A fast reconfigurable multi-satellite multi-payload cooperative method and system for ecological monitoring
By developing a rapid and reconfigurable multi-satellite, multi-payload collaborative method and system for ecological monitoring, and utilizing agent deep reinforcement learning and target neural networks, we have achieved precise matching of ecological elements and payload types and collaborative scheduling of multiple payloads. This solves the problems of low resource utilization and weak dynamic response in multi-satellite monitoring technology, and improves monitoring efficiency and data accuracy.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- XIDIAN UNIV
- Filing Date
- 2025-11-17
- Publication Date
- 2026-08-04
AI Technical Summary
Existing multi-satellite monitoring technologies suffer from low resource utilization and weak dynamic response, making it difficult to meet the routine and sudden needs of ecological monitoring. They also lack a global control strategy and cannot achieve multi-satellite collaborative observation.
Design a rapid, reconfigurable multi-satellite, multi-payload collaborative method and system for ecological monitoring. By combining a monitoring and reconfiguration request module, a detection and reconfiguration control module, and a task execution module with agent deep reinforcement learning and target neural network, the system achieves accurate matching of ecological elements and payload types and collaborative scheduling of multiple payloads.
It enables the rapid generation of scheduling results during sudden ecological disasters, ensuring the precise allocation of satellite resources and payloads, improving the accuracy and completeness of monitoring data, increasing the utilization efficiency of satellite resources, and solving the problems of resource idleness and duplicate observations.
Smart Images
Figure CN121530450B_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of mission planning technology, specifically relating to a rapid, reconfigurable multi-satellite, multi-payload collaborative method and system for ecological monitoring. Background Technology
[0002] my country has a complex climate and ecosystem, and some areas (such as the plateau permafrost region) are inherently fragile. It is necessary to not only conduct routine monitoring to understand ecological evolution, but also to conduct rapid emergency observation for sudden ecological disasters such as forest fires and floods.
[0003] Currently, low-orbit remote sensing satellite payloads cover optical, infrared, and SAR types. Multi-satellite joint Earth observation has become the mainstream monitoring method due to its high spatiotemporal resolution and high observation frequency. However, existing multi-satellite monitoring technologies have significant defects, such as low resource utilization (prone to duplicate observations or idle resources), weak dynamic response (difficult to quickly adjust observation plans to cope with sudden tasks), and lack of global management strategies (routine and sudden tasks are managed separately, making it impossible to coordinate and schedule satellite resources). They are difficult to adapt to the combined needs of "normalization + suddenness" in ecological monitoring. Therefore, designing a rapid and reconfigurable multi-satellite and multi-payload collaborative method for ecological monitoring has become an urgent technical problem to be solved. Summary of the Invention
[0004] To address the aforementioned problems in existing technologies, this invention provides a rapid, reconfigurable multi-satellite, multi-payload collaborative method and system for ecological monitoring. The technical problem to be solved by this invention is achieved through the following technical solution: In a first aspect, the present invention provides a rapidly reconfigurable multi-satellite, multi-payload collaborative system for ecological monitoring, comprising: The monitoring reconstruction request module includes an ecological element extraction submodule and a task information request submodule. The ecological element extraction submodule is used to extract the ecological elements to be observed based on the ecosystem of the target area being monitored. The task information request submodule is used to summarize the time window resources and imaging strip resources required by the task and output the first task element of each task. The detection and reconstruction control module includes a sudden ecological element analysis submodule, a payload matching submodule, and a reconstruction execution submodule. The sudden ecological element analysis submodule is used to analyze the ecological elements to be observed and output the disaster type occurring in the target area. The payload matching submodule is used to match the corresponding imaging payload type according to the disaster type occurring in the target area and output the second task element of each task. The reconstruction execution submodule is used to output the task scheduling sequence according to the first task element and the second task element of each task. The task execution module includes a scheduling instruction uploading submodule and a task execution submodule. The scheduling instruction uploading submodule is used to generate scheduling instructions from the task scheduling sequence and send observation instructions to the corresponding satellites through uploading. The task execution submodule is used to control the corresponding satellites to perform routine or emergency observation tasks according to the observation instructions.
[0005] Secondly, this invention also provides a rapid, reconfigurable, multi-satellite, multi-payload collaborative method for ecological monitoring, comprising: Monitor the ecosystem of the target area and extract the ecological elements to be observed; Based on the ecological elements to be observed, describe the task scenario, obtain the time window resources and imaging strip resources required for the task, and construct the first task elements for each task. The types of disasters occurring in the target area are obtained by analyzing the ecological elements to be observed. Based on the type of disaster occurring in the target area, the corresponding imaging payload type is matched to obtain the second mission elements for each task; Based on the first task element and the second task element of each task, we obtain the observation task with complete task elements. We then use the trained model to process the observation task with complete task elements to obtain the task scheduling decision. Based on the task scheduling decision, we obtain the task scheduling sequence. The scheduling sequence of the mission is used to generate scheduling instructions, and observation instructions are sent to the corresponding satellites via uploading. According to the observation instructions, the corresponding satellite performs routine or emergency observation tasks.
[0006] The beneficial effects of this invention are: This invention provides a rapid, reconfigurable, multi-satellite, multi-payload collaborative method and system for ecological monitoring. By describing the scheduling process of sudden ecological monitoring tasks as a deep reinforcement learning process of an intelligent agent, the scheduling model fully considers the functional characteristics and applicable scenarios of different payloads such as optical, infrared, and SAR, and establishes a precise correspondence mechanism between ecological elements and payload types. Combined with a target neural network, a sudden ecological observation task scheduling model adapted to low-Earth orbit remote sensing satellite constellation scenarios is constructed. When a sudden ecological observation task arrives, the system can quickly generate scheduling results, ensuring the accurate allocation of satellite resources and adapted payloads in a short time. It can also achieve multi-payload collaborative complementarity, avoiding the problem of single payloads being limited by environmental factors such as illumination, and ensuring the accuracy and integrity of monitoring data.
