Target rolling clustering task planning method based on sub-satellite point trajectory

Abstract

The invention relates to a target rolling clustering task planning method based on a sub-satellite point trajectory. The invention relates to the technical field of multi-target observation. Point targets which are close to each other and associated with the same sub-satellite point trajectory are integrated into a clustering observation target; distributing observation satellites according to the association degree of the clustering target and satellites corresponding to different sub-satellite point trajectories, and realizing pre-distribution of observation tasks; and task planning is carried out on all processed clustering targets and single targets, and conflicts between adjacent tasks are solved through a sliding window. As the sub-satellite point trajectory moves along with the rotation of the earth, the target clustered for each trajectory is updated in a rolling manner. Simulation examples show that by means of the clustering method, the number of tasks can be reasonably and effectively reduced, target observation conflicts can be relieved, calculation time consumption can be reduced, and the effect is better when the number of the targets is larger.

Description

Technical Field

[0001] This invention relates to the field of multi-target observation technology and is a target rolling clustering task planning method based on nadir point trajectories. Background Technology

[0002] With the deepening application of Earth observation satellites and the rapid development of satellite technology, the number of remote sensing satellites in global orbit has surged. The rise of mega-constellations in low Earth orbit marks the beginning of the "constellation era" for remote sensing, significantly enhancing multi-target observation capabilities. Utilizing constellations for multi-target observation requires mission planning to maximize observational benefits under limited resources. Current research mainly focuses on selecting and improving effective mission planning algorithms, establishing constraint models and objective functions for specific problems to achieve efficient utilization of satellite resources.

[0003] Planning observation missions for multiple ground targets within a constellation is a classic NP-hard problem, especially for large constellations where the solution space is enormous and difficult to solve directly through traversal. Heuristic algorithms, metaheuristic algorithms, and machine learning are typically employed to search the solution space. Heuristic methods were first applied to mission planning for remote sensing satellites. Existing techniques, using greedy algorithms and constructive heuristics based on residual mission density, have yielded better mission solutions within acceptable computation time. While heuristic algorithms cannot guarantee finding the optimal solution, they are rarely used in routine applications. However, their low computational complexity and speed make them suitable for some relatively urgent missions. Metaheuristic algorithms are widely used in constellation mission planning, including genetic algorithms, particle swarm optimization, ant colony optimization, tabu search, simulated annealing, and adaptive large neighborhood search. Some researchers have also used hybrid algorithms combining different algorithms for specific problems or objectives to combine their advantages. Existing technologies propose a Hybrid Genetic Tabu Search (HGTSA) algorithm for single-channel imaging satellite planning, maximizing the number of observation targets under limited energy constraints. For regional target observation planning, a novel Integer-Encoded Genetic Algorithm (IRICGA) based on individual weight allocation is proposed, reducing the computational cost of satellite scheduling. Existing technologies also propose a learning meme algorithm based on variable populations and neighborhoods for multi-target scheduling of large-scale heterogeneous imaging satellites, achieving co-evolution of two populations. With the development of machine learning methods, represented by reinforcement learning, planning algorithms have shifted from the "low intelligence" of metaheuristic algorithms to "high intelligence." Existing technologies propose a satellite scheduling method based on deep reinforcement learning, solving the single-satellite multi-target scheduling problem. The Q-Learning algorithm, the advanced actor-critic (A2C) algorithm, and a self-attention mechanism are used to solve the multi-satellite multi-target observation problem.

[0004] The aforementioned research methods, whether heuristic algorithms, metaheuristic algorithms, or reinforcement learning methods for end-to-end solution generation, essentially all directly search the solution space without processing it. In reality, the solution space can be reduced through reasonable methods to improve solution efficiency and even solution quality. These methods can be divided into two categories: task pre-assignment and task clustering. Task pre-assignment involves allocating targets to satellites that may perform the observation task in advance according to certain rules, and then scheduling the observation tasks for each satellite. Existing technologies predict the probability of different satellites completing tasks based on historical satellite scheduling data, assigning targets to different satellites and transforming the multi-satellite planning problem into single-satellite task planning. Existing technologies use probabilistic models to assign targets to different orbits, which also reduces the solution space to some extent. Existing technologies have proposed adaptive task allocation algorithms that can reallocate unscheduled tasks to obtain higher observation benefits. Pre-assignment mainly focuses on discrete targets and is suitable for scenarios with a limited number of targets. If the number of targets is too large, it will increase the difficulty of subsequent single-satellite task scheduling. Therefore, for dense multi-target scenarios, some researchers have adopted task clustering methods. Existing technologies have proposed multi-stage mission planning methods, but mission clustering methods all aggregate missions that can be observed using the same satellite attitude. In essence, this method merges missions with overlapping observation windows by extending the observation time, resulting in significant energy and memory consumption and prolonged occupation of satellite resources. Existing technologies have designed a mission clustering method that groups nearby targets for observation, allowing a single observation mission to observe multiple targets and improving planning efficiency. These technologies not only aggregate nearby targets but also reselect the observation positions of the clustered targets, improving observation quality. Existing technologies have applied mission clustering to space debris tracking, similarly aggregating debris with similar orbital positions into a single observation target. However, the clustering methods proposed in existing technologies are static and do not address the conflict problem of the same target appearing in different selectable clusters.

