CN108021433B - Target observation method for multi-satellite cluster - Google Patents

Target observation method for multi-satellite cluster Download PDF

Info

Publication number
CN108021433B
CN108021433B CN201711252589.4A CN201711252589A CN108021433B CN 108021433 B CN108021433 B CN 108021433B CN 201711252589 A CN201711252589 A CN 201711252589A CN 108021433 B CN108021433 B CN 108021433B
Authority
CN
China
Prior art keywords
satellite
target
observation
emergency
algorithm
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active
Application number
CN201711252589.4A
Other languages
Chinese (zh)
Other versions
CN108021433A (en
Inventor
贺仁杰
褚骁庚
张忠山
陈英武
陈宇宁
吕济民
陈盈果
陈成
王涛
刘晓路
邢立宁
姚锋
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
National University of Defense Technology
Original Assignee
National University of Defense Technology
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by National University of Defense Technology filed Critical National University of Defense Technology
Priority to CN201711252589.4A priority Critical patent/CN108021433B/en
Publication of CN108021433A publication Critical patent/CN108021433A/en
Application granted granted Critical
Publication of CN108021433B publication Critical patent/CN108021433B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F9/00—Arrangements for program control, e.g. control units
    • G06F9/06—Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
    • G06F9/46—Multiprogramming arrangements
    • G06F9/48—Program initiating; Program switching, e.g. by interrupt
    • G06F9/4806—Task transfer initiation or dispatching
    • G06F9/4843—Task transfer initiation or dispatching by program, e.g. task dispatcher, supervisor, operating system
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F9/00—Arrangements for program control, e.g. control units
    • G06F9/06—Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
    • G06F9/46—Multiprogramming arrangements
    • G06F9/50—Allocation of resources, e.g. of the central processing unit [CPU]
    • G06F9/5005—Allocation of resources, e.g. of the central processing unit [CPU] to service a request
    • G06F9/5027—Allocation of resources, e.g. of the central processing unit [CPU] to service a request the resource being a machine, e.g. CPUs, Servers, Terminals
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F9/00—Arrangements for program control, e.g. control units
    • G06F9/06—Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
    • G06F9/46—Multiprogramming arrangements
    • G06F9/50—Allocation of resources, e.g. of the central processing unit [CPU]
    • G06F9/5005—Allocation of resources, e.g. of the central processing unit [CPU] to service a request
    • G06F9/5027—Allocation of resources, e.g. of the central processing unit [CPU] to service a request the resource being a machine, e.g. CPUs, Servers, Terminals
    • G06F9/5038—Allocation of resources, e.g. of the central processing unit [CPU] to service a request the resource being a machine, e.g. CPUs, Servers, Terminals considering the execution order of a plurality of tasks, e.g. taking priority or time dependency constraints into consideration

Landscapes

  • Engineering & Computer Science (AREA)
  • Software Systems (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Radio Relay Systems (AREA)

Abstract

The invention discloses a target observation method for a multi-satellite cluster, which comprises the following steps: step1, generating a pre-planned observation scheme for the conventional target by each satellite, and executing the current observation scheme; step2, when a new emergency target arrives, the main satellite screens load information capable of executing the task and sends the load information to the execution satellite; step3, after receiving the information, each satellite locally calculates the visibility relation with the target, and generates and sends an observation report to the main satellite; step4, determining the satellite to execute the observation task according to each observation report by the main satellite, and distributing the task result to each satellite which determines to execute the new emergency target; and Step5, adding the new cooperative task into the subsequent task to be planned by the observation task execution satellite, updating the own observation scheme, and executing the new observation planning scheme. The invention can not only reduce the labor cost, but also greatly improve the use efficiency of satellite resources so as to better and more reliably meet the observation requirements of different application scenes.

