an ordered sequence. .

[0036] S2.1.2 Traversal Each task in Calculate its candidate satellite set Current normalized memory load rate of each satellite : , in, For satellites currently allocated The task set.

[0037] S2.1.3 Selection The smallest satellite As a task The initial allocation objective is to let the decision variables be... The rest are 0.

[0038] S2.2 Establish the frequency allocation memory matrix and adaptive probability model to construct a Long-term memory matrix .

[0039] Matrix elements The task was recorded during the algorithm iteration process. Assigned to satellite The cumulative number of occurrences. Define the inverse correlation selection probability density function. This function is used to guide the neighborhood search in subsequent iterations. It consists of two non-linearly weighted parts: a frequency penalty term and a resource potential reward term. , in, It is a very small positive number to prevent the denominator from being zero. and These are the adjustment weights for the frequency factor and the heuristic factor, respectively, used to coordinate the balance between seeking new solutions and optimizing known solutions in the algorithm. This formula uses the reciprocal of the assigned frequency as the baseline probability. If the task... It has been allocated to satellites multiple times. If the probability of selection decreases, the algorithm can escape the local optimum trap caused by the greedy strategy and explore other untried satellite resources. Simultaneously, introducing a heuristic factor increases the probability of selecting satellites with abundant remaining resources, thus guiding the system to achieve load balancing at the probabilistic level.

[0040] In this embodiment, the frequency factor weight is set to... = 2.0, the heuristic factor weights are 2.0. =1.5. The value is used to force the algorithm to explore the unknown solution space by strongly penalizing historically high-frequency allocation schemes, thus preventing premature convergence; while The value is to ensure that the load balancing mechanism plays a role throughout the entire iteration process, and to avoid falling into local extrema by excessively pursuing simple load balancing and ignoring global benefits.

[0041] It is a heuristic factor, defined as: , This factor reflects the satellite Remaining capacity for tasks The ability to accept.

[0042] S2.3 Perform multi-strategy neighborhood search and solution iteration. In each iteration, new solutions are generated through the following three operators to escape local optima: Single-task transfer operator: based on the above probabilities It randomly selects a task and moves it from the current satellite to another candidate satellite.

[0043] Dual-task exchange operator: Randomly select two tasks assigned to different satellites and exchange their assigned satellites. This operator is executed only if the tasks are candidates for each other.

[0044] Multi-task chain reorganization operator: Construct a task chain of length L, and cyclically move the task ownership in turn to achieve a larger range of perturbations.

[0045] The iteration termination condition is set to reach the maximum number of iterations. Or the objective function value has not improved for several consecutive generations.

[0046] S3 Phase 2: Multi-level Time Optimization Scheduling After determining the affiliation between missions and satellites, this stage requires scheduling precise execution time windows for the mission sequence on each satellite. ,like Figure 4 As shown.

[0047] S3.1 Initialization of the data structure for the satellite timeline For each satellite Maintain a timeline linked list The linked list contains a series of occupied and idle time slices. This is for the time slices allocated to satellites. task subset Implement a dual sorting strategy: First keyword: Deadline Sort in ascending order to ensure that urgent tasks are prioritized.

[0048] Second keyword: Observation duration Arrange in descending order, with longer tasks given priority, as they have poor flexibility and require priority scheduling.

[0049] S3.2 Optimal Insertion Position Based on Temporal Balance Scoring For tasks to be scheduled Within its effective window, there may be multiple feasible insertion positions, such as the window start point, end point, and midpoint. To avoid fragmentation, this invention proposes a comprehensive scoring index. To evaluate each candidate insertion time The advantages and disadvantages. For any candidate time... Its corresponding end time is .definition: Left side idle time This refers to the gap between the insertion point and the end time of the preceding task.

[0050] Right side idle time This refers to the gap between the insertion point and the start time of the subsequent task.

[0051] Total free space .