[0007] The present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description
[0008] Figure 1This is a schematic diagram of a rapidly reconfigurable multi-satellite multi-payload collaborative system for ecological monitoring provided in an embodiment of the present invention; Figure 2 This is a flowchart of a rapid, reconfigurable, multi-satellite, multi-payload collaborative method for ecological monitoring provided in an embodiment of the present invention; Figure 3 This is a schematic diagram of a task scenario provided in an embodiment of the present invention; Figure 4 This is a flowchart illustrating the time window resources and imaging strip resources required for a computational task provided in an embodiment of the present invention. Figure 5 This is a flowchart of a training model provided in an embodiment of the present invention. Detailed Implementation
[0009] The present invention will be further described in detail below with reference to specific embodiments, but the implementation of the present invention is not limited thereto.
[0010] The following technical problems urgently need to be solved in the existing technology.
[0011] First, ecological monitoring, as a key area of Earth observation, has specific and clear imaging requirements. However, in multi-satellite Earth observation systems, the orbital parameters and imaging payloads of different satellites vary significantly. Furthermore, when facing observation missions covering multiple large-scale ecological regions, issues such as overlapping coverage or uneven observation are easily encountered, directly leading to the underutilization of existing satellite resources. Therefore, how to rationally allocate satellite resources according to the specific needs of different ecological monitoring missions and achieve multi-satellite collaborative observation of ecological regions has become a critical issue that urgently needs to be addressed.
[0012] Second, existing multi-satellite monitoring technologies lack a precise matching mechanism between "ecological elements and payload types," and do not fully consider the differences in characteristics and collaborative adaptation requirements of different types of payloads such as optical, infrared, and SAR. This fails to meet the detection needs of different ecological elements requiring specific imaging payloads, and also struggles to address the limitations of single payloads due to environmental factors such as weather (e.g., optical payloads cannot effectively image in the absence of light). This necessitates the design of imaging mission planning methods for emergency ecological monitoring that can achieve collaborative scheduling of multiple payload types based on different ecological elements and lighting conditions. Therefore, there is an urgent need to design an imaging mission planning method for emergency ecological monitoring that can achieve collaborative scheduling of multiple payload types based on the needs of different ecological elements and different environmental conditions.
[0013] Third, ecological monitoring emergencies are characterized by their suddenness and urgency, requiring satellites to respond quickly and adjust their observation tasks. Currently, most scheduling methods for these emergencies often employ conventional heuristic algorithms and fail to adequately consider that some satellites may not be at ideal observation angles in multi-satellite collaborative observations across multiple ecological regions. This leads to poor imaging quality, low coverage, low satellite resource utilization, and overall unsatisfactory observation results. Therefore, a planning method for emergency ecological monitoring tasks must be designed to jointly optimize both emergency and routine observation tasks.
[0014] Please see Figure 1 , Figure 1 This is a schematic diagram of a rapidly reconfigurable multi-satellite, multi-payload collaborative system for ecological monitoring provided in an embodiment of the present invention. The rapidly reconfigurable multi-satellite, multi-payload collaborative system for ecological monitoring provided by the present invention includes: The monitoring reconstruction request module includes an ecological element extraction submodule and a task information request submodule. The ecological element extraction submodule is used to extract the ecological elements to be observed based on the ecosystem of the target area being monitored. The task information request submodule is used to summarize the time window resources and imaging strip resources required by the task and output the first task element of each task. The detection and reconstruction control module includes a sudden ecological element analysis submodule, a payload matching submodule, and a reconstruction execution submodule. The sudden ecological element analysis submodule is used to analyze the ecological elements to be observed and output the disaster type occurring in the target area. The payload matching submodule is used to match the corresponding imaging payload type according to the disaster type occurring in the target area and output the second task element of each task. The reconstruction execution submodule is used to output the task scheduling sequence according to the first task element and the second task element of each task. The task execution module includes a scheduling instruction uploading submodule and a task execution submodule. The scheduling instruction uploading submodule is used to generate scheduling instructions from the task scheduling sequence and send observation instructions to the corresponding satellites through uploading. The task execution submodule is used to control the corresponding satellites to perform routine or emergency observation tasks according to the observation instructions.
[0015] For details, please continue to see Figure 1 In this embodiment, the monitoring and reconstruction request module is responsible for generating observation task requests, including an ecological element extraction submodule and a task information request submodule. The ecological element extraction submodule automatically extracts the ecological elements to be observed based on the ecosystem of the target area to be monitored. The task information request submodule summarizes information such as the time, target area, and task type required for the observation task. The two parts of information are sent together to the monitoring and reconstruction control module to generate the observation task.
[0016] The detection and reconstruction control module is responsible for further analyzing task elements, generating observation tasks, and planning tasks. It includes a sudden ecological element analysis submodule, a payload matching submodule, and a reconstruction execution submodule. The sudden ecological element analysis submodule is responsible for analyzing the ecological elements input by the ecological element extraction submodule to obtain the disaster type occurring in the target area, and sending the results to the payload matching submodule. The payload matching submodule outputs the corresponding imaging payload type according to the disaster type based on the output results of the sudden ecological element analysis submodule. The reconstruction execution submodule has the functions of generating observation tasks and planning tasks.
[0017] The task execution module is responsible for scheduling instruction uploading and task execution, including a scheduling instruction uploading submodule and a task execution submodule. The scheduling instruction uploading submodule generates scheduling instructions based on the task scheduling results output from the reconstruction execution submodule, and sends observation instructions to the corresponding satellites through instruction uploading, thereby completing the reconstruction of the monitoring network. The task execution submodule controls the satellites to perform routine and emergency observation tasks according to the observation instructions. While performing routine observation tasks, it can also monitor the target area. If a sudden ecological disaster occurs, it will notify the monitoring reconstruction request module to start a new round of monitoring network reconstruction.