[0005] It can be seen that compared with research on task planning algorithms, research on the preprocessing of observation targets is relatively limited and needs improvement. Therefore, a new target clustering method needs to be designed for observation task planning. For the problem of dense point target observation in large-scale satellite constellations, we propose a target rolling clustering method based on nadir trajectories, combining task pre-assignment and task clustering methods. First, target positions are statically clustered, and the observation orbits to which targets belong are assigned based on the matching degree between the target and different nadir trajectories. As the satellite moves, new nadir trajectories will form on the ground in different orbits, thus requiring dynamic updates to the target clusters. By dynamically rolling and updating the target clusters, the clustering of dense targets and task allocation in each orbit are realized, which can reasonably and effectively reduce the number of tasks, alleviate target observation conflicts, and reduce computation time. The denser the satellite orbits, the better the rolling clustering effect.

[0006] With the development of remote sensing technology, the demand for observation has gradually increased, and it is often concentrated in certain hotspot areas. This type of problem can be called the Dense Target Observation Problem (DTOP). To solve the problem of observing dense targets from star clusters, it is first necessary to analyze the problem, define the input, output and process of the solution method, and explain in detail the key assumptions, constraints and objective function in the planning process. Summary of the Invention

[0007] This invention divides the dense target observation problem into three main parts: target clustering, task allocation, and a task master scheduling algorithm. In the target clustering part, multiple observation targets are aggregated into one observation target, and the observation center position is reselected. A dynamic clustering method is used to re-cluster the targets within each orbital period. In the task allocation part, the new set of clustered targets is initially allocated tasks according to their matching degree with different observation orbits. In the task scheduling part, different observation tasks allocated to different orbits are centrally scheduled to maximize the observation targets. A target rolling clustering task planning method based on sub-satellite point trajectories is disclosed.

[0008] This invention provides the following technical solutions: A target rolling clustering task planning method based on sub-satellite point trajectories, the method comprising the following steps: Step 1: Aggregate multiple observation targets into one observation target, and reselect the observation center location. Use dynamic clustering to re-cluster the targets in each orbital period. Step 2: Perform preliminary task allocation based on the matching degree with different observation orbits of the new clustered target set; Step 3: Centrally schedule different observation tasks assigned to different orbits to maximize the observation targets.

[0009] Preferably, the input includes the scene start and end times, the set of observation targets, and satellite group parameters, and the output is the reselected observation center point and its included set of target points, observation time, observation satellites, and imaging attitude.

[0010] Preferably, the planning start and end times are: ,in and The start and end times for observation planning must be set; all observation tasks must be completed within the specified timeframes. The process is completed in the middle, and a suitable satellite is selected to observe the dense targets within this time frame; The input dense point target set is represented as , ; in, For the number of point targets, Indicates the first One point target, express longitude, express latitude, express Minimum required observation duration; The input star cluster is represented as , ; in, For the number of satellites, Indicates the first star in the star cluster One satellite; express The set of orbits, ,in, Indicates satellite exist The inner circle, The satellite orbit is represented using a fixed Earth coordinate system to indicate the position and velocity at any given time. for Maximum attitude maneuver angle; for Camera width; for attitude maneuver angular velocity; for The maximum observation time for a single orbit; The output task planning results are represented as follows: , ; in, For the number of tasks, Indicates the first One observation task; Indicates the execution of an observation task Satellite serial number; This represents the set of indexes of the covered point targets, i.e., the clustering of dense targets; Indicates to The center of the gaze; Indicates the moment when attitude maneuver begins; Indicates the moment when the attitude maneuver ends; Indicates the time the camera was turned on; Indicates the time the camera is turned off; Indicates the satellite attitude during the observation process. , ,in Indicates time, and For satellite exist The roll attitude angle and pitch attitude angle at any given moment.

[0011] Preferably, for any satellite in the constellation When performing a task Therefore, the constraints that need to be considered during the planning process are as follows: Attitude constraints: Tasks during the task execution period The gaze is always focused on the satellite. Within the field of view: (1) in, For a moment, ; Attitude transfer time constraints: The same satellite performing adjacent tasks sequentially and The attitude adjustment time should be less than Imaging start time and The time interval between the end of imaging: (2) (3) Imaging duration constraint: For any task, the observation duration should be greater than the minimum imaging time required for all targets within the cluster observed by that task. (4) in, ; Maximum observation time constraint per orbit: Due to on-board storage and data transmission plans, the total observation time performed by all satellites in the constellation during each orbit cannot exceed the maximum observation time. (5) in, For satellite The total number of tasks executed in a single cycle.

