Multi-satellite planning method and device based on deep learning neural network, and medium

Abstract

The invention discloses a multi-satellite planning method and device based on a deep learning neural network and a medium, and the method comprises the steps: obtaining the historical planning data of multi-satellite planning, and obtaining the resource information of the multi-satellite planning deep learning neural network after preprocessing; determining a constraint function as a model constraint condition of the deep learning neural network according to constraint information needing to be considered in multi-satellite planning; marking a flag bit, and selecting a resource information characteristic value as a deep learning neural network input according to an observed target task; according to the observation task satisfaction degree of multi-satellite planning and the satellite load utilization rate, an objective function is determined, and the deep learning neural network is trained; and performing multi-satellite planning by using the trained deep learning neural network. According to the novel multi-satellite planning method based on the deep learning neural network, the multi-satellite planning task can have better redundancy, robustness and task execution capacity, and the multi-satellite planning task requirement of a user can be met more quickly and more accurately.

Description

Technical Field

[0001] The present invention relates to the field of establishing a deep learning neural network for multi-satellite planning in satellite imaging work, and in particular to a multi-satellite planning method, device and medium based on a deep learning neural network. Background Art

[0002] With the vigorous development of satellite technology, the mature application of multi-satellite technology, and the rise of commercial aerospace companies, the types of imaging satellites have gradually become richer and the scale has gradually grown, which has not only led to a significant increase in users' demand for satellite imaging tasks, but also made the classification of imaging tasks more complex and detailed. Especially in some special cases, users have higher requirements for the accuracy, clarity, timeliness and other aspects of imaging tasks. Therefore, in the process of satellite imaging, the limitation of independent work of a single satellite has been broken away, forming a comprehensive system of coordinated satellites of various types. In addition, due to the increasing requirements for high-resolution imaging and the limited number of satellites within a short transit range, the use of a single satellite can hardly meet the complex task requirements of regional image acquisition, while multi-satellite collaboration has better redundancy, robustness and mission execution capabilities. Therefore, it is necessary to use multiple satellites with the same sensor type and similar spatial resolution for collaborative imaging to ensure faster completion of user mission requirements.

[0003] Compared with traditional networking satellites and single satellite platforms, multi-satellite collaboration has its own characteristics: First, the functions of multi-satellite collaboration units are becoming increasingly complex, the degree of space mission coupling is higher, and the requirements for task coordination and allocation are higher. The satellite needs to have autonomous task planning capabilities. Secondly, multi-satellite multi-task units make it difficult to expose potential conflicts between satellites or tasks. The mission planning satellite needs to process a large amount of satellite interaction data. In the absence of multi-satellite simulation equipment, traditional simulation test methods are difficult to discover deep-seated conflicts and the impact on multi-satellite units. Thirdly, autonomous mission planning, as the core of completing space missions and tapping the potential of multi-satellite collaboration units, needs to have high reliability and security. If the algorithm function of the multi-satellite collaboration unit is not fully tested, it will cause the satellite function test to fail at best, and cause serious damage to the products and equipment on the satellite at worst. Finally, there are not many simulation verification methods for multi-satellite collaboration. Existing satellite simulation tests are mostly conducted on formations or single satellites. There is insufficient simulation capability to support multi-satellite units with heterogeneous functions, the need for on-orbit collaboration, and frequent information interaction. Against the backdrop of growing demand for satellite imaging and limited capabilities of single satellites, multi-satellite collaboration is an important means to meet diversified observation needs. However, due to different priorities of observation needs and different capability constraints of satellites, there is a problem of conflicts in the use of multiple resources when performing tasks.

[0004] Patent document number CN115016910A discloses a multi-satellite collaborative task planning method and device for multi-observation tasks. The application simplifies the multi-objective planning problem for multi-observation tasks into multiple serial single-objective planning problems. The model can be solved quickly and has strong interpretability. However, it does not use neural grids to build a multi-satellite planning model.

[0005] Patent document with document number CN115271313A The present invention relates to a multi-satellite collaborative task planning method and device based on observation demand conflict degree, which improves the satisfaction rate of observation demand and improves the efficiency of satellite resource utilization. However, the application also does not use neural grid to build a multi-satellite planning model. Summary of the invention

[0006] In view of the above problems, the purpose of the present invention is to provide a multi-satellite planning method, device and medium based on deep learning neural network, so as to solve the conflict problem when multiple satellites perform tasks, thereby shortening the multi-satellite planning time and improving mission satisfaction.

[0007] Embodiments of the present invention provide a multi-satellite planning method, device and medium based on deep learning neural network.