[0018] In an optional embodiment of the present invention, the reconstructed execution submodule includes a task generation unit, a model loading unit, and a task planning unit. The task generation unit is used to output an observation task with complete task elements based on the first task elements and the second task elements of each task. The model loading unit is used to process the observation task with complete task elements using a trained model and output a task scheduling decision. The task planning unit is used to output a task scheduling sequence based on the task scheduling decision.
[0019] In an optional embodiment of the present invention, the task execution submodule includes a sudden task execution unit and a regular task execution unit. The sudden task execution unit is used to execute sudden observation tasks, and the regular task execution unit is used to execute regular observation tasks. At the same time, the sudden ecological disaster is fed back to the detection and reconstruction control module to initiate a new monitoring reconstruction.
[0020] Based on the same inventive concept, please refer to Figure 2 , Figure 2 This is a flowchart of a rapid reconfigurable multi-satellite, multi-payload collaborative method for ecological monitoring provided in an embodiment of the present invention. The present invention also provides a rapid reconfigurable multi-satellite, multi-payload collaborative method for ecological monitoring, applied to the rapid reconfigurable multi-satellite, multi-payload collaborative device for ecological monitoring provided in the above embodiments of the present invention. Embodiments of the device can be referred to above and will not be repeated here. The method includes: S101. Monitor the ecosystem of the target area and extract the ecological elements to be observed.
[0021] Specifically, in this embodiment, the ecological elements to be observed can be deserts, saline land, rivers and lakes, forests, etc.
[0022] S102. Based on the ecological elements to be observed, describe the task scenario, obtain the time window resources and imaging strip resources required for the task, and construct the first task elements for each task.
[0023] Specifically, in this embodiment, the task scenario is described, including: Set the satellite's orbital plane, the number of satellites on each plane, the satellite's altitude and orbital inclination, and the target area to describe the mission scenario; among these, The satellite is represented as: ; in, Indicates the total number of satellites. Indicates the satellite index; The target region is represented as: ; in, Indicates the total number of target areas; against ;in, , in, Represents the coordinates of the boundary points of the target region. Indicates the type of ecological elements in the target area. Indicates the types of disasters that may occur in the target area. This indicates the probability of each disaster occurring. The target region is divided into sub-regions, denoted as... , This indicates the number of sub-regions divided into corresponding target regions. Indicates the first division The coordinates of the boundary points of each sub-region Indicates the index of the target region.
[0024] Optional, such as Figure 3 As shown, Figure 3This is a schematic diagram of a mission scenario provided by an embodiment of the present invention. The mission scenario consists of low Earth Orbit (LEO) satellites and a target area. The LEO satellites include 10 orbital planes, with 15 satellites in each plane, at an altitude of 500 km and an orbital inclination of 53°, totaling 150 satellites. Three target areas are defined, each a square area ranging from 2° to 3° in size. Each satellite carries an imaging payload responsible for completing regional observations for both routine and emergency observation tasks. The ground area represents the ecological environment that needs daily monitoring, used to generate routine and emergency observation tasks. Routine observation tasks require a large observation area, have a fixed generation time, and a long duration; emergency observation tasks require a random observation area, have a random generation time, and a short duration. The LEO satellites push-broom their nadir trajectory to form an imaging strip to complete the imaging work. By laterally swinging, they can deviate from the nadir trajectory to image, thereby expanding their imaging coverage. In this mission scenario, because the target area is relatively large, it is difficult for a single satellite to cover the entire target area; therefore, multiple satellites are needed to perform multiple imaging operations to complete the observation of the entire target area.
[0025] Further, please see Figure 4 , Figure 4 This is a flowchart illustrating the time window resources and imaging strip resources required for a computational task according to an embodiment of the present invention. In this embodiment, the time window resources and imaging strip resources required for the task are obtained to construct the first task elements for each task, including: Using SKT, the visible time windows of all satellites and the target area are calculated, along with the nadir trajectories of the satellites within each visible time window. Each satellite can have one visible time window calculated for each mission, and within each visible time window, the satellite corresponds to one nadir trajectories. The visible time window is represented as follows: ; in, Indicates the total number of visible time windows; against ;in, ; in, Indicates the visible time window The corresponding target area Indicates the visible time window The corresponding satellite number, Indicates the visible time window The corresponding start time, Indicates the visible time window The corresponding start time and satellite nadir coordinates. Indicates the visible time window The corresponding end time, Indicates the visible time window The corresponding end time is the satellite's nadir coordinates. Indicates the visible time window The corresponding satellite's payload, The index representing the visible time window; Obtain the nadir coordinates of the start and end times of the visible time window corresponding to the satellite; According to the preset side view range and side view discrete angle The discrete side view sequence is calculated; Traverse the discrete side view sequence, for the current side view Calculate the vertex coordinates of the corresponding strip, expressed as: ; ; ; ; in, Indicates the satellite's field of view. , These represent the angles on either side of the strip corresponding to the current side view. Indicates the satellite imaging angle. Indicates the observation angle. Indicates the satellite's coverage angle. This indicates the coverage angle corresponding to the left strip of the strip at the current side view. Represents the Earth's radius. Indicates satellite altitude. Indicates satellite Latitude and longitude coordinates at the current moment Indicates the orbital inclination of the satellite. Indicates satellite The longitude of the vertex of the left band of the time strip. Indicates satellite The latitude of the leftmost apex of the band at time t. Iterate through all visible time windows until you obtain the vertex coordinates of all stripes corresponding to each visible time window; where a stripe is represented as: ; in, This represents the total number of stripes corresponding to the visible time window.