[0012] Preferably, the target point only needs to be observed once; repeated observations are not counted, i.e., the effective observation target is calculated as follows: (6) The overall goal of dense target observation is to maximize observation gains, and the effective observation targets should account for the largest proportion of the total number of targets: (7)

[0013] Preferably, the observed targets for clustering are recalculated within each orbital period, and the final clustering result is represented as follows: , , , , The number of orbits the constellation can complete during the mission period. Indicates the first The first orbital period within the [number] orbital cycle Task clustering; This indicates the center of the gaze point for this clustering task; This indicates the dense point targets included in the clustering task. ; Indicates the first The satellite in the The nadir trajectory under each orbital period, i.e., the nadir trajectory closest to the clustering task; During the first orbital period, the star cluster pairs are calculated. , and The access window, while in the second orbital period, only the star cluster pairs are calculated. , and Access window.

[0014] Preferably, based on the clustering results As input, the task allocation result is obtained and the output is... , , , Indicates the first The first orbital period within the [number] orbital cycle One observation task; The clustering target observed in this task; For the satellites that will carry out this mission; and Satellites right The start and end times of the observation window; For all observation tasks Add a 0-1 variable As a decision variable, it is used to determine whether the task should be executed. All variables determined by the task planning algorithm are... The task is to select the task to perform the observation. To obtain the final task planning results: .

[0015] A target rolling clustering task planning system based on sub-satellite point trajectories, the system comprising: The target clustering module aggregates multiple observation targets into one observation target and reselects the observation center location. It uses a dynamic clustering method to re-cluster the targets in each orbital period. The task allocation module performs preliminary task allocation on the clustered new target set according to the matching degree with different observation orbits; The task scheduling module centrally schedules different observation tasks assigned to different orbits to maximize the observation targets.

[0016] A computer-readable storage medium having a computer program stored thereon, which is executed by a processor to implement a target rolling clustering task planning method based on nadir point trajectories.

[0017] A computer device includes a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement a target rolling clustering task planning method based on nadir point trajectories.

[0018] The present invention has the following beneficial effects: This invention proposes a target rolling clustering observation task planning method based on nadir point trajectories. First, it analyzes and compares the clustering results with static clustering methods, demonstrating the rationality of the proposed dynamic clustering method. Then, it establishes a simulation scenario and employs different task planning algorithms to verify the effectiveness of task pre-allocation based on nadir point trajectories, significantly reducing planning time, alleviating target observation conflicts, and improving observation planning capabilities. Finally, by analyzing the planning results with different numbers of targets, it verifies that the target rolling clustering observation task planning method based on nadir point trajectories improves the observation task planning capability more effectively in scenarios with a larger number of targets. Attached Figure Description

[0019] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0020] Figure 1 The process shown is that of this invention; Figure 2 The inputs and outputs of this invention are shown below; Figure 3 Displayed as clustering based on nadir point trajectories; Figure 4 This is displayed as a clustering method based on nadir point trajectories; Figure 5 Displayed as a scrolling target cluster; Figure 6 The diagram shows the sliding of the imaging window. Figure 7 Displayed as a static clustering result distribution chart; Figure 8 The results are displayed as a distribution chart of dynamic clustering (period 1). Figure 9 The results are displayed as a distribution chart of dynamic clustering (period 2). Figure 10 This is a comparison chart of the silhouette coefficients of different clustering methods; Figure 11 This is a comparison chart showing the distribution of cluster centers using different clustering methods. Figure 12 The diagram shows the process of solving the observation task using a genetic algorithm. Figure 13 This chart shows a comparison of the iterative solution process for observation tasks using different planning algorithms. Figure 14 This diagram shows the process of solving the observation task using a genetic algorithm. Detailed Implementation

[0021] The technical solution of the present invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0022] The present invention will be described in detail below with reference to specific embodiments. Specific Implementation Example 1: The rapid increase in the number of low-Earth orbit remote sensing satellites has provided more options for Earth observation, but it has also increased the complexity of mission planning. To rationally and effectively arrange observation missions for large-scale remote sensing constellations, a target rolling clustering method based on nadir point trajectories is proposed. First, point targets that are close to each other and associated with the same nadir point trajectory are integrated into a single cluster of observation targets. Then, observation satellites are allocated according to the correlation between the clustered targets and satellites corresponding to different nadir point trajectories, achieving pre-allocation of observation tasks. Finally, task planning is performed on all processed clustered targets and individual targets, and conflicts between adjacent tasks are resolved using a sliding window. Since the nadir point trajectories move with the Earth's rotation, the targets clustered for each trajectory are updated rollingly. Simulation examples show that the clustering method of this invention can reasonably and effectively reduce the number of tasks, alleviate target observation conflicts, and reduce computation time; the more targets, the better the results.

[0024] according to Figures 1 to 14 As shown, the specific optimization technical solution adopted by the present invention to solve the above-mentioned technical problems is: The present invention relates to a target rolling clustering task planning method based on the nadir point trajectory.