[0008] Aspect 1: A multi-star planning method based on deep learning neural network, comprising: S1. Obtain historical planning data of multi-satellite planning, and after preprocessing, obtain resource information of multi-satellite planning deep learning neural network; S2. Determine the constraint function based on the constraint information that needs to be considered in multi-satellite planning as the model constraint condition of the deep learning neural network; S3, define the flag bit, and select the resource information feature value as the input of the deep learning neural network according to the observed target task; S4. Determine the objective function and train the deep learning neural network based on the observation mission satisfaction and satellite payload utilization of multi-satellite planning; S5. Use the trained deep learning neural network for multi-star planning.

[0009] The resource information includes: a satellite payload set, an observed target set, a single-satellite mission set, and a visible arc set of the satellite for the target.

[0010] Optionally: the: The satellite payload set is defined as , in, The total number of satellites is , , For satellite The identifier of For satellite status, For satellite storage, is the payload type of satellite n, For satellite Priority; The observed target set is defined as , in, The total number of targets is , , For the task The identifier of For the goal status, For the goal Type, For the goal Priority; The single-star mission set is defined as , in, The total number of tasks is , , For the task The identifier of For the task Satellite For the task Type, For the task The start time of For the task The end time of For the task The execution status of For the task Priority; The set of visible arcs of the satellite to the target is defined as , The total number of arc segments is , , For arc segment The identifier of For arc segment The corresponding satellite, For arc segment The corresponding target, For arc segment The start time of For arc segment duration of For arc segment The end time of For arc segment The usage status, For arc segment priority.

[0011] Optionally, the constraint information to be considered in the multi-satellite planning in S2 includes: The probability of satellite payload failure in multi-satellite planning is 0; A single satellite payload observes only one target at a time; Each observation task is executed only once and only one arc segment is selected for execution; The time interval between executing different tasks must be greater than or equal to the switching time of the device; The effective execution time of the observation task must be greater than or equal to the sum of the task duration and the load switching time.

[0012] Optionally: Based on the constraint information, a constraint function is constructed, and the formula is expressed as: A single satellite payload observes only one target at a time:

[0013] Each observation task is executed only once and only one arc segment is selected for execution:

[0014] The time interval between executing different tasks must be greater than or equal to the switching time of the device :

[0015] The effective execution time of the observation task must be greater than or equal to the sum of the task duration and the load switching time:

[0016] in, For one A constant representing the task Is it in the satellite payload? of Arc segment execution, is the effective execution end time of the observation task, is the effective execution start time of the observation task, is a constant representing the interval between observation tasks, Indicates the effective execution time of the task. Indicates the duration of the task, Indicates the switching time of the satellite payload.

[0017] Optionally: The satisfaction of multi-satellite planned observation mission is expressed as:

[0018] The satellite payload utilization rate of multi-satellite planning is expressed as follows:

[0019] Indicates whether the task is successfully executed, 1 if successful, 0 if failed. Indicates whether the visible arc of the satellite payload is allocated, 1 if allocated, and 0 if not allocated. Plan satellite payload visible arcs for multiple satellites.

[0020] Optionally: Determine the objective function based on the observation mission satisfaction and satellite payload utilization of the multi-satellite planning, and the formula is expressed as:

[0021] in, For one The constant, is the observation task satisfaction, is the satellite payload utilization rate.

[0022] Optionally, the flag bit in S3 includes: Set the arc constraint flag according to the observed target task , the duration of the satellite's visible arc to the target is greater than or equal to the mission duration is 1, less than is 0; According to the observed target mission, the priority of the satellite, target, arc segment and mission is weighted and summed, and this value is recorded as the priority flag bit. ; Set the conflict flag according to the observed target task .

[0023] A second aspect: An electronic device comprises a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the method provided in the first aspect when executing the program.

[0024] A third aspect: A non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the method provided in the first aspect.

[0025] Beneficial effects of the present invention: The present invention uses the historical planning data of multi-satellite planning to form the resource information of the multi-satellite planning deep learning neural network. According to the constraint information to be considered in multi-satellite planning, it is used as the model constraint condition of the deep learning neural network. According to the observed target task, the resource information characteristic value is selected as the input of the deep learning neural network by defining the flag bit. According to the observation task satisfaction and satellite payload utilization rate of the multi-satellite planning, the deep learning neural network is trained as the objective function. Finally, the optimal weight matrix is ​​obtained. and the offset vector A multi-satellite planning network model is constructed, and then the trained deep learning neural network is used for multi-satellite planning. This provides a new multi-satellite planning method based on deep learning neural network, which can make the multi-satellite planning task have better redundancy, robustness and task execution capability, and complete the user's multi-satellite planning task requirements faster and more accurately. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1 A schematic diagram of a multi-star planning method based on a deep learning neural network according to the present invention; Figure 2 A schematic diagram of the process principle of a multi-star planning method based on a deep learning neural network of the present invention; Figure 3 It is a schematic diagram of the structure of the electronic device of the present invention. DETAILED DESCRIPTION

[0027] Embodiments of the present invention are described in detail below, examples of which are shown in the accompanying drawings, wherein the same or similar symbols throughout represent the same or similar elements or elements with the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and cannot be understood as limiting the present invention.