[0052] The following scoring function is constructed to maximize the integrity of the remaining space while maintaining a balanced distribution: , The scoring function evaluates the quality of time fragments through mathematical construction. The balance component is mathematically similar to the harmonic mean. According to the mean-mean inequality, this component reaches its maximum value if and only if the idle time on the left side equals the idle time on the right side. This means the algorithm tends to insert tasks in the exact center of idle time periods, thus avoiding leaving extremely short, unusable time fragments on one side. The deviation penalty term utilizes the rapid growth characteristic of the exponential function to impose a strong, non-linear penalty on any asymmetric insertion position, forcing tasks to be distributed as centrally as possible.

[0053] Iterate through all feasible insertion points and select... The biggest moment as a task The provisional execution time. In this embodiment, the weight coefficients of each component of the scoring function are set to... = 0.5, = 0.3, = 0.2. Wherein, The largest proportion is because the primary goal is to ensure the symmetry of time window segmentation and prevent the generation of small, unusable fragments of time. Secondly, it is used to encourage the retention of larger, continuous remaining space; It is relatively small and serves as a supplementary penalty, used to fine-tune the deviation.

[0054] S4 Conflict Detection Time Flexibility Adjustment Mechanism When the task When attempting to insert, if the optimal position is found to be occupied by a high-priority task or there is overlap, the task is not immediately discarded, but conflict resolution fine-tuning is initiated.

[0055] S4.1 Calculation of Conflict Overlap Metric Set a new task The planned time period is The task already exists. The time period is If the intersection of the two is not empty, calculate the conflict duration. : , S4.2 Two-way flexible sliding window adjustment Define the elastic adjustment operator Try to make a slight translation of the task on the timeline: S4.2.1 Forward Probing: Attempting to advance the task The start time was brought forward Defined as: , in, For safe isolation intervals.

[0056] Inspect the new section Is it still within the active window and does not conflict with other tasks?

[0057] S4.2.2 Backward probing: If forward probing fails, try pushing the task back by the same distance.

[0058] S4.2.3 Compression Probe: If the task allows for service degradation, attempt to reduce the observation duration to [a certain value]. Try inserting again. Defined as: , in, This is the degradation factor. In this embodiment, the degradation factor is set to a value of [value missing]. = 0.8. This coefficient indicates that in the event of irreconcilable conflicts, if the mission allows for degraded service, the mission observation duration will be reduced to 80% of the original plan. According to the push-broom imaging principle of low-Earth orbit satellites, retaining 80% of the duration can usually still cover the core area of ​​the target. Under the premise of ensuring the basic effectiveness of the mission, 20% of the time resources are freed up to resolve overlapping conflicts.

[0059] If all the above adjustments fail, the task is marked as a locally scheduled failure and placed in the rescheduling queue. .

[0060] S5 Inter-satellite Link Cooperative Rescheduling Based on Task Characteristics For missions that cannot be handled locally, the inter-satellite link capability of the low-Earth orbit constellation is used to allocate them to other idle satellites. This step is crucial for improving mission completion rates.

[0061] S5.1 Constructing a dynamic priority queue based on revenue density right The tasks in the process are reordered. Traditional reward-based... The current ranking is too simplistic, so this invention introduces a spatiotemporal efficiency priority factor. : , in, This represents the density of returns per unit of time. The square of the duration is used as the denominator here to greatly increase the weight of short, high-value tasks, because short tasks are easier to insert into fragmented time. It is a congestion penalty item. This refers to the inter-satellite link bandwidth. Tasks with excessively large data volumes will consume too many link resources, and their priority will be appropriately reduced.

[0062] S5.2 Feasibility Assessment of Cross-Satellite Rescheduling according to Traverse the tasks in descending order. For each task... : Retrieve the set of all candidate satellites other than the originally assigned satellites. .

[0063] , in, The satellites were originally allocated.

[0064] Based on the remaining satellite storage capacity The satellites are sorted in descending order, with priority given to those with abundant resources.