[0025] This invention provides a target rolling clustering task planning method based on sub-satellite point trajectories, the method comprising the following steps: Step 1: Aggregate multiple observation targets into one observation target, and reselect the observation center location. Use dynamic clustering to re-cluster the targets in each orbital period. Step 2: Perform preliminary task allocation based on the matching degree with different observation orbits of the new clustered target set; Step 3: Centrally schedule different observation tasks assigned to different orbits to maximize the observation targets.

[0026] The inputs include the start and end times of the scene, the set of observation targets, and the satellite constellation parameters. The outputs are the newly selected observation center point and its set of target points, the observation time, the observed satellites, and the imaging attitude.

[0027] The planning start and end time is ,in and The start and end times for observation planning must be set; all observation tasks must be completed within the specified timeframes. The process is completed in the middle, and a suitable satellite is selected to observe the dense targets within this time frame; The input dense point target set is represented as , ; in, For the number of point targets, Indicates the first One point target, express longitude, express latitude, express Minimum required observation duration; The input star cluster is represented as , ; in, For the number of satellites, Indicates the first star in the star cluster One satellite; express The set of orbits, ,in, Indicates satellite exist The inner circle, The satellite orbit is represented using a fixed Earth coordinate system to indicate the position and velocity at any given time. for Maximum attitude maneuver angle; for Camera width; for attitude maneuver angular velocity; for The maximum observation time for a single orbit; The output task planning results are represented as follows: , ; in, For the number of tasks, Indicates the first One observation task; Indicates the execution of an observation task Satellite serial number; This represents the set of indexes of the covered point targets, i.e., the clustering of dense targets; Indicates to The center of the gaze; Indicates the moment when attitude maneuver begins; Indicates the moment when the attitude maneuver ends; Indicates the time the camera was turned on; Indicates the time the camera is turned off; Indicates the satellite attitude during the observation process. , ,in Indicates time, and For satellite exist The roll attitude angle and pitch attitude angle at any given moment.

[0028] For any satellite in the constellation When performing a task Therefore, the constraints that need to be considered during the planning process are as follows: Attitude constraints: Tasks during the task execution period The gaze is always focused on the satellite. Within the field of view: (1) in, For a moment, ; Attitude transfer time constraints: The same satellite performing adjacent tasks sequentially and The attitude adjustment time should be less than Imaging start time and The time interval between the end of imaging: (2) (3) Imaging duration constraint: For any task, the observation duration should be greater than the minimum imaging time required for all targets within the cluster observed by that task. (4) in, ; Maximum observation time constraint per orbit: Due to on-board storage and data transmission plans, the total observation time performed by all satellites in the constellation during each orbit cannot exceed the maximum observation time. (5) in, For satellite The total number of tasks executed in a single cycle.

[0029] The target point only needs to be observed once; repeated observations are not counted, meaning that the calculation of valid observation targets is as follows: (6) The overall goal of dense target observation is to maximize observation gains, and the effective observation targets should account for the largest proportion of the total number of targets: (7)

[0030] The observed targets are recalculated and clustered within each orbital period, and the final clustering result is represented as follows: , , , , The number of orbits the constellation can complete during the mission period. Indicates the first The first orbital period within the [number] orbital cycle Task clustering; This indicates the center of the gaze point for this clustering task; This indicates the dense point targets included in the clustering task. ; Indicates the first The satellite in the The nadir trajectory under each orbital period, i.e., the nadir trajectory closest to the clustering task; During the first orbital period, the star cluster pairs are calculated. , and The access window, while in the second orbital period, only the star cluster pairs are calculated. , and Access window.

[0031] Based on clustering results As input, the task allocation result is obtained and the output is... , , , Indicates the first The first orbital period within the [number] orbital cycle One observation task; The clustering target observed in this task; For the satellites that will carry out this mission; and Satellites right The start and end times of the observation window; For all observation tasks Add a 0-1 variable As a decision variable, it is used to determine whether the task should be executed. All variables determined by the task planning algorithm are... The task is to select the task to perform the observation. To obtain the final task planning results: .

[0032] This invention also provides a target rolling clustering task planning system based on sub-satellite point trajectories, the system comprising: The target clustering module aggregates multiple observation targets into one observation target and reselects the observation center location. It uses a dynamic clustering method to re-cluster the targets in each orbital period. The task allocation module performs preliminary task allocation on the clustered new target set according to the matching degree with different observation orbits; The task scheduling module centrally schedules different observation tasks assigned to different orbits to maximize the observation targets.

[0033] The present invention also provides a computer-readable storage medium having a computer program stored thereon, which is executed by a processor to implement a target rolling clustering task planning method based on nadir point trajectories.