[0026] S103. Analyze the ecological elements to be observed to obtain the types of disasters occurring in the target area.
[0027] Specifically, in this embodiment, based on the ecological elements to be observed, the types of disasters that may occur in the target area are predicted. For example, deserts may experience drought, forests may experience fires, and rivers and lakes may experience floods.
[0028] S104. Based on the type of disaster occurring in the target area, match the corresponding imaging payload type to obtain the second mission element for each task.
[0029] Specifically, in this embodiment, an appropriate imaging payload is selected based on the types of disasters that may occur in the target area. For example, SAR imaging is suitable for drought, infrared imaging is suitable for fire, and visible light imaging is suitable for flood.
[0030] S105. Based on the first task element and the second task element of each task, an observation task with complete task elements is obtained. The trained model is used to process the observation task with complete task elements to obtain a task scheduling decision. Based on the task scheduling decision, a task scheduling sequence is obtained.
[0031] Specifically, in this embodiment, the observation task with complete task elements is represented as: ; in, This indicates the total number of observation tasks. An index indicating the observation task; against ;in, ; in, Indicates the first Observation tasks The number, Indicates the first Observation tasks The types are divided into routine observation missions and emergency observation missions. Indicates the first Observation tasks The target area to be observed Indicates the first Observation tasks Vertex coordinates of the target region Indicates the first Observation tasks The start time, Indicates the first Observation tasks Duration, Indicates the first Observation tasks Required observation payload type Indicates the first Observation tasks The minimum required imaging time, Indicates the first Observation tasks Required coverage Indicates the first Observation tasks Basic priority, Indicates the first Observation tasks The completion mark.
[0032] In this embodiment, an agent network is used to generate task scheduling decisions. The agent network consists of a predictive neural network and a target neural network, and network training involves optimizing the parameters of these two types of networks. Since deep reinforcement learning requires the training of the above network to be based on the Markov decision process framework, the construction process of the system's mathematical model will be explained before specifically describing the training method of the agent network.
[0033] Based on the task scenario, construct the system objective function and constraints, including: Introducing Boolean logic variables The time for the satellite to perform its imaging mission is determined and expressed as: ; in, Indicates the first One observation task, Indicates the visible time window The next Each stripe This indicates that the visible time window corresponding to the selected strip is , Indicates the visible time window The selected strip number; According to Boolean logic variables The coverage of the target region by the stripe to the sub-regions is calculated, and a Boolean variable is introduced. , representing the coverage status of the sub-regions of the target region, is expressed as: ; in, This indicates that after the target region is discretized into small grids, the first... Line 1 Column grid cells; According to the Boolean variable Determine the observation task Target area coverage , is represented as: ; Get the total coverage of all tasks , is represented as: ; Considering the execution of sudden observation tasks and their timeliness, the system objective function is described as follows, based on the total coverage of all tasks: ; in, Indicates the number of emergency observation tasks. Indicates the index of the observation task. Indicates the number of routine observation tasks. Index indicating routine observation tasks, Indicates an emergency observation mission The response time, i.e., the start time for selecting a satellite to begin the imaging task, Indicates the start time of an emergency observation mission; Constraints are crucial for ensuring the feasibility and accuracy of mission planning results. To ensure the successful completion of both emergency and routine observation missions, it is necessary to ensure that the satellite's payload type meets the mission payload requirements, and that the selected imaging opportunities satisfy the minimum imaging time required by the mission. The constraints of the objective function are defined as follows: This indicates that when performing a mission, the type of payload carried by the satellite is consistent with the type of payload required by the mission. This means that the execution of any mission requires the satellite's payload to remain visible to the mission target area, and the visible time window must be within the duration of the observation mission; This means that a satellite can only select one imaging opportunity to perform imaging within a time window, that is, the side view remains fixed within the imaging time window; This indicates the time required for a satellite to perform a side-view maneuver if it needs to perform another imaging operation after one imaging operation has ended. This means that the total energy required for observation and maneuvering of the satellite within one orbital period must be less than the maximum energy the satellite possesses. This means that the duration of each imaging session is greater than or equal to the minimum duration required by the mission. in, Indicates the angular velocity of the maneuver from the satellite side perspective. express Each time slot This indicates the energy required for the satellite to image the target area once. Indicates satellite selection strip The motor energy required to achieve the desired yaw angle This indicates the maximum energy that can be consumed within one orbital period.
[0034] Further, please see Figure 5 , Figure 5 This is a flowchart of a training model provided in an embodiment of the present invention. The trained model is obtained by training an initial prediction neural network and a target neural network, guided by the system objective function and constraints. The training process includes: S1. Initialize reinforcement learning hyperparameters and predict neural network. and target neural network Initialize the task scenario, obtain the list of tasks visible in each time slot and the list of tasks visible to satellites, and save them respectively. and And set the current time slot to 0; S2. Obtain all tasks in the current time slot and the corresponding visible satellites for those tasks, and compile a current satellite list. To obtain all tasks within the current time slot and time interval, a task list is generated. Extracting current task elements and satellite status to obtain the agent's current state, represented as: ; in, This indicates the number of tasks that the satellite with the time slot to be allocated can perform. This indicates the total number of satellites that can be allocated time slots. This represents the total number of tasks that all satellites in a time slot can perform. This indicates the number of tasks that the currently assigned satellites can perform. This indicates the base reward multiplier for performing this task. Indicate whether the task requires multiple payloads or a single payload. Indicates whether the task is active. This indicates the current coverage of the task. This indicates the number of conflicting tasks, that is, the impact of executing this task on other tasks. This indicates the number of satellites that a mission can use in a time slot, i.e., the amount of resources available. This indicates the number of satellites already used for the mission.