[0028] The functions of multi-satellite collaborative units are becoming increasingly complex, the degree of space mission coupling is higher, and the requirements for task coordination and allocation are higher; in the absence of multi-satellite simulation equipment, traditional simulation test methods are difficult to detect deep-seated conflicts and their impact on multi-satellite units; the existing simulation verification methods for multi-satellite collaboration are not rich, and existing satellite simulation tests are mostly conducted on formations or single satellites. The simulation capabilities to support multi-satellite units with heterogeneous functions, the need for on-orbit coordination, and frequent information interaction are insufficient.

[0029] In view of the above problems, the present invention provides a multi-star planning method based on deep learning neural network. Figure 1 A schematic diagram of a multi-star planning method based on a deep learning neural network provided in an embodiment of the present invention, the method comprising: S1. Obtain historical planning data of multi-satellite planning, and after preprocessing, obtain resource information of the deep learning neural network of multi-satellite planning.

[0030] The main principle of deep learning neural network is to self-learn the features of a large number of samples to achieve complex function approximation between input and output. Specifically, deep learning neural network learns more useful features by building multi-layer models and massive training data, thereby ultimately improving the accuracy of classification or prediction, avoiding the tedious process of manual feature extraction, and greatly improving the efficiency and accuracy of data processing and analysis. Deep learning neural network has a wide range of applications, and the structure of deep learning neural network is not introduced in detail here.

[0031] After obtaining the historical data and results of multi-satellite planning, the failed tasks are deleted, and a set of data with a 100% success rate of multi-satellite planning is constructed as resource information. The collected resource information mainly includes: satellite payload set, observed target set, single-satellite mission set and satellite visible arc set for targets.

[0032] The satellite payload set is defined as ,

[0033] in, The total number of satellites is , , For satellite The identifier of For satellite status, For satellite storage, is the payload type of satellite n, For satellite Priority; The observed target set is defined as , in, The total number of targets is , , For the task The identifier of For the goal status, For the goal Type, For the goal Priority; The single-star mission set is defined as , in, The total number of tasks is , , For the task The identifier of For the task Satellite For the task Type, For the task The start time of For the task The end time of For the task The execution status of For the task Priority; The set of visible arcs of the satellite to the target is defined as , The total number of arc segments is , , For arc segment The identifier of For arc segment The corresponding satellite, For arc segment The corresponding target, For arc segment The start time of For arc segment duration of For arc segment The end time of For arc segment The usage status, For arc segment priority.

[0034] S2. Based on the constraint information that needs to be considered in multi-satellite planning, determine the constraint function as the model constraint condition of the deep learning neural network.

[0035] Building a deep learning neural network model needs to meet the objective constraints of multi-satellite planning. The constraints that need to be considered in multi-satellite planning mainly include: The probability of satellite payload failure in multi-satellite planning is 0; a single satellite payload only observes one target at the same time; each observation mission is only executed once and only one arc segment is selected for execution; the time interval for executing different tasks must be greater than or equal to the switching time of the equipment; the effective execution time of the observation mission must be greater than or equal to the sum of the mission duration and the payload switching time, etc.

[0036] The specific constraint information can be expressed by the constraint function as follows: A single satellite payload observes only one target at a time:

[0037] Each observation task is executed only once and only one arc segment is selected for execution:

[0038] The time interval between executing different tasks must be greater than or equal to the switching time of the device :

[0039] The effective execution time of the observation task must be greater than or equal to the sum of the task duration and the load switching time:

[0040] in, For one A constant representing the task Is it in the satellite payload? of Arc segment execution, is the effective execution end time of the observation task, is the effective execution start time of the observation task, is a constant representing the interval of observation tasks, Indicates the effective execution time of the task. Indicates the duration of the task, Indicates the switching time of the satellite payload.

[0041] By constraining the deep learning neural network with the above constraint information, the deep learning neural network has the conditions for multi-satellite planning and coordination, and has better redundancy, robustness and mission execution capabilities.

[0042] S3. Define the flag bit and select the resource information feature value as the input of the deep learning neural network based on the observed target task.