[0065] For each candidate satellite Check if a time window exists that satisfies both the observation and transmission constraints: , And it must meet the inter-satellite transmission delay constraint: , in, This refers to the inter-satellite link transmission rate. The handshake time for establishing a link.

[0066] Once a satellite that meets the criteria is found If the task fails, the resources are locked, the global scheduling scheme is updated, and the task is removed from the failure queue.

[0067] S6 scheduling instruction encoding, encapsulation, and annotation This step involves converting the mathematically optimal scheduling scheme generated by ground calculations into binary machine code executable by the spacecraft's underlying hardware, which involves complex protocol stack encapsulation and anti-interference link transmission.

[0068] S6.1 Instruction Serialization and Encapsulation The ground station first converts the final global scheduling scheme into instructions that can be recognized by the satellite. The instruction format is defined as a high-dimensional vector tuple. : , in, The operation code indicates the payload type, such as turning on the synthetic aperture radar beam or unlocking the optical camera shutter. To ensure absolute execution time, accurate to the millisecond level, and to eliminate satellite-to-ground clock discrepancies, the ground station needs to calculate transmission delay and Doppler shift before adding the command, and perform time calibration using an onboard crystal oscillator drift model. The target attitude angle is calculated based on quaternions and is used to guide the satellite in yaw maneuvers. This represents the duration of task execution, corresponding to the observation duration mentioned earlier. . This indicates imaging mode parameters, such as strip mode, spotlight mode, or sliding spotlight mode.

[0069] S6.2 Encryption Modulation and Closed-Loop Transmission Verification The encapsulated command packet is first processed by the AES-256 encryption algorithm to prevent unauthorized access, and then transmitted to the satellite via the ground station antenna when the satellite passes overhead.

[0070] S6.3 On-board Demodulation and ARQ Retransmission Mechanism After demodulation by the satellite receiver, the onboard computer performs BCH error correction coding verification and CRC cyclic redundancy check. If the verification passes, the satellite returns a "command acknowledgment frame (ACK)" to the ground via the downlink telemetry link and writes the command into the command buffer queue; if the verification fails, it returns a "retransmission request (NACK)" and triggers the ground's automatic retransmission request (ARQ) mechanism until the command is successfully transmitted, thereby ensuring a zero bit error rate in command transmission.

[0071] S7 onboard attitude maneuvering, collaborative observation, and data caching After receiving the command, the satellite enters the autonomous execution phase, where various subsystems work collaboratively to form a pipeline operation encompassing attitude maneuvering, payload imaging, data processing, and storage management. When the onboard clock reaches the specified time... At any given moment, the onboard sensors are activated to image designated ground targets.

[0072] The raw observation data first enters the onboard FPGA for preprocessing to remove invalid pixels. The preprocessed data stream then enters the data compression unit, typically employing lossless or near-lossless compression algorithms to reduce the data rate. The compressed data stream is then written to the onboard high-capacity solid-state storage.

[0073] S8 High-Speed ​​Data Transmission Downlink and Storage Resource Release This step describes how the satellite utilizes a limited transit window to safely and efficiently transmit massive amounts of observational data back to Earth via a highly dynamic link, and achieves automated recycling of onboard resources.

[0074] When the satellite enters the ground station's communication range, the onboard data transmission antenna uses ephemeris data to calculate the pointing vector and automatically acquires and tracks the ground station. After establishing a physical link, the satellite data transmission subsystem initiates an adaptive coding and modulation mechanism to transmit the stored observation data back via the downlink. The onboard file management system divides the large volume of observation data stored in the solid-state recorder into several protocol data units. The transmission mode adopts a reliable transmission mode, i.e., it supports a retransmission mechanism. The ground station performs real-time error detection during reception. If packet loss or bit errors are detected, a "Selective Retransmission Request (SelectiveNACK)" is sent via the uplink. The satellite only retransmits the lost data blocks, rather than the entire file, thereby significantly improving transmission efficiency.