[0034] The present invention also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement a target rolling clustering task planning method based on nadir point trajectory. Specific Implementation Example 2: The only difference between Embodiment 2 and Embodiment 1 of this application is that: With the development of remote sensing technology, the demand for observation has gradually increased, and it is often concentrated in certain hotspot areas. This type of problem can be called the Dense Target Observation Problem (DTOP). To solve the problem of observing dense targets from star clusters, it is first necessary to analyze the problem, define the input, output and process of the solution method, and explain in detail the key assumptions, constraints and objective function in the planning process.

[0036] according to Figure 1 The overall research process determined the inputs and outputs of the method of this invention, such as... Figure 2 As shown. The inputs include the scene start and end times, the set of observation targets, and satellite constellation parameters. The output is the reselected observation center point and its included set of target points, the observation time, the observed satellites, and the imaging attitude.

[0037] (1) The planning start and end time is ,in and Schedule start and end times for observations. All observation tasks need to be completed within [timeframe]. The algorithm needs to select appropriate satellites to observe dense targets within a given timeframe.

[0038] (2) The input dense point target set is represented as , , The number of point targets. Indicates the first One point target, express longitude, express latitude, express Minimum required observation duration.

[0039] (3) The input star cluster can be represented as , , This refers to the number of satellites. Indicates the first star in the star cluster One satellite; express The set of orbits, ,in Indicates satellite exist The inner circle, The satellite orbit is represented using a fixed Earth coordinate system to indicate the position and velocity at any given time. for Maximum attitude maneuver angle; for Camera width; for attitude maneuver angular velocity; for The maximum observation time for a single orbit.

[0040] (4) The output task planning result can be expressed as , , The number of tasks. Indicates the first One observation task; Indicates the execution of an observation task Satellite serial number; This represents the set of point target indices that can be covered, i.e., the clustering of dense targets; Indicates to The center of the gaze; Indicates the moment when attitude maneuver begins; Indicates the moment when the attitude maneuver ends; Indicates the time the camera was turned on; Indicates the time the camera is turned off; Indicates the satellite attitude during the observation process. , ,in Indicates time, and For satellite exist The roll attitude angle and pitch attitude angle at any given moment.

[0041] To focus on the research of target clustering, task allocation, and task planning algorithms, this invention makes the following assumptions about the satellite capabilities and observation task requirements in the constellation: (1) The z-axis of the satellite body system coincides with the optical axis of the camera, and the initial state of the satellite body system Z-axis coincides with the Z-axis of the orbit system.

[0042] (2) The attitude calculation adopts the attitude of this system relative to the orbital system, and the sequence is YPR. The yaw angle is not corrected.

[0043] (3) The target imaging geometric coverage of the cluster center point is considered as successful observation of all targets in the cluster, regardless of the actual image quality.

[0044] (4) The energy obtained by the satellite from the sun is sufficient to support the execution of the mission.

[0045] (5) Considering factors such as fuel, cost, and safety, it is assumed that the star cluster does not have orbital maneuvering capabilities.

[0046] (6) A single satellite can only perform one observation task at a time.

[0047] (7) Satellite mission data can be transmitted to the ground via a link, so only the maximum on-board storage constraint of a single track is considered.

[0048] Planning constraints For any satellite in the constellation If the task is executed Therefore, the constraints that need to be considered during the planning process are as follows: (1) Attitude constraints: during the task execution period The staring center needs to be able to always be on the satellite Within the field of view: (1) in, For a moment, .

[0049] (2) Attitude transfer time constraint: The same satellite performs adjacent tasks in succession. and The attitude adjustment time should be less than Imaging start time and The time interval between the end of imaging: (2) (3) (3) Imaging duration constraint: For any task, the observation duration should be greater than the minimum imaging time required for all targets within the cluster observed by that task. (4) in, .

[0050] (4) Maximum observation time constraint per orbit: Due to on-board storage and data transmission plans, the total observation time performed by all satellites in the constellation during each orbit cannot exceed the maximum observation time. (5) in, For satellite The total number of tasks executed in a single cycle.

[0051] objective function Because target clustering is dynamic, the same target may be assigned to different clusters in subsequent clustering processes. Accordingly, the final task planning involves observing these dynamically changing clustering targets, which may result in repeated observations of some target points. However, target points only need to be observed once; therefore, repeated observations should not be included in the calculation of effective observation targets. (6) The overall goal of dense target observation is to maximize observation gains, and the effective observation targets should account for the largest proportion of the total number of targets: (7) In this section, we discuss the shortcomings of existing research methods and propose a target rolling clustering method based on nadir point trajectories. Based on this method, we pre-assign tasks to clustering targets and design a task planning algorithm to solve the problem.

[0052] Task clustering model Existing research methods rely solely on the geographical similarity of dense targets for clustering. Because the clustering process only considers the inclusion criteria of non-clustered points, the clustering results are affected by the order in which points are grouped during algorithm execution. For example... Figure 3 As shown, the clustering radius defined by the algorithm is visible. , , , Each can be aggregated into an observation task. Assume the order of the point sets processed by the algorithm during the clustering process is as follows: ,but and It can be aggregated into ,and Obviously not satisfied Clustering conditions, but and With clustering as The conditions are as follows. Different clustering results will affect the efficiency of subsequent task planning, so the randomness of the results caused by clustering algorithms is unacceptable.