[0043] After the resource information and constraint information involved in the multi-satellite planning task are defined, the resource information that has a significant impact on the multi-satellite planning results is further selected as the feature value input into the deep learning neural network to train the deep learning neural network; specifically, the resource information feature value can be obtained by setting the flag bit, and the flag bit includes: Set the arc constraint flag The quality of the satellite's visible arc to the target has a great impact on the planning results. Depending on the mission of the observed target, the satellite's visible arc duration to the target is greater than or equal to the mission duration. is 1, less than When it is 0, the resource information can be filtered and limited according to the observed target task, which is more conducive to the training of deep learning neural networks.

[0044] Similarly, the priority of the observation task also affects the quality of the multi-satellite planning task results. According to the observed target task, the priority of the satellite, arc segment and task can be weighted and summed, and this value is recorded as the priority flag. , according to the flag bit Differentiate the priorities of observation tasks.

[0045] Conflicts may occur during multi-satellite planning. Conflict flags can be set based on the mission of the target being observed. , 1 in case of conflict, 0 otherwise.

[0046] In addition, the utilization rate of satellite payloads can be considered during the planning process to prevent satellite payload resource work from stagnation and set the load balancing flag. , when the load utilization rate is greater than or equal to 80% is 1, otherwise is 0.

[0047] S4. Determine the objective function and train the deep learning neural network based on the observation mission satisfaction and satellite payload utilization of multi-satellite planning; Among them, the satisfaction degree of multi-satellite planning observation mission is expressed as follows:

[0048] The satellite payload utilization rate of multi-satellite planning is expressed as follows:

[0049] in, Indicates whether the task is successfully executed, 1 if successful, 0 if failed. Indicates whether the visible arc of the satellite payload is allocated, 1 if allocated, and 0 if not allocated. Plan satellite payload visible arcs for multiple satellites.

[0050] Based on the observation mission satisfaction and satellite payload utilization of multi-satellite planning, the objective function formula is expressed as:

[0052] in, For one The constant, is the observation task satisfaction, is the satellite payload utilization rate.

[0053] like Figure 2 As shown, the deep learning neural network is trained according to the objective function until the objective function is lower than the set threshold, and the deep learning neural network training can be considered completed.

[0054] S5. Use the trained deep learning neural network for multi-star planning.

[0055] like Figure 2As shown, after the deep learning neural network is trained, it can be verified and tested. The trained deep learning neural network is tested using the test set formed by the resource information. After the test, if the deep learning neural network does not meet the use requirements, it is retrained. After the deep learning neural network meets the use requirements, it can be used for multi-star planning mission execution.

[0056] After acquiring the multi-satellite planning historical data, the present invention deletes the task data of failed planning, constructs a set of resource information data with a planning success rate of 100%, and then defines the flag bit according to the observed target task, selects the resource information feature value as the input of the deep learning neural network according to the flag bit, and then determines the objective function according to the observation task constraint conditions of the multi-satellite planning to meet the observation task satisfaction and satellite payload utilization of the multi-satellite planning, trains the deep learning neural network, and finally obtains the optimal weight matrix and the offset vector Deep learning neural network model for multi-satellite planning.

[0057] The present invention also provides an electronic device, Figure 3 A schematic diagram of the structure of an electronic device provided by an embodiment of the present invention, such as Figure 3 As shown, the electronic device may include: a processor, a communications interface, a memory, and a communications bus, wherein the processor, the communications interface, and the memory communicate with each other via the communications bus. The processor may call the logic instructions in the memory to execute the following method: S1. Obtain historical planning data of multi-satellite planning, and after preprocessing, obtain resource information of multi-satellite planning deep learning neural network; S2. Determine the constraint function based on the constraint information that needs to be considered in multi-satellite planning as the model constraint condition of the deep learning neural network; S3, define the flag bit, and select the resource information feature value as the input of the deep learning neural network according to the observed target task; S4. Determine the objective function and train the deep learning neural network based on the observation mission satisfaction and satellite payload utilization of multi-satellite planning; S5. Use the trained deep learning neural network for multi-star planning.

[0058] In addition, the logic instructions in the above-mentioned memory can be implemented in the form of software functional units and can be stored in a computer-readable storage medium when sold or used as an independent product. Based on this understanding, the technical solution of the present invention is essentially or the part that contributes to the prior art or the part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions to enable a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk and other media that can store program codes.