Description

Target observation method for multi-satellite cluster
Technical Field
The invention relates to the technical field of satellites, in particular to a target observation method for a multi-satellite cluster.
Background
As shown in fig. 1, a satellite cluster is composed of a main satellite and a plurality of observation performing satellites (the main satellite may also perform the task of target observation). In the application scene of the agile satellite, each satellite in the satellite constellation has a respective daily observation target set and a corresponding observation task. Meanwhile, in order to enhance the response capability of the satellite to the emergency target and quickly identify the target to determine the target state, the design requirement of the satellite on-satellite autonomous planning capability is provided. However, most of the current autonomous satellites are in a single-satellite autonomous mode, that is, emergency targets can only be handed to a specific satellite for execution. Due to the constraint conditions such as the satellite operation orbit and the observation resources (corresponding to the time window constraint, the fixed storage and other resource constraints), a single satellite often cannot respond to the observation requirement of the target quickly in time. With the development of satellite technology and the increase of the number of satellites, the traditional control mechanism of each satellite chimney type is no longer suitable for future dynamic application scenarios.
Disclosure of Invention
It is an object of the present invention to provide a method of target observation of a cluster of multiple stars to overcome or at least alleviate at least one of the above mentioned drawbacks of the prior art.
In order to achieve the above object, the present invention provides a target observation method for multiple satellite clusters, wherein the control structure of the satellite clusters is a centralized-distributed structure, which includes a main satellite and multiple execution satellites; the observation tasks of the satellite constellation comprise a conventional target observation task and an emergency target cooperative observation task, and the target observation method of the multi-satellite constellation comprises the following steps:
step1, generating a pre-planned observation scheme for the conventional target by each satellite by using a rolling planning algorithm, and executing the current planned observation scheme;
step2, when a new emergency target arrives, the main satellite firstly screens load information capable of executing the task, then sends the emergency target information to a corresponding execution satellite through an inter-satellite link in a broadcasting mode, and sets the receiving time of an observation and analysis report;
step3, the execution satellite receiving the emergency target information transmitted by the main satellite estimates the own observation cost of the emergency task by using local computing resources, generates an observation report and sends the observation report back to the main satellite;
step4, determining the execution satellite for executing the emergency target observation task by the main satellite according to the observation report of each execution satellite and by using a task cooperative allocation algorithm, and assigning the emergency target observation task to the corresponding execution satellite; the conditions met by the execution satellite for executing the emergency target observation task comprise: 1. the requirement of an emergency target on the observation load is met; 2. the requirement of the observation deadline of the emergency target is met; 3. the existing emergency target observation condition of the satellite is not influenced;
step5, adding a new emergency target observation task into a task to be planned later by the execution satellite receiving the emergency target observation task assignment, updating the own observation scheme by using an online scheduling method, executing the new observation planning scheme, and continuously executing the original observation planning scheme by other satellites;
wherein the main constraints of each executing satellite are a time window constraint, a time-dependent maneuvering time constraint and a fixed memory constraint; the emergency target information includes an emergency target ID, location information, observed load demand, observed revenue, and latest imageable time.
The invention can improve the resource utilization rate of all satellites, and can better ensure the response time and quality of emergency targets by means of inter-satellite networking and autonomous collaborative planning. The satellite cooperative networking can improve the overall use efficiency of the executed satellite, and compared with the emergency response of a single satellite, the emergency cooperative response capability of the satellite cluster can better ensure that important tasks are executed smoothly and the imaging information of important targets is acquired in time. Through interconnection and intercommunication among satellite resources and autonomous coordination capacity, the satellite network autonomously arranges proper observation resources and time windows for different emergency targets according to the state and target requirements of each satellite. The star cluster structure can not only reduce the labor cost but also greatly improve the service efficiency of satellite resources by means of the autonomous task planning capability, so that the observation requirements of different application scenes can be met more efficiently and more reliably.
Drawings
Fig. 1 is a schematic diagram of the structure of the satellite constellation of the present invention.
FIG. 2 is a flow chart diagram of a target observation method of a multi-satellite cluster, wherein a single arrow dotted line in the diagram represents the upper-note emergency target information; the solid single-arrow line represents a determination of a target-executing satellite that satisfies both the observed load and time requirements and does not affect the assigned emergency targets; the double headed arrows represent screening of carrier and broadcast tasks from the master star and return of observation reports from the star.
FIG. 3 is a schematic view of an observed slope.
Fig. 4 is a schematic diagram of time window overlap.
FIG. 5 is a graphical illustration of the impact of the adaptive filtering mechanism on the algorithms corresponding to each heuristic strategy.
FIG. 6 is a graph illustrating the increase in observed yield of the algorithm corresponding to each heuristic strategy by the adaptive filtering mechanism.
Fig. 7 is a schematic diagram of ranking distribution of revenue ranking statistics of algorithm 1 and algorithm 2 in a heuristic strategy in different scenarios.
Fig. 8 is a schematic diagram of ranking distribution of revenue ranking statistics of algorithm 3 and algorithm 4 in a heuristic strategy in different scenarios.
Fig. 9 is a schematic diagram of ranking distribution of revenue ranking statistics of algorithm 5 and algorithm 6 in a heuristic strategy in different scenarios.
Fig. 10 is a schematic diagram of ranking distribution of revenue ranking statistics of algorithm 7 and algorithm 8 in a heuristic strategy in different scenarios.
Fig. 11 is a schematic diagram of ranking distribution of revenue ranking statistics of algorithms 9 and 10 in a heuristic strategy in different scenarios.
FIG. 12 is a representation of other algorithms when Algorithm 7 in the heuristic strategy ranks 6 bits later (inclusive).
FIG. 13 is a graphical illustration of the benefit comparison of an algorithm selector to other algorithms.
FIG. 14 is a schematic diagram of a fault statistical analysis of the algorithm selector.
Detailed Description
In the drawings, the same or similar reference numerals are used to denote the same or similar elements or elements having the same or similar functions. Embodiments of the present invention will be described in detail below with reference to the accompanying drawings.
In the description of the present invention, the terms "central", "longitudinal", "lateral", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", etc., indicate orientations or positional relationships based on those shown in the drawings, and are only for convenience in describing the present invention and simplifying the description, but do not indicate or imply that the device or element being referred to must have a particular orientation, be constructed in a particular orientation, and be operated, and therefore, should not be construed as limiting the scope of the present invention.
Based on the application requirements of the current satellite, each executing satellite has a respective exclusive conventional observation task and also has a quick response capability to the emergency target observation task. Therefore, a collaborative satellite network which considers both the conventional observation task and the emergency observation task needs to be designed, and the satellite network has high feasibility in the current satellite hardware technical condition.
The structure of the common satellite constellation mainly comprises three structures of complete centralization, complete distribution and distribution centralization. The characteristics of the three structures are briefly analyzed, and a suitable structure is selected as an experimental structure of the satellite constellation according to the current technical conditions.
The completely centralized structure means that all computations (mainly including time window computation of each satellite, emergency target cooperative allocation and generation of each single-satellite observation scheduling scheme) of a satellite cluster are completed in a main satellite. The structure has the advantages that the communication load of the inter-satellite link can be effectively reduced, the design of other executed satellites except the main satellite is simpler, the manufacturing cost is lower, and the global optimization of a satellite cluster scheme is facilitated (the main satellite can directly carry out multiple iterative optimization solution on a main satellite computer under the condition that the calculation capacity of the main satellite allows). However, the disadvantage of the star cluster structure is also obvious, and mainly, the calculation load of the main star is too high, the requirement on the calculation capacity of the main star is too high, the cost of the single star of the main star is too high, and the robustness of the star cluster is poor. Meanwhile, as the calculation load of the main satellite is continuously increased along with the increase of the number of satellites, the size of the satellite cluster is limited by the calculation capacity of the main satellite, and an execution satellite cluster with considerable scale cannot be built in a short time according to the structure according to the relevant capacity of the current satellite-borne computer.
The completely distributed structure means that the positions of all satellites in a satellite cluster are equal, the role of a main satellite is not existed, and the calculation work of visual time windows of all satellites, task planning of a single satellite and the like is completed locally on all satellites. In the aspect of cooperative allocation of emergency targets, the satellite cluster enables the satellites to obtain consistent consensus through a mode of multiple inter-satellite information interaction between each satellite and adjacent satellites, and determines an execution satellite of a certain emergency task. The structure has the advantages that the calculation load of the star cluster is uniform, the robustness of the star cluster is high, and the expansion capability is strong. The structure has the disadvantages that each satellite can obtain a consistent decision result through multiple negotiations, the inter-satellite communication cost is high, and the inter-satellite link load is large. Moreover, if the satellite cannot make a local decision, the decision speed is affected to a certain extent, and if the satellite makes a local fast decision, the decision quality is not high. Meanwhile, each satellite has certain decision-making capability, which also increases the cost of each satellite.
The distributed centralized structure means that each satellite can locally calculate a time window and an observation scheduling scheme, and the main satellite determines an execution satellite of an emergency target according to the evaluation of the emergency target corresponding to each satellite. The structure has the advantages that the calculation load of each satellite is balanced, the communication cost of the inter-satellite link is relatively controllable, the decision efficiency and the decision quality of an emergency target are considered, and the robustness of the satellite cluster is completely centralized (a main satellite is easily replaced). Moreover, the main satellite only needs to add one cooperative decision function relative to the executive satellite, the expansibility of the satellite cluster is good, and the new satellite can be added into the satellite cluster as long as the new satellite is registered at the main satellite and an inter-satellite link is established with other satellites. The structure has the defects that certain requirements are required for the computing capacity and the inter-satellite communication capacity of each satellite, maintenance and corresponding backup are required to be carried out on a main satellite at fixed time in order to guarantee the robustness of a satellite cluster, and partial satellites are required to participate in computing in each decision.