[0075] After the ground station successfully receives and verifies all data frames, it sends a successful reception confirmation frame and a storage release command via the uplink. Upon receiving the confirmation, the satellite's file management system marks the corresponding storage block as erasable and rewritable, thereby releasing storage space, updating the remaining capacity, and freeing up resources for subsequent scheduling tasks.

[0076] The raw data received by the S9 ground station is a binary bit stream, which needs to be decoded, calibrated and image processed before the processed image products are delivered to the end user according to the user's needs.

[0077] To verify the effectiveness of the solution described in this embodiment, the following experiments were conducted: The simulation experiment is as follows: Using satellite orbit parameters and related data provided by the Tiansuan constellation East China Station (Yantai), a low-Earth orbit satellite constellation consisting of 10 satellites was constructed. Ground target points were randomly and uniformly generated within the range of (3°N–53°N, 74°E–133°E) as observation tasks. The memory consumption and task benefits for each task were configured according to actual needs.

[0078] In this paper, TASS represents the two-stage adaptive satellite scheduling algorithm proposed in this invention, while Multi-Agent Reinforcement Learning for Satellite Scheduling (MARS), Improved Differential Evolutionary Algorithm (IDE), Two-Layer Evolutionary Algorithm (BEATS), and Hybrid Greedy Weighted Scheduling (HGWS) represent comparative algorithms. Experimental results are as follows: Figure 5 and Figure 6 As shown.

[0079] Figure 5The experiment demonstrates the total task revenue of different scheduling algorithms under varying task sizes. The results show that the proposed method achieves higher total revenue compared to other comparative algorithms. Particularly when the task size reaches 1000, the total revenue curve of the proposed method (TASS) is significantly higher than that of other comparative algorithms, with an average total revenue improvement of 4.59% compared to the BEATS algorithm.

[0080] Figure 6 The paper demonstrates the task completion rates of different scheduling algorithms under varying task sizes. As the task load increases, the completion rate of traditional algorithms decreases linearly. However, the inter-satellite link collaborative rescheduling strategy proposed in this invention, based on task characteristics, effectively recovers previously discarded short tasks, resulting in a smoother completion rate curve that remains above 90% even under heavy loads. Experiments show that the task completion rate of the proposed method outperforms the comparative algorithms across different task sizes.

[0081] Example 2 This embodiment provides an adaptive scheduling system for large-scale low-orbit satellite Earth observation missions, including: The data acquisition module is configured to receive observation requests from multi-source heterogeneous users and satellite data; The preprocessing module is configured to preprocess the received multi-source heterogeneous user observation requests. The mapping module is configured to process the preprocessed data to generate a task candidate resource mapping table. The allocation module is configured to generate task allocation results based on the task candidate resource mapping table, including generating an initial allocation scheme and establishing an allocation frequency memory matrix using a greedy strategy with time and load as the two factors; using the allocation frequency memory matrix to construct an inverse correlation selection probability model and output a task allocation result that takes into account both global benefits and load balancing. The optimization module is configured to optimize task scheduling based on task allocation results, including fine-tuning conflict resolution using a conflict-aware time elastic adjustment mechanism and coordinating rescheduling of inter-satellite links based on task characteristics. The encoding module is configured to encode scheduling instructions based on the optimized task allocation results.

[0082] A computer-readable storage medium storing a plurality of instructions adapted for loading and execution by a processor of a terminal device, the aforementioned adaptive scheduling method for large-scale low-Earth orbit satellite Earth observation missions.

[0083] A terminal device includes a processor and a computer-readable storage medium, wherein the processor is used to implement various instructions; the computer-readable storage medium is used to store multiple instructions, which are adapted to be loaded and executed by the processor to provide an adaptive scheduling method for large-scale low-Earth orbit satellite Earth observation missions.

[0084] The above are all preferred embodiments of the present invention and are not intended to limit the scope of protection of the present invention. Therefore, all equivalent changes made in accordance with the structure, shape and principle of the present invention should be covered within the scope of protection of the present invention.