[0035] It should be noted that the number of tasks that can be assigned to each satellite is not fixed. When the number of tasks is greater than or equal to... At that time, choose the one with the highest return multiplier. If the number of tasks is less than 1; Each state is padded with 0s at the corresponding positions to ensure that the dimension of the state space remains unchanged.
[0036] Write the current number of satellites, the total number of tasks, and the number of tasks that can be executed by the currently unassigned satellites into the state space; then retrieve the task list of the unassigned satellites, sort it from largest to smallest according to the base reward multiplier of the tasks, and select the top-ranked tasks. One. Traverse this Each task calculates and retrieves its current task status, current task coverage, number of conflicting tasks, number of available resources, number of available satellite imaging opportunities, and number of satellites already in use. These are then written sequentially into the state space to obtain the current state.
[0037] S3. Determine the current satellite list. Is it empty? If yes, increment the time slot by 1, update the agent's current state, i.e., execute S2; otherwise, proceed according to... The strategy selection action involves choosing the corresponding satellite and executing the corresponding task at the corresponding imaging angle; the action space is represented as follows: ; in, This indicates the number of available stripes for the target area from the satellite within the current visible time window; It should be noted that, from the side view of certain imaging opportunities, the satellite stripes may not cover the target area, and these stripes will be removed. Therefore, there may be insufficient imaging opportunities between the satellite and the mission area. In this case, a mask is needed to block out these actions, so that the action space remains unchanged while ensuring that the agent does not select these invalid actions.
[0038] S4. Determine if the current action value is 0; if yes, and if none of the agent's current actions can effectively cover the task, then reward the agent to encourage it to choose not to perform the task in the current state; otherwise, punish it; if no, reward the agent based on the observed results; where the reward function is expressed as: ; in, Indicates stripe The number of newly added coverage areas, Indicates the task before imaging The remaining coverage, This represents a constant used to smooth out immediate rewards. This represents the revenue decay value, specifically the energy penalty and the number of satellites already used. Available satellite resources and number of conflict missions The cumulative value after multiplying by the corresponding coefficient. This represents the penalty value for the agent choosing the wrong action.
[0039] It should be noted that the design of the reward function should take into account the timing of task execution and minimize the impact on other tasks. That is, the more new coverage, the less duplicate coverage, the faster the execution time, and the smaller the impact on other tasks, the greater the reward should be.
[0040] In this embodiment, the idea of resource reservation is incorporated into the algorithm design. When matching a satellite for a certain task, the monitoring coverage effect obtained by the satellite at the current imaging angle is first judged. If the imaging effect does not meet the ecological monitoring accuracy requirements, the satellite resource is reserved to avoid inefficient occupation and prevent other tasks from having no resources available at the same time due to excessive concentration of resources on a single task. This effectively solves the suboptimal problem of excessive concentration on sudden tasks, which leads to low overall observation benefits.
[0041] Furthermore, by employing a resource reservation method, the demand for tasks to be executed in the next few minutes is simultaneously predicted when allocating satellites for the current mission. This allows for the pre-allocation of some satellite resources for subsequent short-term tasks, ensuring that tasks not yet allocated resources in the short term still have a chance to acquire observation resources later, preventing tasks from being missed due to sudden resource shortages. In addition, the method proposed in this invention trains a policy network using deep reinforcement learning algorithms and relies on this network for task planning. The aforementioned resource retention and reservation algorithm design can simultaneously improve the response speed to sudden ecological monitoring tasks and the overall observation benefits of the observation missions.
[0042] S5. Store the experience values, obtain the next state of the agent, and if the sampling requirements are met, calculate the predicted Q-value and the target Q-value, calculate the loss value, update the network parameters of the prediction neural network, and simultaneously softly update the network parameters of the target neural network. ; S6. Determine if the current time slot has reached the maximum time slot for one round of training. If not, increment the time slot by 1 and update the current state of the agent, i.e., execute S2. If yes, end the current round of training and check if the current model has converged. If yes, end the training. If not, continue the next round of training with the updated network parameters of the prediction neural network and the target neural network.
[0043] Furthermore, in this embodiment, after obtaining the trained model, the observation task information, time window resources, imaging strip resources, and satellite status are input into the trained model for processing to obtain task scheduling decisions.
[0044] In this embodiment, the scheduling process of sudden ecological monitoring tasks is described as a deep reinforcement learning process of an intelligent agent. The scheduling model incorporates the precise adaptation logic of different types of payload characteristics such as optical, infrared, and SAR with ecological elements. Combined with the target neural network, a scheduling model for sudden ecological observation tasks suitable for low-orbit remote sensing satellite constellation scenarios is established. This shortens the planning time of multi-satellite monitoring tasks, meets the core requirement of rapid response for sudden ecological monitoring, ensures that sudden tasks can complete satellite resource matching and observation plan generation in a short time, and achieves precise matching of "ecological elements-specific payloads" and collaborative observation of multiple payloads, overcoming the shortcomings of single payloads being limited by environmental factors.
[0045] Furthermore, in this embodiment, the task scheduling sequence is obtained based on the task scheduling decision, including: Based on the actions selected by the mission scheduling decision, the corresponding satellite and the corresponding imaging angle are obtained and saved to the result list of the corresponding mission, thus obtaining the scheduling sequence of all missions.
[0046] S106. Generate scheduling instructions from the task scheduling sequence and send observation instructions to the corresponding satellites via uploading.
[0047] S107. Based on the observation instructions, the corresponding satellite performs routine or emergency observation tasks.