According to the related technology of the current satellite platform and the inter-satellite link, in the process of designing and analyzing the cooperative allocation algorithm, the satellite cluster is assumed to be a distributed centralized networking structure. The structure has strong feasibility and is convenient to separate the cooperative decision function of the target to be exhausted from the on-line scheduling function of a single satellite. Therefore, the structure is also helpful for the invention to carry out deeper research on the emergency target cooperative allocation algorithm of the autonomous satellite constellation by virtue of the research content.
In a scene of satellite cluster cooperative task planning, a satellite cluster is composed of a plurality of autonomous satellites, one satellite in the satellite cluster is a main satellite, and other satellites are executive satellites. The observation task of the satellite cluster mainly comprises a conventional target observation task and an emergency target cooperative observation task of each satellite. The operation target of the star cluster is to maximize the operation efficiency of the system in one period, namely to maximize the observation yield of the star cluster to the target. The main satellite is responsible for the cooperative observation planning of the emergency targets, the visible time windows of the executed satellites and the targets and the respective observation scheduling schemes are calculated by the executed satellite loads, and the executed satellites and the main satellite can carry out real-time information communication.
The conventional targets refer to respective observation task targets of the satellites, and all task information is known by the satellites before the cooperative task planning scene begins. The task information mainly comprises: task ID, task position, task observation time window, task observation duration and task observation income.
Relevant information for emergency objectives is unknown prior to the start of the collaborative mission planning scenario. As shown in fig. 2, during the operation of the constellation, after a user initiates an emergency target observation requirement, the annotating device of the ground station transmits information of the emergency target to the main satellite, and the main satellite allocates tasks to appropriate execution satellites according to task requirements (observation load, observation deadline, and the like) and states of the satellites.
As can be seen from the above discussion, the management and control structure of the satellite constellation is a centralized-distributed structure, where the centralized structure is mainly realized by the main satellite taking charge of cooperative scheduling to complete the allocation work of the emergency targets, and the distributed structure is realized by the respective autonomous satellites providing observation and evaluation reports of the emergency targets through local computation and adjusting the relevant scheduling observation schemes.
As shown in fig. 2, the target observation method for a multi-star cluster provided in this embodiment includes:
step1, generating a pre-planned observation scheme for the conventional target by each satellite by using a rolling planning algorithm, and executing the current planned observation scheme;
step2, when a new emergency target arrives, the main satellite firstly screens load information capable of executing the task, then sends the target information to a corresponding execution satellite through an inter-satellite link in a broadcasting mode, and sets the receiving time of an observation and analysis report;
step3, after receiving the information, each satellite locally calculates the visibility relation with the target, and sends the relevant information back to the main satellite after generating an observation report;
step4, determining target execution satellites for executing the observation tasks by the main satellite according to the observation reports of the satellites and by using a task cooperative allocation algorithm, and broadcasting task allocation results to the target execution satellites; the conditions met by the target executive satellite include: (1) observing load and time requirements, (2) not affecting assigned emergency objectives;
step5, adding a new cooperative task into a task to be planned later by the target execution satellite determined by Step4, updating an own observation scheme by using an online scheduling method, executing a new observation planning scheme, and continuously executing the original observation planning scheme by other satellites;
the main constraints of each satellite are among others time window constraints, time-dependent maneuvering time constraints and fixed memory constraints.
For example: as shown in FIG. 1, the satellite constellation consists of three satellites Sat1, Sat2 and Sat3, wherein Sat1 is the main satellite and Sat2 and Sat3 are executive satellites. As shown in fig. 1, the dotted line below the satellite shown in the off-satellite direction of each satellite is shown, and the observation path is shown by the solid line below the satellite. During the operation of the satellite constellation, the user finds the emergency target 9 and injects the specific information of the emergency target to the master satellite Sat1 through the satellite-to-ground communication network. The main satellite Sat1 allocates the emergency target 9 to the executive satellite Sat2 according to the target information and the current state of each satellite in the cluster; after executing Sat2 and receiving target information, adding a target 9 into a sequence to be observed, generating a new observation scheme by using an on-satellite autonomous task planning algorithm, and adjusting an original observation path {3 → 6 → 8 → 4} into {3 → 6 → 9 → 4 }; the observation paths of the other satellites (Sat1, Sat3) in the constellation remain unchanged.
When a scene starts, each satellite plans a conventional task observation scheme according to the conventional target task information of the satellite, and executes the current planning observation scheme when an emergency cooperative task does not exist. When a certain emergency target is noted on the main satellite on the ground, the main satellite distributes the emergency target to each satellite through an inter-satellite link, and each satellite estimates the observation cost of the main satellite for the emergency task by using local computing resources to generate a corresponding observation report. After receiving the observation reports of each satellite, the main satellite triggers a task cooperative allocation algorithm to select a proper satellite to execute the observation task, and relevant information of emergency target observation is reported to the ground. And after the execution satellite receives the emergency target observation task assignment, adjusting the subsequent observation scheme of the execution satellite. The main constraints of each satellite are among others time window constraints, time-dependent maneuvering time constraints and fixed memory constraints.
Each satellite generates a pre-planned observation scheme for a conventional target by using a rolling planning algorithm at the beginning of a scene; each satellite executes the current planning observation scheme; when a new emergency target arrives, the main satellite sends target information to each satellite; after receiving the information, each satellite locally calculates the visibility relation with the target, and sends the relevant information back to the main satellite after generating an observation report; the main satellite determines the satellite to execute the task by using an allocation algorithm according to the observation report of each satellite, and broadcasts the task allocation result to each satellite; the observation task execution satellite adds a new cooperative task into a subsequent task set to be planned, updates an own observation scheme by using an online scheduling method, and executes a new observation planning scheme; other satellites continue to execute the original observation planning scheme.
The cooperative flow of the star clusters is shown in fig. 2. After the observation requirement of the emergency target exists, the user injects the observation requirement of the task to the main satellite through the satellite-ground link (as shown by a short dashed line 1 in the figure), and the main information of the emergency target comprises an emergency target ID, position information, an observation load requirement, observation income and latest imaging time. After receiving the emergency target information, the main satellite firstly screens load information capable of executing the task, then sends the target information to a corresponding execution satellite in a broadcasting mode through an inter-satellite link, and sets the receiving time of an observation analysis report; after receiving the tasks, each satellite calculates the visibility relationship with the target by using local computing resources, generates an observation analysis report, and returns the report to the main satellite (as shown by a bidirectional solid line 2 in the figure). After the main satellite receives the observation analysis report of each satellite (if the observation analysis report of a certain satellite arrives at the main satellite after the reception time, the satellite is considered to be unable to observe the target), under the premise that the observation requirement of the target can be met and the arranged emergency target is not affected, the satellite for executing observation is determined by using a task cooperative allocation algorithm, and the corresponding satellite is assigned to adjust the task planning scheme thereof, so as to observe the emergency target (as shown by a one-way solid line 3 in the figure, the executed satellite is Sat 2).
In the cooperative flow, after receiving information of an emergency target, a main satellite (Sat1) interacts with each satellite through an inter-satellite link to obtain an observation analysis report of the target once, and after the report is collected, a corresponding execution satellite is selected by using a task cooperative allocation algorithm and assigned with a task. Namely, 3 times of unidirectional information transmission (two times of unidirectional transmission are needed for observation and analysis report interaction) is needed between the satellites for each emergency target, so that a proper target observation execution satellite can be determined. Because the data volume of the observation analysis report and the target assignment information is small, the load of the cooperative flow on the inter-satellite transmission is small, and the realization in engineering is facilitated. Meanwhile, the main satellite summarizes the analysis report of each executable target execution satellite, so that the main satellite can also perform simple global optimization to ensure the operating efficiency of the system. Meanwhile, in the cooperative flow, each satellite utilizes local resources to calculate observation and analysis reports, so that the load of a calculation method of a main satellite is greatly reduced, the calculation resources of the satellite cluster can be fully utilized to improve the operation efficiency, and the robustness of the satellite cluster is also improved.
If the observation target abandoned due to observation of the emergency target is required to be observed by other satellites, a satisfactory solution can be obtained only by carrying out information interaction among the satellites for many times, so that a large load is caused on links among the satellites, the calculation load of each satellite is increased, and the decision efficiency of a satellite cluster is also reduced. Based on the current hardware conditions, a collaborative flow that can be converged after multiple iterations among the satellites is not recommended.
In an embodiment, in order to improve the operation efficiency of a star cluster system, the embodiment provides a single-star online scheduling method applied to a star cluster collaborative task planning. In order to facilitate later application expansion, the satellite can not only accept dynamically arrived emergency targets, but also execute dynamically arrived conventional targets, and the online scheduling algorithm of the autonomous satellite is preferentially used in the selection of the satellite single-satellite online scheduling method. The single-satellite online scheduling method of each satellite mainly adopts an agile satellite online branch-and-bound planning algorithm, and the triggering principle of the algorithm is that when a current target is observed, the next target to be observed is determined according to later task information. The on-line branch and bound planning algorithm of the agile satellite can effectively process the time sequence constraint of the agile satellite, the satellite looks ahead for a certain time each time, a local optimal solution is calculated by using the on-line branch and bound planning algorithm of the agile satellite, a first target in the solution is selected for observation, the state of the satellite is updated to the state after the target is selected after the observation is finished, and the looking ahead is continued until the observation judgment of the last target in the scene is finished. The main constraints of each satellite are a time window constraint, a time-dependent maneuvering time constraint and a fixed memory constraint.