[0053] In the clustering method of this invention, the set of points clustered into one observation task is not fixed but dynamically changing. This involves the conditions for unclustered points to enter existing cluster centers and the conditions for clustered points to leave existing cluster centers, requiring the design of reasonable judgment conditions. Since the observation quality is better when the satellite is passing overhead, the entry and exit of different points into the same cluster will affect the position of the cluster center. Therefore, this invention uses the distance between the cluster center and the trajectory of the nadir points as the evaluation criterion. The clustering process is as follows: Figure 4 As shown.

[0054] Because star clusters create different nadir trajectories on the ground during observation missions, the clustering results of dense targets remain constant within the same orbital period but change dynamically across different orbital periods. For example... Figure 5 As shown, For different satellites of the constellation In the The nadir trajectories within each orbital period. It can be seen that points that could cluster into the same observation task within the first orbital period no longer meet the aggregation conditions for the nadir trajectories formed in the new orbital period. Therefore, it is necessary to recalculate the observation targets that can be clustered within each orbital period. The final clustering result can be represented as... , , , , The number of orbits that the constellation can complete within the mission period. Indicates the first The first orbital period within the [number] orbital cycle Task clustering; This indicates the center of the gaze point for this clustering task; This indicates the dense point targets included in the clustering task. ; Indicates the first The satellite in the The nadir trajectory under each orbital period is the nadir trajectory that the clustering task is closest to.

[0055] Task allocation method After dynamic clustering, dense targets yield new gaze centers. Based on the constellation's satellite orbits, attitude maneuver constraints, swath width, and target positions, the access windows of all satellites to the cluster centers can be calculated. Previous work, due to the static nature of clustering, typically involved directly calculating the access windows of all satellites to the cluster centers throughout the entire mission cycle.

[0056] Within different orbital periods, the location of the cluster centers in the dynamic clustering method proposed in this invention changes dynamically. Therefore, calculating all access windows throughout the entire mission cycle is redundant; it is only necessary to calculate the access windows of all satellites to the cluster centers within each orbital period. Figure 5 As shown, during the first orbital period, it is necessary to calculate the star cluster pairs. , and The access window, and during the second orbital period, due to the update of cluster centers, the calculation of star cluster pairs... The access window is meaningless, so only the star cluster pairs need to be calculated. , and Access window.

[0057] Within the same orbital period, since the minimum distance between the cluster center and the nadir point trajectory is calculated, a nadir point trajectory with the smallest imaging angle can be matched for each clustering task, thus determining the optimal observation satellite. By using this method to allocate observation targets within the same orbital period, the observation task sequence of each satellite in each orbital period can be determined.

[0058] Therefore, based on the clustering results As input, the task allocation result can be obtained and the output is... , , . Indicates the first The first orbital period within the [number] orbital cycle One observation task; The clustering target observed in this task; For the satellites that will carry out this mission; and Satellites right The start and end times of the observation window.

[0059] Task planning algorithm Because the task clustering process is dynamic, the same target may be clustered into different observation tasks within different orbital periods. Directly using task allocation methods would result in redundant observation tasks, and due to limitations in onboard memory and data transmission plans, it is impossible for all scheduled observation tasks within the same orbital period to be executed. Centralized planning is needed to determine the optimal observation sequence. Therefore, for all observation tasks... Add a 0-1 variable These are used as decision variables to determine whether the task should be performed. All variables determined using task planning algorithms... The task is to select the task to perform the observation. Then the final task planning result can be obtained. , , and It can be done get, , , and The calculation can be performed by sliding the imaging window

[32] , as shown in the figure. Attitude can be calculated using satellite staring; the specific calculation method has been explained in relevant works and will not be repeated here.

[0060] Simulation Analysis Clustering results comparison To verify the advantages of the proposed clustering method compared to traditional geographic similarity clustering methods, a simulation scenario was designed for verification. The simulation scenario ran from 00:00:00 on October 9, 2025 to 00:00:00 on October 10, 2025. The scenario used a Walker constellation of 24 satellites as low-Earth orbit remote sensing satellites to perform observation tasks. The constellation had 8 orbital planes and a phase factor of 2. The orbital parameters and camera capabilities of the seed satellites in the Walker constellation are shown in Table 1. For attitude maneuver angular velocity, For the maximum maneuver angle, The camera's field of view is defined as follows. The locations of the target points were randomly generated, with 2000 target points randomly generated within the region of 80°–140° E and 10°–60° N. Furthermore, to simplify the calculation of dynamic clustering, the dynamic clustering within the simulation scene was divided into 14 dynamic clustering cycles of 100 minutes each, based on the constellation orbital altitude and period. This means that the clustering results are updated every 100 minutes based on the current nadir point trajectory.