[0048] In summary, the rapid, reconfigurable multi-satellite, multi-payload collaborative method and system for ecological monitoring provided by this invention has the following beneficial effects: First, this invention describes the scheduling process of sudden ecological monitoring tasks as a deep reinforcement learning process of an intelligent agent. The scheduling model fully considers the functional characteristics and applicable scenarios of different payloads such as optical, infrared, and SAR, and establishes a precise correspondence mechanism between ecological elements and payload types. Combined with the target neural network, a scheduling model for sudden ecological observation tasks adapted to low-orbit remote sensing satellite constellation scenarios is constructed. When a sudden ecological observation task arrives, it can quickly generate scheduling results, ensuring the accurate allocation of satellite resources and adapted payloads in a short time. It can also achieve multi-payload synergy and complementarity, avoiding the problem of single payloads being limited by environmental factors such as light, and ensuring the accuracy and integrity of monitoring data.
[0049] Secondly, the present invention adopts a resource reservation method. When selecting satellites and imaging angles for observation, the current coverage effect is obtained. By reserving satellites with poor coverage effects, the inefficient use of satellites is avoided. This effectively solves the problem of excessive resource tilting and low overall observation benefits caused by sudden tasks, and improves the overall utilization efficiency of satellite resources.
[0050] Third, compared with the problems of slow response time and limited response resources in existing emergency ecological observation task planning methods, this invention introduces a resource reservation method. By including tasks in the next few minutes in the resource allocation, some resources are reserved in advance for subsequent tasks, avoiding the omission of tasks due to instantaneous resource shortages and ensuring rapid response to emergency observation tasks.
[0051] It should be noted that, in this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations are intended to cover non-exclusive inclusion, such that an article or device comprising a list of elements includes not only those elements but also other elements not expressly listed. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the article or device comprising said element. Terms such as "connected" or "linked" are not limited to physical or mechanical connections but can include electrical connections, whether direct or indirect. The orientations or positional relationships indicated by terms such as "upper," "lower," "left," and "right" are based on the orientations or positional relationships shown in the accompanying drawings and are used only for the convenience of describing the invention and for simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be construed as limiting the invention.
[0052] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the present invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Furthermore, those skilled in the art can combine and integrate the different embodiments or examples described in this specification.
[0053] The above description, in conjunction with specific preferred embodiments, provides a further detailed explanation of the present invention. It should not be construed that the specific implementation of the present invention is limited to these descriptions. For those skilled in the art, various simple deductions or substitutions can be made without departing from the concept of the present invention, and all such modifications and substitutions should be considered within the scope of protection of the present invention.
Claims
1. A rapidly reconfigurable multi-satellite, multi-payload collaborative system for ecological monitoring, characterized in that, include: The monitoring reconstruction request module includes an ecological element extraction submodule and a task information request submodule. The ecological element extraction submodule is used to extract the ecological elements to be observed based on the ecosystem of the target area being monitored. The task information request submodule is used to summarize the time window resources and imaging strip resources required by the task and output the first task element of each task. The detection and reconstruction control module includes a sudden ecological element analysis submodule, a payload matching submodule, and a reconstruction execution submodule. The sudden ecological element analysis submodule is used to analyze the ecological elements to be observed and output the disaster type occurring in the target area. The payload matching submodule is used to match the corresponding imaging payload type according to the disaster type occurring in the target area and output the second task elements of each task. The reconstruction execution submodule is used to generate observation tasks with complete elements according to the first task elements and the second task elements of each task. Based on observation mission parameters, satellite imaging side-swing characteristics, target area gridding coverage rules, satellite imaging energy consumption, and satellite side-view maneuver time constraints, a system objective function is constructed with the optimization goals of maximizing the total coverage of the entire observation area and minimizing the response time of sudden ecological observation missions. Six hard execution constraints are configured: matching the satellite payload type with the mission-required payload type; the duration of the observation mission covering the entire visible time window of the target area; the constraint that a single satellite can only use a fixed side-view to complete a single imaging operation within a single visible time window; the constraint of reserving side-view maneuver time between two satellite imaging operations; and the constraint of satellite single orbital cycle... During the period, the total energy consumption of imaging and side-swing maneuvering shall not exceed the maximum available energy consumption constraint of the satellite, and the duration of a single satellite imaging session shall not be less than the minimum imaging duration specified in the mission. A Markov decision process framework shall be built based on the objective function of the system and six execution constraints. A dual-network reinforcement learning model consisting of a prediction neural network and a target neural network shall be installed. The model shall be trained to convergence through a process of time-slot cyclic iteration, sample experience storage, real-time reward calculation, loss value solution, prediction network parameter update, and target network soft update. The trained reinforcement learning model shall be called to process the complete observation mission to obtain the mission scheduling decision, and the mission scheduling sequence shall be output according to the mission scheduling decision. The task execution module includes a scheduling instruction uploading submodule and a task execution submodule. The scheduling instruction uploading submodule is used to generate scheduling instructions from the scheduling sequence of the task and send observation instructions to the corresponding satellite through uploading. The task execution submodule is used to control the corresponding satellite to perform routine or emergency observation tasks according to the observation instructions.
2. The rapidly reconfigurable multi-satellite, multi-payload collaborative system for ecological monitoring according to claim 1, characterized in that, The reconstruction execution submodule includes a task generation unit, a model loading unit, and a task planning unit. The task generation unit is used to output an observation task with complete task elements based on the first task elements and the second task elements of each task. The model loading unit is used to process the observation task with complete task elements using a trained model and output a task scheduling decision. The task planning unit is used to output a task scheduling sequence based on the task scheduling decision.
3. The rapidly reconfigurable multi-satellite, multi-payload collaborative system for ecological monitoring according to claim 1, characterized in that, The task execution submodule includes an emergency task execution unit and a regular task execution unit. The emergency task execution unit is used to execute emergency observation tasks, and the regular task execution unit is used to execute regular observation tasks. At the same time, the emergency ecological disaster is fed back to the detection and reconstruction control module to initiate a new monitoring reconstruction.