Due to agile satelliteThe line branch and bound planning algorithm has high calculation efficiency (only one local task distribution is calculated each time), and can be used for an online scheduling method of autonomous satellites. When the execution satellite receives an emergency target observation requirement, the emergency targets are added into a target sequence to be observed according to the sequence of the overhead time, and a new observation scheme is generated by utilizing the idea of rolling planning. However, since the information of the emergency target is not known in advance and the satellite has a fixed energy constraint, the algorithm easily causes the fixed resource of the satellite to be consumed prematurely, so that the emergency target with a later trigger time cannot be effectively responded. In view of the related experience of the scheduling algorithm used on the ground learning satellite, the realization difficulty of the algorithm is reduced, and a self-adaptive task filtering mechanism is designed for the single-satellite online scheduling method of the satellite cluster. The task filtering mechanism enables the satellite to perform primary screening on a list of targets to be observed according to the residual solid storage amount, the income solid storage ratio of each target and other related information when deciding the next observation target, and selects the targets with higher observation cost performance for observation. The ratio of the observed profit to the target of profit-to-solid ratio to the observed solid consumption of the satellite for the target (p)k/durk*crjWherein p iskFor the observed yield of target k, durkThe imaging duration of target k, crjThe consumed inventory for imaging the satellite j unit time), a higher revenue inventory may be expected than a higher observation revenue per 1 unit of memory consumed by the satellite.
The "online scheduling method" in Step5 is the above-mentioned "task filtering mechanism", and specifically includes:
arranging all targets according to a descending order of the profit-to-solid ratio from high to low, multiplying the current solid residue of the execution satellite by a coefficient exSD to obtain a solid screening value, selecting the first N targets in a sorted target list, enabling the sum of the required observation solid residues of the targets to be just larger than the solid screening value, simultaneously enabling N to meet the formulas (49) and (50), and deleting the subsequent non-emergency targets in the list; after the tasks are preliminarily screened, a new observation planning scheme is generated by using an online scheduling method.
Figure GDA0002763071740000091
Figure GDA0002763071740000092
Wherein the expression of the profit-solid ratio is pk/durk*crjWherein p iskFor the observed yield of target k, durkThe imaging duration of target k, crjImmobilization consumed for imaging of satellite j unit time, sdRjThe remaining stocks of satellites at the decision time.
In one embodiment, on-board computing resources are limited while ensuring fast response to emergency objectives. Therefore, the task cooperative allocation algorithm of the main star in the star cluster should meet the related design requirements of fast and efficient calculation, timely observation of emergency targets, no excessive adjustment (no change of observation time windows) on the arranged emergency targets and the like. Aiming at the design requirement, on the basis of taking reference to other documents, 10 task cooperative allocation algorithms suitable for the inter-satellite computing environment are designed, and the characteristics of different algorithms are briefly analyzed. In the text, 10 heuristic distribution strategies are designed in total to guide the main satellite to carry out cooperative distribution on the emergency targets, so that the main satellite selects a proper execution satellite from execution satellites meeting the observation requirements of the emergency targets to observe the emergency targets by using different heuristic rules.
The executive satellite meeting the observation requirement of the emergency target meets the following three conditions: the requirement of an emergency target on the observation load is met; the requirement of the observation deadline of the emergency target is met; the existing emergency target observation condition of the satellite is not influenced.
The "task collaborative allocation algorithm" in Step4 includes the following 10 heuristic allocation strategies, which are respectively labeled heuA1To heuA10The parameter superscript j represents the serial number of the executing satellite, the parameter subscript i represents the current emergency target needing cooperative distribution, and each heuristic scoreThe algorithm corresponding to the configuration strategy is as follows:
algorithm 1: randomly selecting an execution satellite meeting the requirement of observing the emergency target to execute the observation task of the emergency target, wherein a specific execution formula is as follows:
Figure GDA0002763071740000093
wherein Sat is a set of satellites.
And 2, algorithm: arranging the emergency target at an executing satellite which can start observation at the earliest time, wherein the specific executing formula is as follows:
Figure GDA0002763071740000094
wherein the content of the first and second substances,
Figure GDA0002763071740000095
the start time of the visible time window for satellite j to target i.
Algorithm 3: the emergency target is arranged to the executive satellite with the most remaining solid deposits, the algorithm can balance the solid deposit load of the satellite to a certain extent, and the specific executive formula is as follows:
Figure GDA0002763071740000101
wherein, sdRjRepresenting the remaining inventory of satellite j.
And algorithm 4: arranging the emergency target to an execution satellite with lower target density, wherein the target density to be observed of the satellite is represented by using an observation slope (side swing angle difference value/time difference value) as an index; for satellite j, the observed slope between target i and target k:
Figure GDA0002763071740000102
wherein the content of the first and second substances,
Figure GDA0002763071740000103
for the observed slopes of target i and target k under satellite j,
Figure GDA0002763071740000104
for the observation yaw angle of satellite j for target i,
Figure GDA0002763071740000105
is the midpoint of the time window for satellite j to target i.
I.e. the absolute value of the difference between the yaw angles of the two targets compared to the absolute value of the difference between the midpoints of the upper time windows. As shown in fig. 3 below, the dotted line is the track direction of the satellite Sat2, and the observation slopes of the satellite for the target 1 and the target 9 are the same
Figure GDA0002763071740000106
Where Δ R is the yaw angle difference between the satellite and the two targets, and Δ t is the over-the-top time difference between the satellite and the two targets. When the observation slope of the target to the two targets is larger, it indicates that the lateral swing angle difference of the two targets is larger or the over-top time between the targets is closer, which means that the satellite is difficult to complete corresponding attitude maneuver between the two targets in continuous observation. Meanwhile, provision is made here for: and arranging the targets to be observed according to the ascending sequence of the over-top time, wherein the observation slope of the target i and the previous target is the subsequent observation slope of the target i, and the observation slope of the next target i is the subsequent observation slope. In FIG. 3
Figure GDA0002763071740000107
I.e. the subsequent observed slope of the target 9,
Figure GDA0002763071740000108
is the subsequent observed slope of target 9.
Arranging the targets to be observed according to the ascending order of the over-top time, respectively calculating the successive observation slope and the subsequent observation slope of the emergency target, and taking the larger value of the two observation slopes as a reference value, namely
Figure GDA0002763071740000109
Wherein the content of the first and second substances,
Figure GDA00027630717400001010
for the observation slope of satellite j to target i, pre (i) represents the target i's predecessor in the time series, and sub (i) is the target i's successor in the time series.
Figure GDA00027630717400001011
For the observed slope of target i and the following target,
Figure GDA00027630717400001012
representing the observed slope of target i with the subsequent target. When the emergency target is the first target in the target sequence to be observed, the successive observation slope of the emergency target is the difference value between the yaw angle of the target and the current yaw angle of the satellite, and the difference value between the target over-top time and the current time is compared with the difference value between the target over-top time and the current time. Meanwhile, when the emergency target is the last target in the target sequence to be observed, the subsequent observation slope of the emergency target is 0. The specific execution formula is as follows:
Figure GDA0002763071740000111
the heuristic distribution strategy enables the main satellite to select the satellite with the smaller emergency target observation slope to execute the observation task. The method ensures that the executing satellite has higher probability to directly insert the target into the original scheduling scheme under the condition of not influencing the timing constraint, and the allocation strategy is very effective when the conventional target of the satellite is sparse.
And algorithm 5: arranging the emergency target in a time window with the minimum competition degree, wherein the competition index of the time window is as follows:
Figure GDA0002763071740000112
wherein the content of the first and second substances,
Figure GDA0002763071740000113
denotes the competition index, | wT, of the visible time window of satellite j and target ijL represents the number of visible objects with satellite j,
Figure GDA0002763071740000114
representing the overlap time of the time windows of the two targets i, k. As shown in FIG. 4, the overlap time of the time windows of object 1 and object 6 in satellite Sat2 is
Figure GDA0002763071740000115
That is, the competition index of a certain time window is the sum of the overlapping time of the time window and the time windows of all the satellites. If the competition index of the time window of a certain target is smaller, the overlapping time of the time window and the time windows of other targets is less, namely, the targets are arranged in the time window, so that the observation of other targets by the satellite is less easily influenced. The specific execution formula is as follows:
Figure GDA0002763071740000116
and 6, algorithm: arranging the emergency target in an execution satellite with the longest time window length, wherein the time window with a smaller absolute value of the sidesway angle is longer, and the specific execution formula is as follows:
Figure GDA0002763071740000117
the maximum inclination angle is mainly determined by the composite angle of two observation angles of a yaw angle and a pitch angle due to the limitation of the maximum inclination angle of the attitude maneuver of the satellite. That is, when the absolute value of the image-side tilt angle between the satellite and the target is larger, the pitch angle of the satellite for observing the target is limited to a smaller range, i.e., the time length of the visible time window between the satellite and the target is shortened. Therefore, the time window is related to the absolute value of the roll angle of the satellite relative to the target, the time window with a large absolute value of the roll angle is shorter, and the time window with a small absolute value of the roll angle is longer. The specific execution formula is as follows:
Figure GDA0002763071740000121
and algorithm 7: arranging the emergency target at the executive satellite that will bring the largest gain increase to the satellite. The heuristic distribution strategy utilizes a greedy rule to improve the overall observation yield of the star cluster, namely the observation yield of the star cluster is inclined to be improved to the maximum extent when one emergency target is distributed, and the specific execution formula is as follows:
Figure GDA0002763071740000122
wherein the content of the first and second substances,
Figure GDA0002763071740000123
showing that the satellite j is in the original target list to be observed TLSjThe following observation planning scheme is used in the method,