[0061] Table 1. Satellite Orbit Parameters

[0062] Based on the methods provided in existing research, 2000 target points in the above scenario were clustered, resulting in 531 clustered observation targets, such as... Figure 7 As shown. Using the dynamic clustering method proposed in this invention, different clustering observation targets were obtained within 14 periods, such as... Figure 8 and Figure 9 As shown in Table 2, the number of clusters varies in different periods.

[0063] Table 2. Comparison of the number of clustering results

[0064] Clustering performance comparison This invention uses the profile coefficient and the distance from the cluster center to the nearest sub-satellite point to analyze the performance of different clustering methods.

[0065] The silhouette coefficient is typically used to measure the tightness of each sample with its own cluster and its separation from the nearest cluster. It combines the evaluation of cluster tightness and inter-cluster separation, with results ranging from [-1, 1], where a higher value indicates better clustering performance. By randomly generating multiple groups of points with varying numbers within a region, performing clustering, and calculating the average silhouette coefficient, the result is obtained. Figure 10 As can be seen, dynamic clustering is always superior to static clustering. Moreover, when the number of target points exceeds 2000, the results obtained by the sub-point rolling clustering are significantly better than those obtained by static clustering. That is, the more the number of target points, the more obvious the advantage of dynamic clustering becomes.

[0066] Since the observation quality is best when the target is directly overhead on a remote sensing satellite, this invention uses the distance from the cluster center to the nearest nadir point to analyze the observation quality that the task generated through clustering can obtain at the optimal observation position. That is, the smaller the distance from the cluster center to the nearest nadir point, the better the optimal observation quality that the clustering task can obtain. Figure 11 It can be seen that the distance from the cluster center to the nearest sub-star point in the dynamic clustering method is always less than that in the static clustering method, and the advantage of the dynamic clustering method becomes more obvious as the number of target points increases.

[0067] Task planning Based on target rolling clustering based on nadir point trajectories, in order to verify the improvement effect of the rolling clustering and task allocation method designed in this invention on actual task planning, simulation verification was carried out on the designed scenario. The observation task requires a minimum observation time of 15 seconds for each target and a maximum observation duration of 500 seconds for a single satellite orbit.

[0068] Comparison of observation task planning results of different clustering methods This invention first employs a genetic algorithm to plan the observation task for clustering targets obtained by different clustering methods. The genetic algorithm parameters are set identically, differing only in the input observation window. The population size is 30, the maximum number of iterations is 150, the crossover probability is 0.8, and the mutation probability is 0.01. The task planning solution process and results are shown in Tables 3 and 4, respectively. Figure 12 As shown, it can be seen that using dynamic clustering and pre-assigning tasks can significantly reduce the number of input tasks, allowing more task objectives to be observed with fewer observation tasks. This indirectly proves the rationality of the dynamic clustering results and significantly improves the convergence speed of the task planning algorithm, thus reducing the task planning time.

[0069] Table 3. Comparison of observation task planning results for different clustering methods

[0070] Comparison of task planning results of different optimization algorithms A genetic algorithm was used to plan tasks based on the results of different clustering methods. Using the same simulation scenario and task input, differential evolution and simulated annealing methods were employed to analyze the task results generated by the proposed target rolling clustering observation task planning method based on nadir point trajectories using different planning algorithms, as shown in Table 4. Figure 13 As shown, when using different algorithms to plan the same task, the method proposed in this invention has significant advantages in both convergence speed and planning results.

[0071] Table 4. Comparison of Task Planning Results of Different Optimization Algorithms

[0072] Comparison of task planning results with different numbers of objectives Having analyzed the observation planning results when the number of target points was 2000, we will increase the number of targets to 4000 here to analyze the task planning results under different numbers.

[0073] First, target clustering was performed. A total of 630 clustered observation targets were obtained using the static clustering method. Then, the dynamic clustering method proposed in this invention was used to obtain different clustered observation targets in 14 periods, as shown in Table 5.

[0074] Table 5. Comparison of the number of clustering results

[0075] Subsequently, based on the results, a more effective genetic algorithm was used to solve the observation task planning problem for the clustering target. The genetic algorithm parameters were set identically, differing only in the input observation window: population size of 30, maximum number of iterations of 150, crossover probability of 0.8, and mutation probability of 0.01. The task planning process and results are shown in Tables 6 and 6, respectively. Figure 14 As shown in the figure. In comparison, it can be seen that using dynamic clustering and pre-assigning tasks can significantly reduce the number of input tasks, allowing more task objectives to be observed with fewer observation tasks. Moreover, the improvement of the planning results by the proposed method becomes more significant as the number of objectives increases.

[0076] Table 6. Comparison of observation task planning results of different clustering methods

[0077] The above description is merely a preferred embodiment of a target rolling clustering task planning method based on nadir point trajectories. The scope of protection for this method is not limited to the above embodiments; all technical solutions falling within this framework are within the scope of protection of this invention. It should be noted that for those skilled in the art, any improvements and variations made without departing from the principles of this invention should also be considered within the scope of protection of this invention.