Claims

1. An adaptive scheduling method for large-scale low-orbit satellite Earth observation missions, characterized in that, include: Receive observation requests from multi-source heterogeneous users and satellite data; Data preprocessing is performed on the received observation requests from multiple heterogeneous users. The preprocessed data is used to generate a task candidate resource mapping table; The task allocation results are generated based on the task candidate resource mapping table, including generating an initial allocation scheme and establishing an allocation frequency memory matrix using a greedy strategy with time and load as the two factors; using the allocation frequency memory matrix to construct an inverse correlation selection probability model, and outputting task allocation results that take into account both global benefits and load balancing. Task scheduling optimization is performed based on task allocation results, including fine-tuning conflict resolution using a conflict-aware time elasticity adjustment mechanism. Inter-satellite link cooperative rescheduling based on mission characteristics; The scheduling instructions are encoded based on the optimized task allocation results.

2. The adaptive scheduling method for large-scale low-orbit satellite Earth observation missions according to claim 1, characterized in that, The receiving of multi-source heterogeneous user observation requests and satellite data includes obtaining the real-time two-line orbital elements TLE of the satellite constellation through ground control stations or ephemeris servers, for each satellite in the constellation set S. (j=1,…,M), establish a high-dimensional state vector containing static and dynamic characteristics, and define physical properties that limit working capacity to determine the boundary of constraints, specifically including: maximum storage capacity. This indicates the upper limit of storage for satellite-borne solid-state recorders; maximum energy threshold. Used to indicate the upper limit of satellite battery pack energy; data transmission link rate Used to represent the average rate of satellite-to-ground data transmission; sensor field of view. Used to define the geometric coverage capability of satellite-borne optical cameras or synthetic aperture radar payloads, specifically the semi-cone angle between the sensor's optical axis and the edge of the field of view.

3. The adaptive scheduling method for large-scale low-orbit satellite Earth observation missions according to claim 2, characterized in that, The data preprocessing of the received multi-source heterogeneous user observation requests includes processing the received user observation task requests t via the ground station. i For each i=1,…,N, data cleaning is performed. First, syntax integrity is checked to ensure that key fields are not missing and the format conforms to the interface specification. Then, geofencing detection is performed to automatically filter out tasks located in non-observation areas or invalid coordinate points. At the same time, unstructured task requests that pass the check are converted into standardized six-tuple vectors that can be recognized internally by the system. Defined as: ， in,( ) represents the geographical latitude and longitude of the ground target point, [ This indicates the user-required task observation time window, meaning the task must be completed within this time period. This indicates the duration of continuous observation required for the task. This indicates the value of the task; finally, it is based on the imaging resolution required by the task. Based on the swath width and satellite sensor parameters, the estimated storage resource consumption of the mission is calculated. : ， in, For ground speed, For width, For pixel bit depth, This represents the compression ratio of the satellite image.

4. The adaptive scheduling method for large-scale low-orbit satellite Earth observation missions according to claim 3, characterized in that, The process of generating a task candidate resource mapping table based on preprocessed data includes constructing a spatiotemporal visibility window set, first calculating satellite... For the task To improve geometric visibility, a complex spatiotemporal reference frame transformation is performed, where the original TLE data is defined in the TME coordinate system while the observation mission is defined in the geographic coordinate system. This involves using the SGP4 orbit prediction model, considering the effects of Earth's non-spherical perturbations, atmospheric drag, and lunar gravitational perturbations, to extrapolate the TLE data to the scheduling time window. At any time t within the range, the position vector of the satellite in the J2000 geocentric inertial coordinate system is calculated by rotating using the precession, nutation, and polar motion matrices. and velocity vector Then, by incorporating Greenwich Mean Time (GMT), it was converted into a state vector in the WGS-84 Earth-Fixed Coordinate System (ECEF), and the Earth-Satellite geometric relationship was calculated. Finally, the satellite was defined. For the task Visibility discriminant function , is represented as: ， in, For satellite position vectors, The ground target position vector, The elevation angle of the satellite relative to the ground station. The minimum observation elevation angle threshold is determined; and by integrating the time over the scheduling cycle, the inequality is solved to obtain the satellite... For the task All physically visible time windows : For each task Find the intersection of the user-requested window and the physically visible window, i.e., the effective execution window. , , If there exists a constraint that satisfies the duration constraint A continuous period of time, i.e. Make Then the satellite Join the mission candidate satellite set At the same time, empty sets are removed, and a preprocessed task resource mapping table is generated.