Figure GDA0002763071740000124
the function Pro represents the sum of the observation gains of a certain observation planning scheme.
Algorithm 8: and arranging the emergency target to the executive satellite with the highest profit-to-solid ratio ranking of the target. The algorithm first calculates the observed solid memory consumption (dur) of satellite j for each targetk*crjWherein durkThe imaging duration of the object k, crjFor the solid spent imaging the satellite j unit time), the profit-to-solid ratio (p) for each target is calculatedk/durk*crjWherein p iskIs the observed profit of goal k) and sorts all the goals to be observed in ascending order of the profit-solid-reserve of the goals. In the new target sequence, the profit-survival ranking index of the target in the satellite j is calculated through the sequencing position of the emergent target i.
Figure GDA0002763071740000125
Wherein the content of the first and second substances,
Figure GDA0002763071740000126
percentage of revenue solid rank for Emergency goal i in satellite j, proSd (i) revenue solid rank order for goal i, | TLSjAnd + i | is the total number of targets to be observed of the satellite j. The algorithm arranges the emergency target to the satellite with the highest profit-to-solid ratio ranking, and can improve the use efficiency of the solid resources of the corresponding satellite to a greater extent, thereby improving the utilization efficiency of the solid resources of the whole satellite cluster. The specific execution formula is as follows:
Figure GDA0002763071740000127
algorithm 9: and arranging the emergency target to the execution satellite with the highest target observation rate so as to balance the conventional target observation condition of each execution satellite. The observation rate of the target is:
Figure GDA0002763071740000128
wherein, the obsPjRepresents the conventional target observation rate, TLS, of satellite jjA list of objects to be observed representing a satellite j,
Figure GDA0002763071740000131
indicating that the satellite j is considering the observation plan scheme of the emergency target i, and | l represents the total number of targets. The task cooperative allocation algorithm guides cooperative allocation of emergency targets by taking the target observation rate as an index, and can properly improve the completion rate of the targets so as to ensure effective observation of most targets. The specific execution formula is as follows:
Figure GDA0002763071740000132
the algorithm 10: in the above-mentioned distributed allocation strategy, a heuristic allocation strategy is randomly selected to complete the cooperative allocation of emergency targets. The specific execution formula is as follows:
Figure GDA0002763071740000133
in order to further optimize the cooperation strategy subsequently, the influence of the cooperation strategy and the target adaptive screening mechanism in the single-satellite online scheduling method provided in the embodiments on the operation efficiency of the satellite cluster is firstly subjected to experimental analysis.
1. Design of experiments
In the experimental design part, the effectiveness of a self-adaptive screening mechanism is verified, and the performance conditions of different cooperative strategies are analyzed. The main experimental parameters are scene attributes, emergency target information and conventional target information.
In a simulation verification scene, the collaborative execution satellite network is composed of 5 executable satellites, the 5 satellites are optical execution satellites of the same agile platform (relevant parameters are shown in a table), the orbital height of the satellites is 500km, the initial fixed memory is 900Gb, and the imaging fixed memory write code rate is 3 Gb/s. In the aspect of mobility, the maximum speed of the satellite attitude maneuver is 1 degree/s, and the acceleration of the satellite attitude maneuver is 0.5 degree/s2The acceleration at deceleration is 0.25 DEG/s2. The stable time is 5s, and the range of the yaw angle is [ -30 DEG, 30 DEG ]]The maximum compound inclination angle is 40 degrees. The conventional target number of each satellite follows Gaussian distribution with the mean value of 40 and the standard deviation of 10, the target distribution to be observed of each satellite is composed of a uniform area distribution and two key area distributions, and the specific gravity of the three is 6:1: 3.
TABLE 14 scene and satellite related parameters
Figure GDA0002763071740000141
Before describing the conventional target distribution of the satellite in detail, the related information of the emergency target is described, and the related parameters are shown in table 15. The number of emergency targets in the scene obeys a uniform distribution of 10 to 40, the visibility probability of each target to different satellites is 0.8, the triggering time of the targets obeys a uniform distribution of 600s to 900s after the scene starts, and the imaging deadline obeys a uniform distribution of 2600s to 4300s after the scene starts. The imaging duration of the emergency target is subject to Gaussian distribution with the mean value of 40s and the standard deviation of 10s, the observation yield is subject to Gaussian distribution with the mean value of 100 and the standard deviation of 15, the imaging side swing angles of the emergency target and each satellite are subject to uniform distribution of-30 degrees to 30 degrees, and the over-vertex time of each satellite which has a visual relationship with the emergency target to the target is subject to uniform distribution of 200s to 4300s after the scene begins.
Table 15 emergency objective related parameters
Figure GDA0002763071740000142
The conventional target consists of three areas, namely a uniform area, a key area 1 and a key area 2, and accounts for 60%, 10% and 30% of the conventional target number of each satellite respectively. The combination mode represents a more complex target distribution situation, and the performance of the algorithm can be better tested.
The targets in the uniform area distribution are mainly point targets (related parameters are shown in table 16), the overhead time of the satellite is subjected to uniform distribution of 30s to 4000s after the scene starts, the imaging time of the targets is 10s, the tilt angle of the targets and the observation side of the satellite is subjected to uniform distribution of-30 degrees to 30 degrees, the imaging yield is subjected to Gaussian distribution with the mean value of 30 and the standard deviation of 10.
TABLE 16 Uniform region target distribution-related parameters
Figure GDA0002763071740000151
Target distribution parameters of the key area 1 are shown in the table, the target is also a point target, the satellite vertex-passing time obeys Gaussian distribution with the mean value of 3500s and the standard deviation of 600 s; the observation side sway angle of the satellite follows Gaussian distribution with the mean value of-20 degrees and the standard deviation of 5 degrees; the imaging duration of the target was 10s, the target benefit obeyed a gaussian distribution with a mean of 40 and a standard deviation of 5. The targets in the region are relatively uniformly distributed and more concentrated, and certain challenges are provided for processing scheduling algorithm timing constraints.
TABLE 17 distribution-related parameters of objects in region of interest 1
Figure GDA0002763071740000152
The target distribution parameters of the region of interest 2 are shown in the table, and the target is composed of a point target and a strip target, wherein the strip target accounts for 50% of the targets in the region. The overhead time of the target in the area is subject to uniform distribution from 130s to 1430s after the scene begins; observing the side swing angle to obey the uniform distribution of minus 20 degrees to 10 degrees; the imaging time of the point target is still 10s, the imaging yield obeys Gaussian distribution with the mean value of 40 and the standard deviation of 10; the imaging duration of the strip target follows Gaussian distribution with the mean value of 20s and the standard deviation of 3 s; the imaging yield follows a gaussian distribution with a mean value of 60 and a standard deviation of 10.
TABLE 18 distribution-related parameters of the objects in the region of emphasis 2
Figure GDA0002763071740000161
And generating 100 experimental scenes according to the scene design, and counting the total observation yield of the star cluster by using different cooperative allocation algorithms under the condition of the existence and non-existence of a single-star self-adaptive target filtering mechanism. Through comparative analysis, the effectiveness of the adaptive target filtering mechanism is tested, and then the related performances of different distribution cooperative distribution algorithms are analyzed.
2. Results and analysis of the experiments
The adaptive filtering mechanism observes the revenue impact for different algorithms in 100 test scenarios as shown below. The algorithm number is as follows: the algorithm 1 randomly selects an execution satellite meeting the observation requirement; algorithm 2 arranges the emergency targets to the executive satellites that can begin imaging earliest; algorithm 3 arranges the emergency targets to the executive satellites with the most remaining stocks; the algorithm 4 uses the time slope as a target density index, and arranges the emergency target to the execution satellite with the minimum observation slope; the algorithm 5 arranges the emergency target to the execution satellite with the minimum competition index of the time window; algorithm 6 is to schedule the emergency object at the executive satellite with the longest time window; the algorithm 7 arranges the emergency target in an execution satellite which can bring the maximum gain increment to the satellite; the algorithm 8 arranges the emergency target to the executive satellite with the highest profit-to-solid ratio ranking; the algorithm 9 arranges the emergency target to the execution satellite with the highest target observation rate; the algorithm 10 randomly selects a heuristic algorithm to accomplish the cooperative allocation of emergency objectives.
As can be seen from fig. 5, the imaging gain of the adaptive filtering mechanism for all task cooperative allocation algorithms is improved in different extents, which indicates that the method can effectively improve the utilization efficiency of satellite persistent resources, and the overall imaging gain is improved by imaging a target with a higher profit-to-solid ratio. As can also be seen from the figure, after the adaptive filtering mechanism is added, the inter-algorithm measurement gap is reduced, which shows that the adaptive filtering mechanism can achieve a better optimization effect on the cooperative scheduling and the single-star scheduling of most scenes, and has a certain general value.
When the adaptive filtering mechanism is not used, the gains of the algorithm are ranked from high to low as: 7,2,4,6,8,10,5,9,3,1. After using the adaptive filtering mechanism, the gains of the algorithm are ranked from high to low as: 8,2,5,7,6,10,4,1,9,3. This shows that the combination of the filtering mechanism and different allocation strategies produces different effects, the original algorithm 8 focuses on the optimization of the benefit-to-solid ratio, and the heuristic rule and the filtering mechanism are mutually overlapped, so that the performance of the algorithm 8 is not greatly improved after the filtering mechanism is added to other algorithms. The algorithm 7 is a heuristic index which takes the increment of imaging profit as the cooperative allocation of emergency targets, but the index easily causes the satellite cluster to pay too much attention to short-term profit, and under the condition of all target information at the position, the satellite can consume a large amount of solid resources too early, so that the observation of high-profit targets at the later stage of the scene cannot be performed. Through the cooperation with the adaptive filtering mechanism, the satellite can also use the solid resources in a longer period when the short-term benefit is considered to be improved, so that a higher performance level can be obtained.
As can be seen from the difference analysis fig. 6, the adaptive filtering mechanism has a large influence on the algorithm 4 and a small influence on the algorithm 8.
The revenue ranking statistics in different scenarios for each algorithm are shown in fig. 7-11. As known from the ranking distribution of the task collaborative allocation algorithms, each task collaborative allocation algorithm has a relatively suitable application scene, namely, no algorithm can adapt to the application of various scenes. Where algorithm 7 performed well, the highest overall benefit was obtained in 25 of the 100 scenarios, but algorithm 7 ranked 6 (including 6) later in 29 scenarios.
4. Selection algorithm of cooperative allocation algorithm