Claims

1. A target rolling clustering task planning method based on sub-satellite point trajectories, characterized by: The method includes the following steps: Step 1: Aggregate multiple observation targets into one observation target, and reselect the observation center location. Use dynamic clustering to re-cluster the targets in each orbital period. Step 2: Perform preliminary task allocation based on the matching degree with different observation orbits of the new clustered target set; Step 3: Centrally schedule different observation tasks assigned to different orbits to maximize the observation targets.

2. The method according to claim 1, characterized in that: The inputs include the start and end times of the scene, the set of observation targets, and the satellite constellation parameters. The outputs are the newly selected observation center point and its set of target points, the observation time, the observed satellites, and the imaging attitude.

3. The method according to claim 2, characterized in that: The planning start and end time is ,in and The start and end times of the observation plan; All observation tasks need to be performed The process is completed in the middle, and a suitable satellite is selected to observe the dense targets within this time frame; The input dense point target set is represented as , ; in, For the number of point targets, Indicates the first One point target, express longitude, express latitude, express Minimum required observation duration; The input star cluster is represented as , ; in, For the number of satellites, Indicates the first star in the star cluster One satellite; express The set of orbits, ,in, Indicates satellite exist The inner circle, The satellite orbit is represented using a fixed Earth coordinate system to indicate the position and velocity at any given time. for Maximum attitude maneuver angle; for Camera width; for attitude maneuver angular velocity; for The maximum observation time for a single orbit; The output task planning results are represented as follows: ， ； in, For the number of tasks, Indicates the first One observation task; Indicates the execution of an observation task Satellite serial number; This represents the set of indexes of the covered point targets, i.e., the clustering of dense targets; Indicates to The center of the gaze; Indicates the moment when attitude maneuver begins; Indicates the moment when the attitude maneuver ends; Indicates the time the camera was turned on; Indicates the time the camera is turned off; Indicates the satellite attitude during the observation process. , ,in Indicates time, and For satellite exist The roll attitude angle and pitch attitude angle at any given moment.

4. The method according to claim 3, characterized in that: For any satellite in the constellation When performing a task Therefore, the constraints that need to be considered during the planning process are as follows: Attitude constraints: Tasks during the task execution period The gaze is always focused on the satellite. Within the field of view: (1) in, For a moment, ; Attitude transfer time constraints: The same satellite performing adjacent tasks sequentially and The attitude adjustment time should be less than Imaging start time and The time interval between the end of imaging: (2) (3) Imaging duration constraint: For any task, the observation duration should be greater than the minimum imaging time required for all targets within the cluster observed by that task. (4) in, ; Maximum observation time constraint per orbit: Due to on-board storage and data transmission plans, the total observation time performed by all satellites in the constellation during each orbit cannot exceed the maximum observation time. (5) in, For satellite The total number of tasks executed in a single cycle.

5. The method according to claim 4, characterized in that: The target point only needs to be observed once; repeated observations are not counted, meaning that the calculation of valid observation targets is as follows: (6) The overall goal of dense target observation is to maximize observation gains, and the effective observation targets should account for the largest proportion of the total number of targets: (7)。 6. The method according to claim 5, characterized in that: The observed targets are recalculated and clustered within each orbital period, and the final clustering result is represented as follows: , , , , The number of orbits the constellation can complete during the mission period. Indicates the first The first orbital period within the [number] orbital cycle Task clustering; This indicates the center of the gaze point for this clustering task; This indicates the dense point targets included in the clustering task. ; Indicates the first The satellite in the The nadir trajectory under each orbital period, i.e., the nadir trajectory closest to the clustering task; During the first orbital period, the star cluster pairs are calculated. , and The access window, while in the second orbital period, only the star cluster pairs are calculated. , and Access window.

7. The method according to claim 6, characterized in that: Based on clustering results As input, the task allocation result is obtained and the output is... , , , Indicates the first The first orbital period within the [number] orbital cycle One observation task; The clustering target observed in this task; For the satellites that will carry out this mission; and Satellites right The start and end times of the observation window; For all observation tasks Add a 0-1 variable As a decision variable, it is used to determine whether the task should be executed. All variables determined by the task planning algorithm... The task is to select the task to perform the observation. To obtain the final task planning results: .

8. A target rolling clustering task planning system based on sub-satellite point trajectories, characterized by: The system includes: The target clustering module aggregates multiple observation targets into one observation target and reselects the observation center location. It uses a dynamic clustering method to re-cluster the targets in each orbital period. The task allocation module performs preliminary task allocation on the clustered new target set according to the matching degree with different observation orbits; The task scheduling module centrally schedules different observation tasks assigned to different orbits to maximize the observation targets.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, The program is executed by the processor to implement the method as claimed in any one of claims 1-7.

10. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that: When the processor executes the computer program, it implements the method of any one of claims 1-7.

Images