[0059] An embodiment of the present invention further provides a non-transitory computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the method provided in each of the above embodiments is implemented, for example, including: S1. Obtain historical planning data of multi-satellite planning, and after preprocessing, obtain resource information of multi-satellite planning deep learning neural network; S2. Determine the constraint function based on the constraint information that needs to be considered in multi-satellite planning as the model constraint condition of the deep learning neural network; S3, define the flag bit, and select the resource information feature value as the input of the deep learning neural network according to the observed target task; S4. Determine the objective function and train the deep learning neural network based on the observation mission satisfaction and satellite payload utilization of multi-satellite planning; S5. Use the trained deep learning neural network for multi-star planning.

[0060] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.

[0061] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0062] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A multi-star planning method based on deep learning neural network, characterized in that: include: S1. Obtain historical planning data of multi-satellite planning, and after preprocessing, obtain resource information of multi-satellite planning deep learning neural network; S2. Determine the constraint function based on the constraint information that needs to be considered in multi-satellite planning as the model constraint condition of the deep learning neural network; S3, define the flag bit, and select the resource information feature value as the input of the deep learning neural network according to the observed target task; S4. Determine the objective function and train the deep learning neural network based on the observation mission satisfaction and satellite payload utilization of multi-satellite planning; S5. Use the trained deep learning neural network for multi-star planning; The resource information includes: a satellite payload set, an observed target set, a single-satellite mission set, and a visible arc set of the satellite for the target.

2. A multi-star planning method based on deep learning neural network according to claim 1, characterized in that: Said: The satellite payload set is defined as , in, The total number of satellites is , , For satellite The identifier of For satellite status, For satellite storage, is the payload type of satellite n, For satellite Priority; The observed target set is defined as , in, The total number of targets is , , For the task The identifier of For the goal status, For the goal Type, For the goal Priority; The single-star mission set is defined as , in, The total number of tasks is , , For the task The identifier of For the task Satellite For the task Type, For the task The start time of For the task The end time of For the task The execution status of For the task Priority; The set of visible arcs of the satellite to the target is defined as , The total number of arc segments is , , For arc segment The identifier of For arc segment The corresponding satellite, For arc segment The corresponding target, For arc segment The start time of For arc segment duration of For arc segment The end time of For arc segment The usage status, For arc segment priority.

3. The multi-star planning method based on deep learning neural network according to claim 1, characterized in that: The constraints that need to be considered in the multi-satellite planning in S2 include: The probability of satellite payload failure in multi-satellite planning is 0; A single satellite payload observes only one target at a time; Each observation task is executed only once and only one arc segment is selected for execution; The time interval between executing different tasks must be greater than or equal to the switching time of the device; The effective execution time of the observation task must be greater than or equal to the sum of the task duration and the load switching time.

4. The multi-star planning method based on deep learning neural network according to claim 2, characterized in that: Based on the constraint information, the constraint function is constructed, and the formula is expressed as: A single satellite payload observes only one target at a time: ； Each observation task is executed only once and only one arc segment is selected for execution: ； The time interval between executing different tasks must be greater than or equal to the switching time of the device : ； The effective execution time of the observation task must be greater than or equal to the sum of the task duration and the load switching time: ； in, For one A constant representing the task Is it in the satellite payload? of Arc segment execution, is the effective execution end time of the observation task, is the effective execution start time of the observation task, is a constant representing the interval of observation tasks, Indicates the effective execution time of the task. Indicates the duration of the task, Indicates the switching time of the satellite payload.

5. The multi-star planning method based on deep learning neural network according to claim 4 is characterized in that: The satisfaction degree of multi-satellite planned observation mission is expressed as: ； The satellite payload utilization rate of multi-satellite planning is expressed as follows: ； Indicates whether the task is successfully executed, 1 if successful, 0 if failed. Indicates whether the visible arc of the satellite payload is allocated, 1 if allocated, and 0 if not allocated. Plan satellite payload visible arcs for multiple satellites.

6. The multi-star planning method based on deep learning neural network according to claim 5, characterized in that: The objective function is expressed as: ； in, For one The constant, is the observation task satisfaction, is the satellite payload utilization rate.

7. The multi-star planning method based on deep learning neural network according to claim 1, characterized in that: The flag bits in S3 include: Set the arc constraint flag according to the observed target task , the duration of the satellite's visible arc to the target is greater than or equal to the mission duration is 1, less than is 0; According to the observed target mission, the priority of the satellite, target, arc segment and mission is weighted and summed, and this value is recorded as the priority flag bit. ; Set the conflict flag according to the observed target task .

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the program, the steps of a multi-star planning method based on a deep learning neural network are implemented as described in any one of claims 1 to 7.

9. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of a multi-star planning method based on a deep learning neural network as described in any one of claims 1 to 7 are implemented.

Images