From the above experimental analysis, it can be known that no single algorithm can perform well in all scenes, so the idea of algorithm selection in machine learning is introduced to further improve the overall benefit of the task collaborative allocation algorithm. Firstly, various expressions of different task collaborative allocation algorithms are analyzed through a large number of scenes, relevant scene parameters are extracted, and the task collaborative allocation algorithm which has higher probability to obtain higher star cluster observation income is selected at the scene starting time according to known information. Due to the fact that the emergency targets are dynamically arrived, only part of information of the conventional targets of each satellite is known at the scene starting time, feature extraction is conducted on the conventional targets of the satellites, relevant machine learning training is conducted, and then a proper cooperative distribution algorithm is selected according to corresponding information of a new scene.
The traditional machine learning needs to calculate a large amount of training data, but the experimental environment used by the method is limited (the CPU dominant frequency of the experimental environment is 2.0-GHz, and the internal memory is 3.79GB RAM.), and large-scale machine learning cannot be carried out. Here, only the idea of machine learning is applied to verify the validity of the idea and develop a machine learning strategy that can be used in a small personal computer to improve the overall performance of the task co-allocation algorithm.
The "task collaborative allocation algorithm" in Step4 including 10 heuristic allocation strategies will be analyzed below.
Taking the case here as an example, the average performance of algorithm 7 is the best among the 10 algorithms. It is therefore desirable to further enhance the performance of the algorithm 7 through the idea of machine learning. Algorithm 7 ranks after 6 (including 6 th) scenes out of 100 scenes 29 scenes. In the scenario with algorithm 7 ranked 6 or later, the average performance of all other algorithms is shown in fig. 12. As can be seen from the figure, when the rank of algorithm 7 is greater than or equal to 6, the average ranks of the other algorithms are from low to high, 2,4,6,5,8,10,9,7,3, 1; the profit ranks from high to low as 2,4,5,8,6,10,9,7,3, 1. It can be seen that algorithm 2 performs better when algorithm 7 performs poorly, both in terms of average ranking and overall profitability, indicating a greater likelihood of further improving the performance of algorithm 7 when it cooperates with algorithm 2. Algorithm 2 is referred to herein as a complement to algorithm 7.
As can be seen from the above analysis, the emergency target arrives dynamically in each scene, so the position, benefit, latest imaging time requirement, visibility relationship with each satellite, and other information of the emergency target at the scene start time are unknown. The scene starting time can be known only as partial information of the conventional targets of each satellite, and due to the distribution of the conventional targets of different satellites, the information such as observation duration and income is different, and the number of targets is different. Therefore, in order to facilitate the stability of the input data dimension in machine learning, some feature information is selected to describe the conventional targets of different execution satellites.
The selection of the conventional target parameters is described in detail below.
Because the selected feature information can only describe the conventional target list in a one-sided manner, a large amount of specific information is lost, and in order to describe the conventional target relatively fully, the distribution situation of the conventional target should be described from a plurality of angles as much as possible. Here, the following 12 parameters are chosen to describe a regular target list of executing satellites. The method specifically comprises the following steps: 1. a target number; 2. a target total gain; 3. target average profit; 4. standard deviation of target revenue; 5. observing the total solid reserve consumption of all targets; 6. a target profit-solid-memory ratio mean value; 7. standard deviation of target profit-solid-reserve ratio; 8. pre-planning scheme revenue; 9. the number of observation tasks in the pre-planning scheme; 10. residual satellite solid storage after the pre-planning scheme is executed; 11. target average subsequent observation slope; 12. target subsequent observed slope standard deviation. The above parameters are calculated as follows (subscript i in the parameter formula indicates the observed object, superscript j indicates the satellite number, and there are 12 parameters for the conventional object list for each satellite j).
Parameter 1 is the total number of conventional targets, specifically:
para1＝|TLSj|
wherein, TLSjConventional target list, | TLS, representing satellite jjAnd | represents the number of targets in the target list.
The parameter 2 is a general target total yield, and specifically comprises the following steps:
Figure GDA0002763071740000191
wherein p isiThe observation yield of the observation target i is obtained.
The parameter 3 is a conventional target average profit, specifically:
Figure GDA0002763071740000192
the parameter 4 is the statistical standard deviation of the conventional target revenue, and specifically is:
Figure GDA0002763071740000193
parameter 5 is the total solid consumption observed for all conventional targets, specifically:
Figure GDA0002763071740000194
wherein, duriThe imaging duration of the object i, crjImaging code rate (in units) for satellite jThe spent solid stock of time imaging).
The parameter 6 is the average value of the profit-solid-reserve ratio of the conventional target, and specifically is as follows:
Figure GDA0002763071740000195
the parameter 7 is the statistical standard deviation of the profit-to-solid ratio of the conventional objective, specifically:
Figure GDA0002763071740000201
the parameter 8 is a plan benefit corresponding to a pre-planning plan in which the satellite only considers a conventional target, and specifically includes:
Figure GDA0002763071740000202
wherein
Figure GDA0002763071740000203
Indicating that satellite j is in the regular object list TLSjThe following pre-planning scheme, which is obtained by the above mentioned rolling planning algorithm in combination with the adaptive target filtering mechanism.
The parameter 9 is the total number of the conventional targets observed in the pre-planning scheme, and specifically includes:
Figure GDA0002763071740000204
wherein the content of the first and second substances,
Figure GDA0002763071740000205
satellite j in the conventional object list TLSjThe number of observed targets in the lower preplanning scheme.
The parameter 10 is the remaining satellite solid memory after the satellite executes the pre-planning scheme, and specifically includes:
Figure GDA0002763071740000206
wherein sdRIjRepresenting the amount of initial fixed resources for satellite j,
Figure GDA0002763071740000207
representing the total amount of solid resources consumed to observe the targets in all pre-planned scenarios.
The parameter 11 is the average subsequent observation slope of the conventional target, specifically:
Figure GDA0002763071740000208
wherein the content of the first and second substances,
Figure GDA0002763071740000209
for the observation yaw angle of satellite j for target i,
Figure GDA00027630717400002010
the time window midpoint of satellite j to target i, sub (i), the next target of target i, i.e., the next passing target after satellite j passes the top of target i. When the object i is the last object,
Figure GDA00027630717400002011
the parameter 12 is a statistical standard deviation of a subsequent observation slope of a conventional target, and specifically comprises:
Figure GDA0002763071740000211
the parameter 1 and the parameter 2 represent the upper limit of the observation yield of the satellite on the conventional target, the parameter 3 to the parameter 6 represent the average observation cost performance of the satellite on the conventional target, the parameter 7 to the parameter 10 provide a reference value for the execution of the observation of the satellite on the conventional target, and the parameter 11 and the parameter 12 represent the distribution density of tasks and the attitude maneuver difficulty degree of the satellite for continuously observing a plurality of targets through the observation slope (detailed).
The experimental results and analysis are as follows:
in the preparation stage of learning data, a 12-dimensional characteristic information vector is generated for each executing satellite, and the known information is analyzed by using a machine learning method to predict which collaborative allocation algorithm should be used in different scenes so as to obtain better global observation benefits. Here, taking the example of the cooperative observation network composed of 5 executing satellites, the input vector of machine learning is 5 × 12 — 60-dimensional vector. In the generation of the learning vector, if algorithm 2 is better than algorithm 7, the classification result is-1, otherwise, it is 1. The method is used for training a support vector machine to learn the classification capability of the algorithm by analyzing data of a plurality of scenes, and then the classification capability of the algorithm is tested in a new scene set.
In the experimental part, 500 scenes are used for generating a learning vector of a classification algorithm, and 100 scenes are generated for detecting the effectiveness of the cooperative allocation algorithm guided by algorithm classification.
Fig. 13 shows the sum of gains of different task co-allocation algorithms in the new 100 scenarios. Where algorithms 1 through 10 are again the heuristic allocation strategy mentioned above, algorithm 11 is an algorithm selector trained by the support vector machine. The algorithm 11 selects an appropriate algorithm from the algorithms 7 and 2 to perform cooperative allocation of emergency targets of the corresponding scene by using the scene known information at the beginning stage of each scene. From the data, it can be seen that the algorithm 11 performs better than the original best algorithm (algorithm 7) in the new scene test set.
TABLE 19 observed yield for different algorithms in test set
Figure GDA0002763071740000221
Error statistics of algorithm selector as shown in fig. 14, 8 scenes out of 100 scenes select the wrong task co-allocation algorithm, that is, the overall benefit of the selected algorithm is low. The accuracy (92%) of the algorithm selector is better, which shows that the emergency target task cooperative allocation algorithm can be screened by the relevant characteristics of the conventional target information, and the effectiveness of the characteristic information selected by the invention is also proved.
Of these, 6 scenarios select allocation algorithm 7 when allocation algorithm 2 has a higher gain, and 2 scenarios select allocation algorithm 2 when allocation algorithm 7 has a higher gain. This is because the algorithm 7 performs better than the algorithm 2 as a whole, and therefore the trained machine model is more likely to select the algorithm 7 that performs better as a whole.
The following conclusions can be drawn through the analysis of the experimental results,
1. in different scenes, the distribution situation of the conventional targets of the satellite can influence the performance of the collaborative allocation algorithm of different emergency targets.
2. A more appropriate emergency target cooperative allocation algorithm can be determined by analyzing known satellite conventional targets;
3. the feature vector provided by the invention can better describe the distribution condition of the conventional satellite target;
4. and a proper emergency target cooperative allocation algorithm can be more reasonably selected through a machine learning correlation method.
The invention initially discusses the structural design and the cooperative flow of the execution satellite constellation consisting of a plurality of autonomous satellites. On the basis, by means of the research result of the single-satellite online scheduling method, the dynamic emergency target cooperative allocation algorithm of the main satellite in the star satellite cluster is deeply researched. Through experimental verification of 10 designed heuristic task cooperative allocation algorithms, the influence of different task cooperative allocation algorithms on the operation efficiency of the star clusters is analyzed. And finally, designing a task cooperative allocation algorithm selection mechanism based on a related method of a support vector machine in machine learning. The mechanism can select a proper cooperative distribution algorithm in different application scenes according to the conventional target distribution information of each satellite, so that the observation efficiency of the satellite cluster is further improved.
Finally, it should be pointed out that: the above examples are only for illustrating the technical solutions of the present invention, and are not limited thereto. Those of ordinary skill in the art will understand that: modifications can be made to the technical solutions described in the foregoing embodiments, or some technical features may be equivalently replaced; such modifications or substitutions do not depart from the spirit and scope of the corresponding technical solutions of the embodiments of the present invention.