4. A rapid, reconfigurable, multi-satellite, multi-payload collaborative method for ecological monitoring, characterized in that, include: Monitor the ecosystem of the target area and extract the ecological elements to be observed; Based on the ecological elements to be observed, the task scenario is described, and the time window resources and imaging strip resources required for the task are obtained to construct the first task elements for each task. The types of disasters occurring in the target area are obtained by analyzing the ecological elements to be observed. Based on the type of disaster occurring in the target area, the corresponding imaging payload type is matched to obtain the second task element for each task; Based on the first task element and the second task element of each task, an observation task with complete task elements is obtained. Based on observation mission parameters, satellite imaging side-swing characteristics, target area gridding coverage rules, satellite imaging energy consumption, and satellite side-view maneuver time constraints, a system objective function is established with the optimization goals of maximizing the total coverage of the entire observation area and minimizing the response time of sudden ecological observation missions. Six hard execution constraints are also set: matching constraint between satellite payload type and mission-required payload type; constraint on the duration of the observation mission covering the entire visible time window of the target area with the satellite payload; constraint that a single satellite can only use a fixed side-view to complete a single imaging operation within a single visible time window; constraint on the reserved side-view maneuver time between two satellite imaging operations; and constraint on the imaging energy consumption within a single satellite orbital cycle. The total energy consumption of the side-swing maneuver does not exceed the maximum available energy consumption constraint of the satellite, and the duration of a single satellite imaging session is not less than the minimum imaging duration specified in the mission. A Markov decision process framework is constructed based on the system's objective function and six execution constraints. A dual-network reinforcement learning model consisting of a prediction neural network and a target neural network is built. The model is iteratively trained according to a fixed process of time-slot cyclic iteration, sample experience storage, immediate reward calculation, loss value solution, prediction network parameter update, and target network soft update until the model converges. The trained reinforcement learning model is used to process the observation task with complete mission elements to obtain a mission scheduling decision. Based on the mission scheduling decision, the mission scheduling sequence is obtained. The scheduling sequence of the task is used to generate scheduling instructions, and observation instructions are sent to the corresponding satellites via uplink. According to the observation instructions, the corresponding satellite performs routine or emergency observation tasks.
5. The rapid, reconfigurable multi-satellite, multi-payload collaborative method for ecological monitoring according to claim 4, characterized in that, Describe the task scenario, including: Set the satellite's orbital plane, the number of satellites on each plane, the satellite's altitude and orbital inclination, and the target area to describe the mission scenario; among these, The satellite is represented as: ； in, Indicates the total number of satellites. Indicates the satellite index; The target region is represented as: ； in, Indicates the total number of target areas; against ;in, ， in, Represents the coordinates of the boundary points of the target region. Indicates the type of ecological elements in the target area. Indicates the types of disasters that may occur in the target area. This indicates the probability of each disaster occurring. The target region is divided into sub-regions, denoted as... , This indicates the number of sub-regions divided into corresponding target regions. Indicates the first division The coordinates of the boundary points of each sub-region Indicates the index of the target region.
6. The rapid, reconfigurable multi-satellite, multi-payload collaborative method for ecological monitoring according to claim 5, characterized in that, Acquire the time window resources and imaging strip resources required for the task to construct the first task elements for each task, including: Calculate the visible time windows for all satellites and the target area, and the nadir trajectory of the satellites within the visible time windows; wherein, each satellite can have a visible time window calculated for each mission, and within each visible time window, the satellite corresponds to a nadir trajectory; wherein, the visible time window is represented as: ； in, Indicates the total number of visible time windows; against ;in, ； in, Indicates the visible time window The corresponding target area Indicates the visible time window The corresponding satellite number, Indicates the visible time window The corresponding start time, Indicates the visible time window The corresponding start time and satellite nadir coordinates. Indicates the visible time window The corresponding end time, Indicates the visible time window The corresponding end time is the satellite's nadir coordinates. Indicates the visible time window The corresponding satellite's payload, The index representing the visible time window; Obtain the nadir coordinates of the start and end times of the visible time window corresponding to the satellite; According to the preset side view range and side view discrete angle The discrete side view sequence is calculated; Traverse the discrete side view sequence, for the current side view Calculate the vertex coordinates of the corresponding strip, expressed as: ； ； ； ； in, Indicates the satellite's field of view. , These represent the angles on either side of the strip corresponding to the current side view. Indicates the satellite imaging angle. Indicates the observation angle. Indicates the satellite's coverage angle. This indicates the coverage angle corresponding to the left strip of the strip at the current side view. Represents the Earth's radius. Indicates satellite altitude. Indicates satellite Latitude and longitude coordinates at the current moment Indicates the orbital inclination of the satellite. Indicates satellite The longitude of the vertex of the left band of the time strip. Indicates satellite The latitude of the leftmost apex of the band at time t. Traverse all the visible time windows until the vertex coordinates of all stripes corresponding to each visible time window are obtained; wherein, the stripes are represented as: ； in, This represents the total number of stripes corresponding to the visible time window.
7. The rapid, reconfigurable multi-satellite, multi-payload collaborative method for ecological monitoring according to claim 6, characterized in that, The observation task with complete task elements is represented as follows: ； in, This indicates the total number of observation tasks. An index indicating the observation task; against ;in, ； in, Indicates the first Observation tasks The number, Indicates the first Observation tasks The types are divided into routine observation missions and emergency observation missions. Indicates the first Observation tasks The target area to be observed Indicates the first Observation tasks Vertex coordinates of the target region Indicates the first Observation tasks The start time, Indicates the first Observation tasks Duration, Indicates the first Observation tasks Required observation payload type Indicates the first Observation tasks The minimum required imaging time, Indicates the first Observation tasks Required coverage Indicates the first Observation tasks Basic priority, Indicates the first Observation tasks The completion mark.