5. The adaptive scheduling method for large-scale low-orbit satellite Earth observation missions according to claim 4, characterized in that, The method utilizes a greedy strategy based on both time and load factors to generate an initial allocation scheme and establish an allocation frequency memory matrix. This includes decoupling the three-dimensional coupling problem of task-time-resource during the task allocation phase. To accelerate convergence, a greedy strategy based on both time and load factors is employed to generate the initial solution. First, categorize all task sets T according to their arrival times. Sort in ascending order to obtain an ordered sequence. Then iterate through Each task in Calculate the candidate satellite set Current normalized memory load rate of each satellite : ,in, For satellites currently allocated The task set, and then select The smallest satellite As a task The initial allocation objective is to let the decision variables be... The rest are 0; finally, the frequency allocation memory matrix and adaptive probability model are established to construct... Long-term memory matrix Matrix elements Record the task during the algorithm iteration process. Assigned to satellite The cumulative number of occurrences is used to define the inverse correlation selection probability density function. This function, used to guide neighborhood search in subsequent iterations, consists of a non-linear weighted combination of a frequency penalty term and a resource potential reward term. ， in, It is a very small positive number, to prevent the denominator from being zero. and These are the adjustment weights of the frequency factor and the heuristic factor, respectively, used to coordinate the balance between seeking new solutions and optimizing known solutions in the algorithm. It is a heuristic factor, defined as: Used to reflect satellite Remaining capacity for tasks The ability to accept.

6. The adaptive scheduling method for large-scale low-orbit satellite Earth observation missions according to claim 5, characterized in that, The method utilizes the allocation frequency memory matrix to construct an inverse correlation selection probability model, outputting task allocation results that balance global benefits and load balancing. This includes iterative multi-strategy neighborhood search and solution generation based on the allocation frequency memory matrix. In each iteration, three operators are used to generate new solutions to escape local optima. The iteration termination condition is set to reaching the maximum number of iterations. Or the objective function value has not improved for several consecutive generations; after determining the affiliation between the mission and the satellite, a precise execution time window is arranged for the mission sequence on each satellite. , of which, for each satellite Maintain a timeline linked list The linked list contains a series of occupied and idle time slices allocated to satellites. task subset A dual sorting strategy is implemented for tasks to be scheduled. Within its effective window, there can be multiple feasible insertion positions. To avoid fragmentation, a comprehensive scoring index is constructed. To evaluate each candidate insertion time The merits and demerits of any candidate time point Its corresponding end time is Define the left-side idle degree The interval between the insertion point and the end time of the preceding task, and the idle time on the right. The total free space of the window is the interval between the insertion point and the start time of the subsequent task. Finally, a scoring function is constructed to maximize the integrity of the remaining space and maintain a balanced distribution, expressed as: ， By traversing all feasible insertion points, select The biggest moment as a task The tentative implementation time.