Claims (2)

1. A target observation method of a multi-satellite cluster is characterized in that a control structure of the multi-satellite cluster is a centralized-distributed structure and comprises a main satellite and a plurality of execution satellites; the observation tasks of the satellite constellation comprise a conventional target observation task and an emergency target cooperative observation task, and the target observation method of the multi-satellite constellation comprises the following steps:
step1, generating a pre-planned observation scheme for the conventional target by each satellite by using a rolling planning algorithm, and executing the current planned observation scheme;
step2, when a new emergency target arrives, the main satellite firstly screens load information capable of executing the task, then sends the emergency target information to a corresponding execution satellite through an inter-satellite link in a broadcasting mode, and sets the receiving time of an observation and analysis report;
step3, the execution satellite receiving the emergency target information transmitted by the main satellite estimates the own observation cost of the emergency task by using local computing resources, generates an observation report and sends the observation report back to the main satellite;
step4, determining the execution satellite for executing the emergency target observation task by the main satellite according to the observation report of each execution satellite and by using a task cooperative allocation algorithm, and assigning the emergency target observation task to the corresponding execution satellite; the conditions met by the execution satellite for executing the emergency target observation task comprise: 1. the requirement of an emergency target on the observation load is met; 2. the requirement of the observation deadline of the emergency target is met; 3. the existing emergency target observation condition of the satellite is not influenced;
step5, adding a new emergency target observation task into a task to be planned later by the execution satellite receiving the emergency target observation task assignment, updating the own observation scheme by using an online scheduling method, executing the new observation planning scheme, and continuously executing the original observation planning scheme by other satellites;
wherein the main constraints of each executing satellite are a time window constraint, a time-dependent maneuvering time constraint and a fixed memory constraint; the emergency objective information includes an emergency objective ID, location information, observed load demand, observed revenue, and latest imageable time,
the online scheduling method in Step5 specifically comprises the following steps:
arranging all targets according to a descending order of the profit-to-solid ratio from high to low, multiplying the current solid deposit allowance of the execution satellite by a coefficient exSD to obtain a solid deposit screening value, selecting the first N targets in a sorted target list, enabling the sum of the required observation solid deposits of the targets to be just larger than the solid deposit screening value, simultaneously enabling N to meet the following formulas (49) and (50), and deleting the subsequent non-emergency targets in the list; after the tasks are primarily screened, a new observation planning scheme is generated by using an online scheduling method;
Figure FDA0002763071730000021
Figure FDA0002763071730000022
the expression of the profit-solid ratio is pk/durk*crjWherein p iskFor the observed yield of target k, durkThe imaging duration of target k, crjImmobilization consumed for imaging of satellite j unit time, sdRjThe remaining stocks of satellites at the decision time.
2. The target observation method of the multi-star cluster as claimed in claim 1, wherein the "task cooperative distribution algorithm" in Step4 includes 10 heuristic distribution strategies, each heuristic distribution strategy is labeled heuA respectively1To heuA10The parameter superscript j represents the serial number of the satellite, the parameter subscript i represents the current emergency target needing cooperative distribution, and each heuristic scoreThe algorithm corresponding to the configuration strategy is as follows:
algorithm 1: randomly selecting an execution satellite meeting the requirement of observing the emergency target to execute the observation task of the emergency target, wherein a specific execution formula is as follows:
Figure FDA0002763071730000023
wherein Sat is a satellite set;
and 2, algorithm: arranging the emergency target at an executing satellite which can start observation at the earliest time, wherein the specific executing formula is as follows:
Figure FDA0002763071730000024
wherein the content of the first and second substances,
Figure FDA0002763071730000025
the start time of the visible time window for satellite j to target i;
algorithm 3: the emergency target is arranged to the executive satellite with the most remaining solid deposits, the algorithm can balance the solid deposit load of the satellite to a certain extent, and the specific executive formula is as follows:
Figure FDA0002763071730000026
wherein, sdRjRepresents the remaining stocks of satellite j;
and algorithm 4: arranging the emergency target to an execution satellite with lower target density, wherein the target density to be observed of the satellite is represented by taking an observation slope as an index; for satellite j, the observed slope between target i and target k:
Figure FDA0002763071730000027
wherein the content of the first and second substances,
Figure FDA0002763071730000028
for the observed slopes of target i and target k under satellite j,
Figure FDA0002763071730000029
for the observation yaw angle of satellite j for target i,
Figure FDA0002763071730000031
the midpoint of the time window for satellite j to target i;
and algorithm 5: arranging the emergency target in a time window with the minimum competition degree, wherein the competition index of the time window is as follows:
Figure FDA0002763071730000032
wherein the content of the first and second substances,
Figure FDA0002763071730000033
denotes the competition index, | wT, of the visible time window of the executing satellite j and the target ijL represents the number of visible objects that are owned by the executing satellite j,
Figure FDA0002763071730000034
represents the overlap time of the time windows of the two targets i, k; if the smaller the competition index of the time window of a certain target is, the less the overlapping time of the time window and the time windows of other targets is, the specific execution formula is:
Figure FDA0002763071730000035
and 6, algorithm: arranging the emergency target in an execution satellite with the longest time window length, wherein the time window with a smaller absolute value of the sidesway angle is longer, and the specific execution formula is as follows:
Figure FDA0002763071730000036
and algorithm 7: arranging the emergency target to an execution satellite with the largest income increment, wherein the heuristic distribution strategy improves the overall observation income of the satellite cluster by using a greedy rule, and the specific execution formula is as follows:
Figure FDA0002763071730000037
wherein the content of the first and second substances,
Figure FDA0002763071730000038
showing that the satellite j is in the original target list to be observed TLSjThe following observation planning scheme is used in the method,
Figure FDA0002763071730000039
representing an observation planning scheme of a satellite j under consideration of an emergency target i, and a function Pro () representing the sum of observation gains of a certain observation planning scheme;
algorithm 8: the emergency objectives are assigned to the executing satellites that achieve higher profitability and/or better than the desired cost of the solid-to-solid ratio, and the algorithm first calculates the observed solid-to-solid consumption dur for each objective for satellite j according to equation (61)k*crjWherein durkThe imaging duration of the object k, crjFor the solid memory consumed by imaging the satellite i unit time, the profit-solid memory ratio p of each target is calculatedk/durk*crjWherein p iskThe observation gains of the target k are obtained, all targets to be observed are sorted in an ascending order of the gain-to-solid-ratio of the target, in a new target sequence, the gain-to-solid ranking index of the target in the satellite j is calculated according to the sorting position of the emergency target i,
Figure FDA00027630717300000310
wherein the content of the first and second substances,
Figure FDA00027630717300000311
percentage of revenue solid rank for Emergency goal i in satellite j, proSd (i) revenue solid rank order for goal i, | TLSj+ i | is the total number of targets to be observed of the satellite j, and the specific execution formula is as follows:
Figure FDA0002763071730000041
algorithm 9: giving the emergency target to an execution satellite with the highest target observation rate to balance the conventional target observation conditions of each satellite, wherein the target observation rate is as follows:
Figure FDA0002763071730000042
wherein, the obsPjRepresents the conventional target observation rate, TLS, of satellite jjA list of objects to be observed representing a satellite j,
Figure FDA0002763071730000043
the observation planning scheme of the satellite j considering the emergency target i is represented, wherein | l represents the total number of targets, and the specific execution formula is as follows:
Figure FDA0002763071730000044
the algorithm 10: in the above-mentioned distribution strategy, a heuristic distribution strategy is randomly selected to complete the cooperative distribution of the emergency targets, and the specific execution formula is as follows:
Figure FDA0002763071730000045
the index j in the above parameters represents the satellite serial number, and the parameter index i represents the current emergency target needing cooperative allocation.
CN201711252589.4A 2017-12-01 2017-12-01 Target observation method for multi-satellite cluster Active CN108021433B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201711252589.4A CN108021433B (en) 2017-12-01 2017-12-01 Target observation method for multi-satellite cluster