8. The rapid, reconfigurable multi-satellite, multi-payload collaborative method for ecological monitoring according to claim 7, characterized in that, Construct the system objective function and constraints, including: Introducing Boolean logic variables The time for the satellite to perform its imaging mission is determined and expressed as: ； in, Indicates the first One observation task, Indicates the visible time window The next Each stripe Indicates the index of the stripe; According to Boolean logic variables The coverage of the target region by the stripe to the sub-regions is calculated, and a Boolean variable is introduced. , representing the coverage status of the sub-regions of the target region, is expressed as: ； in, This indicates that after the target region is discretized into sub-regions, the first... Line 1 Sub-region cells of a column; According to the Boolean variable Determine the observation task Target area coverage , is represented as: ； Get the total coverage of all tasks , is represented as: ； Based on the total coverage of all tasks, the system objective function can be described as follows: ； in, Indicates the number of emergency observation tasks. Indicates the index of the observation task. Indicates the number of routine observation tasks. Index indicating routine observation tasks, Indicates an emergency observation mission Response time Indicates the start time of an emergency observation mission; The constraints on the objective function are defined as follows: This indicates that when performing a mission, the type of payload carried by the satellite is consistent with the type of payload required by the mission. This means that the execution of any mission requires the satellite's payload to remain visible to the mission target area, and the visible time window must be within the duration of the observation mission. Indicates the first The allowed start time for each observation task to be executed. Indicates the first The allowed end time for each observation task; This means that a satellite can only select one imaging opportunity to perform imaging within a time window, that is, the side view remains fixed within the imaging time window; This indicates the time required for a satellite to perform a side-view maneuver if it needs to perform another imaging operation after one imaging operation has ended. This means that the total energy required for observation and maneuvering of the satellite within one orbital period must be less than the maximum energy the satellite possesses. This means that the duration of each imaging session is greater than or equal to the minimum duration required by the mission. in, Indicates the angular velocity of the maneuver from the satellite side perspective. Indicates satellite selection strip The corresponding observation side perspective, This indicates the observation viewpoint corresponding to the strip selected in the previous imaging mission. Indicates satellite selection strip The imaging start time, Indicates satellite selection strip The moment the imaging ends, Indicates the first The total number of all visible time windows corresponding to each satellite. express Each time slot This indicates the energy required for the satellite to image the target area once. Indicates satellite selection strip The motor energy required to achieve the desired yaw angle This indicates the maximum energy that can be consumed within one orbital period.
9. The rapid, reconfigurable multi-satellite, multi-payload collaborative method for ecological monitoring according to claim 8, characterized in that, The trained model is obtained by training the initial prediction neural network and the target neural network, guided by the system objective function and constraints. The training process includes: Initialize reinforcement learning hyperparameters and predict neural networks and target neural network Initialize the task scenario, obtain the list of tasks visible in each time slot and the list of tasks visible to satellites, and save them respectively. and And set the current time slot to 0; Retrieve all tasks in the current time slot and the corresponding visible satellites for those tasks, and compile a current satellite list. To obtain all tasks within the current time slot and time interval, a task list is generated. Extracting current task elements and satellite status to obtain the agent's current state, represented as: ； in, This indicates the number of tasks that the satellite with the time slot to be allocated can perform. This indicates the total number of satellites that can be allocated time slots. This represents the total number of tasks that all satellites in a time slot can perform. This indicates the number of tasks that the currently assigned satellites can perform. This indicates the base reward multiplier for performing this task. Indicate whether the task requires multiple payloads or a single payload. Indicates whether the task is active. This indicates the current coverage of the task. This indicates the number of conflicting tasks, that is, the impact of executing this task on other tasks. This indicates the number of satellites that a mission can use in a time slot, i.e., the amount of resources available. Indicates the number of satellites already used in the mission; Determine the current satellite list Is it empty? If yes, increment the time slot by 1 and update the agent's current state; otherwise, proceed according to... The strategy selection action involves choosing the corresponding satellite and executing the corresponding task at the corresponding imaging angle; the action space is represented as follows: ； in, This indicates the number of available stripes for the target area from the satellite within the current visible time window; Determine if the current action is 0; if so, and if none of the agent's current actions can effectively cover the task, then reward the agent to encourage it not to perform the task; otherwise, penalize it. If not, reward the agent based on the observed performance. The reward function is expressed as: ； in, Indicates stripe The number of newly added coverage areas, Indicates the task before imaging The remaining coverage, This represents a constant used to smooth out immediate rewards. Indicates the yield decay value. This represents the penalty value for the agent choosing the wrong action; Storing empirical values, obtaining the agent's next state, and if the sampling requirements are met, calculating the predicted Q-value and the target Q-value, calculating the loss value, updating the network parameters of the prediction neural network, and simultaneously softly updating the network parameters of the target neural network, i.e. ,in, Represents the network parameters of a neural network. This represents the network parameters of the target neural network. This indicates the number of iterations and updates in the network. Indicates the soft update coefficient; Determine if the current time slot has reached the maximum time slot for one round of training. If not, increment the time slot by 1 and update the agent's current state. If yes, end the current round of training and check if the current model has converged. If yes, end the training. If not, continue the next round of training with the updated network parameters of the predictive neural network and the target neural network.
10. The rapid, reconfigurable multi-satellite, multi-payload collaborative method for ecological monitoring according to claim 4, characterized in that, Based on the task scheduling decision, a task scheduling sequence is obtained, including: Based on the actions selected by the mission scheduling decision, the corresponding satellite and the corresponding imaging angle are obtained and saved to the result list of the corresponding mission, thus obtaining the scheduling sequence of all missions.