7. The adaptive scheduling method for large-scale low-orbit satellite Earth observation missions according to claim 6, characterized in that, The aforementioned conflict resolution fine-tuning using a conflict-aware time-elastic adjustment mechanism includes when the task... When attempting insertion, if the optimal position is found to be occupied by a higher-priority task or there is overlap, conflict resolution fine-tuning is initiated. First, the conflict overlap metric is calculated, and a new task is set up. The planned time period is The task already exists. The time period is If the intersection of the two is not empty, calculate the conflict duration. : ； Then, a bidirectional elastic sliding window adjustment is performed, and an elastic adjustment operator is defined. Attempts were made to make small translations of the task on the timeline, including forward probing, to try to shift the task... The start time was brought forward Defined as: in, To ensure safety, separate the intervals and inspect the new area. Whether it is still within the valid window and does not conflict with other tasks; backward probing: if forward probing fails, try pushing the task back by the same distance; and compression probing: if the task allows for service degradation, reduce the resolution or shorten the duration, and try to reduce the observation duration to [a certain value]. Try inserting again. Defined as: in, This is the degradation factor; finally, if all the above adjustments fail, the task is marked as a locally scheduled failure and placed in the rescheduling queue. .

8. The adaptive scheduling method for large-scale low-orbit satellite Earth observation missions according to claim 7, characterized in that, The inter-satellite link cooperative rescheduling based on mission characteristics includes constructing a dynamic priority queue based on benefit density, firstly for... The tasks in the process are reordered, and a time-space efficiency priority factor is introduced. : in, Indicates the return density per unit time. It is a congestion penalty item. The inter-satellite link bandwidth is used; then, a feasibility assessment of cross-satellite rescheduling is performed, according to... Traverse the tasks in descending order, for each task : Retrieve the set of all candidate satellites other than the originally assigned satellites , in, The original satellite allocation was adjusted based on the remaining storage capacity of the satellites. Sort in descending order, for each candidate satellite Check if a time window exists that satisfies both the observation and transmission constraints: ， And it must satisfy the inter-satellite transmission delay constraint, expressed as: , in, This refers to the inter-satellite link transmission rate. The handshake time for link establishment, when a satellite that meets the conditions is found. If the task fails, the resources are locked, the global scheduling scheme is updated, and the task is removed from the failure queue.

9. The adaptive scheduling method for large-scale low-orbit satellite Earth observation missions according to claim 8, characterized in that, The scheduling instruction encoding based on the optimized task allocation results includes converting the optimal scheduling scheme generated by ground calculation into binary machine code executable by the spacecraft's underlying hardware. First, the final global scheduling scheme is converted into instructions that the satellite can recognize. The instruction format is defined as a high-dimensional vector tuple. : ,in, For opcodes, For absolute execution time, ( The target attitude angle is calculated based on quaternions. This represents the duration of the task execution. The command packet represents the imaging mode parameters; the encapsulated command packet is encrypted using the AES-256 algorithm to prevent unauthorized access, and is transmitted to the satellite via the ground station antenna when the satellite passes overhead; after demodulation by the satellite receiver, the onboard computer performs BCH error correction coding verification and CRC cyclic redundancy check; after receiving the command, the satellite enters the autonomous execution phase, where various subsystems work together to form a pipeline operation of attitude maneuvering, payload imaging, data processing, and storage management. When the onboard clock reaches the specified time, the satellite will automatically execute the command. At any given moment, the onboard sensors are activated to image designated ground targets.

10. An adaptive scheduling system for large-scale low-orbit satellite Earth observation missions, characterized in that, include: The data acquisition module is configured to receive observation requests from multi-source heterogeneous users and satellite data; The preprocessing module is configured to preprocess the received multi-source heterogeneous user observation requests. The mapping module is configured to process the preprocessed data to generate a task candidate resource mapping table. The allocation module is configured to generate task allocation results based on the task candidate resource mapping table, including generating an initial allocation scheme and establishing an allocation frequency memory matrix using a greedy strategy with time and load as the two factors; using the allocation frequency memory matrix to construct an inverse correlation selection probability model and output a task allocation result that takes into account both global benefits and load balancing. The optimization module is configured to optimize task scheduling based on task allocation results, including fine-tuning conflict resolution by utilizing a conflict-aware time elastic adjustment mechanism. Inter-satellite link cooperative rescheduling based on mission characteristics; The encoding module is configured to encode scheduling instructions based on the optimized task allocation results.

Images