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201711252589.4A CN108021433B (en) 2017-12-01 2017-12-01 Target observation method for multi-satellite cluster

Publications (2)

Publication Number Publication Date
CN108021433A CN108021433A (en) 2018-05-11
CN108021433B true CN108021433B (en) 2021-03-19

Family

ID=62077911

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201711252589.4A Active CN108021433B (en) 2017-12-01 2017-12-01 Target observation method for multi-satellite cluster

Country Status (1)

Country Link
CN (1) CN108021433B (en)

Families Citing this family (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109299861B (en) * 2018-09-04 2021-08-27 中国人民解放军国防科技大学 On-satellite scheduling method of agile satellite
CN109728845B (en) * 2018-11-07 2021-04-13 长沙天仪空间科技研究院有限公司 Satellite efficient scheduling constellation and scheduling method
CN109767128B (en) * 2019-01-15 2021-06-11 中国人民解放军国防科技大学 Imaging satellite autonomous task planning method based on machine learning
CN110782180A (en) * 2019-11-05 2020-02-11 中国人民解放军国防科技大学 Blackboard model-based multi-satellite task online distribution method and system
CN116347623B (en) * 2023-05-29 2023-08-15 之江实验室 Task scheduling method and device, storage medium and electronic equipment

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102867107A (en) * 2012-08-16 2013-01-09 中国人民解放军国防科学技术大学 Multi-imaging satellite emergency task dynamic scheduling method
CN105095643A (en) * 2015-06-15 2015-11-25 中国人民解放军国防科学技术大学 Method for planning autonomous task of imaging satellite in dynamic environment
CN106059652A (en) * 2016-05-30 2016-10-26 中国电子科技集团公司第五十四研究所 Heuristic rapid satellite data receiving resource joint distribution method
CN106516175A (en) * 2016-11-28 2017-03-22 中国人民解放军国防科学技术大学 Autonomous operation and control system of agile imaging satellite, and operation process of system

Family Cites Families (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US8428897B2 (en) * 2008-04-08 2013-04-23 Massachusetts Institute Of Technology Method and apparatus for spectral cross coherence
CN102479085B (en) * 2010-11-30 2015-04-01 中国人民解放军国防科学技术大学 Agile satellite task planning method
CN103744733B (en) * 2013-07-03 2017-03-22 中国人民解放军国防科学技术大学 Method for calling and configuring imaging satellite resources
CN104063749B (en) * 2014-06-28 2017-03-29 中国人民解放军国防科学技术大学 A kind of autonomous mission planning algorithm of the imaging satellite based on roll stablized loop
KR101942208B1 (en) * 2015-01-08 2019-01-24 애리스 엔터프라이지즈 엘엘씨 Server-side Adaptive Bitrate Control for DLNA HTTP Streaming Clients
CN105096012B (en) * 2015-09-23 2018-06-08 中国电子科技集团公司第五十四研究所 A kind of quick satellite dynamic layout method based on varying glide window
CN105512479B (en) * 2015-12-03 2018-09-14 中国地质大学（武汉） A kind of evaluation method and device of more star earth observation mission planning algorithms

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102867107A (en) * 2012-08-16 2013-01-09 中国人民解放军国防科学技术大学 Multi-imaging satellite emergency task dynamic scheduling method
CN105095643A (en) * 2015-06-15 2015-11-25 中国人民解放军国防科学技术大学 Method for planning autonomous task of imaging satellite in dynamic environment
CN106059652A (en) * 2016-05-30 2016-10-26 中国电子科技集团公司第五十四研究所 Heuristic rapid satellite data receiving resource joint distribution method
CN106516175A (en) * 2016-11-28 2017-03-22 中国人民解放军国防科学技术大学 Autonomous operation and control system of agile imaging satellite, and operation process of system

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
"多星应急观测协同规划及动作序列优化调整方法";彭双;《中国优秀硕士学位论文全文数据库基础科学辑》;20170315(第201703期);第27-31、44页 *
"敏捷成像卫星自主任务规划方法";刘嵩;《计算机集成制造系统》;20160430(第201604期);第930页 *

Also Published As

Publication number Publication date
CN108021433A (en) 2018-05-11

Similar Documents

Publication Publication Date Title
CN108055067B (en) Multi-satellite online cooperative scheduling method
CN108021433B (en) Target observation method for multi-satellite cluster
CN112016812B (en) Multi-unmanned aerial vehicle task scheduling method, system and storage medium
CN112766813B (en) Space-day collaborative observation complex task scheduling method
CN109960544B (en) Task parallel scheduling method based on data driving type agile satellite
CN109767128B (en) Imaging satellite autonomous task planning method based on machine learning
CN110134146B (en) Distributed multi-unmanned aerial vehicle task allocation method under uncertain environment
CN112418497B (en) Material distribution path optimization method for manufacturing Internet of things
CN104063749B (en) A kind of autonomous mission planning algorithm of the imaging satellite based on roll stablized loop
CN111311074B (en) Multi-satellite distributed cooperative rescheduling method facing emergency tasks
CN112508369B (en) Multi-unmanned aerial vehicle task allocation method based on improved ant colony algorithm
CN110515708A (en) The in-orbit autonomous mission planning method and system of satellite
CN113852405B (en) Construction method and device of multi-beam relay satellite task scheduling model
CN108304958A (en) A kind of list star multi-load Complex Constraints mission planning method and system
CN108880653A (en) Relay satellite system method for managing resource and system
Yao et al. Task allocation strategies for cooperative task planning of multi-autonomous satellite constellation
CN107679748A (en) Towards the star ground cooperation method of group of stars observation mission contexture by self
CN109039428B (en) Relay satellite single-address antenna scheduling random search method based on conflict resolution
CN114415735B (en) Dynamic environment-oriented multi-unmanned aerial vehicle distributed intelligent task allocation method
CN109347536A (en) A kind of spatial network monitoring resource condition system based on situation knowledge
CN113487220A (en) Static target observation-oriented space-sky heterogeneous earth observation resource cooperative scheduling method
CN114879745B (en) Multi-unmanned aerial vehicle collaborative task planning method based on preplanning and tabu search algorithm
Yang et al. Onboard coordination and scheduling of multiple autonomous satellites in an uncertain environment
CN112862306A (en) Multi-satellite online collaborative task planning method for time-sensitive moving target tracking
CN102164420A (en) Many-to-one communication conflict processing method based on satellite network

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant