Future beneficial distributed satellite online scheduling method for emergency task arrival

Abstract

The invention discloses a future beneficial distributed satellite online scheduling method for emergency task arrival, and the method comprises the steps: building an emergency task arrival-oriented satellite online scheduling model for solving a problem that an SOS-ET problem lacks a standardized mathematical model and a high-performance distributed online task planning algorithm; the invention also provides a future beneficial distributed satellite online scheduling algorithm for emergency task arrival, and the algorithm comprises a distributed satellite online scheduling algorithm with beneficial bid invitation efficiency expectation and a distributed satellite online scheduling algorithm with beneficial bid invitation efficiency expectation. Therefore, the situation that low-income tasks arriving at the early stage occupy precious satellite imaging resources in the SOS-ET problem is avoided, and resources are reserved for high-income tasks at the later stage.

Description

Technical Field

[0001] The present invention belongs to the field of mission planning and scheduling, and particularly relates to a future-beneficial distributed satellite online scheduling method for emergency task arrival. Background Art

[0002] For sudden hotspot events, remote sensing satellites can discover and generate a series of remote sensing requirements through their equipped anomaly detection settings. Such randomly arriving remote sensing requirements are called emergency tasks. With the significant improvement of satellite system capabilities, users' real-time and personalized requirements for satellite services, especially for remote sensing emergency tasks, are continuously increasing. For example, the online tracking task of a sudden moving target belongs to an emergency task. Users need the satellite to continuously track the target and generate subsequent remote sensing satellite scheduling plans based on the previous target remote sensing data to achieve real-time monitoring of the moving target. Such emergency tasks usually have high real-time and strong uncertainty, that is, the tasks usually arrive randomly over time, and the arrival quantity at each moment is unknown, requiring the satellite scheduling algorithm to respond to them in real time. Therefore, researching the satellite online scheduling problem for emergency task arrival (Satellite Online Scheduling problem for Emergency Task, SOS-ET) has relatively significant practical significance.

[0003] When solving the SOS-ET problem, centralized methods are difficult to apply. Specifically, centralized methods rely heavily on the master controller, which makes them easily affected by the failure of the master controller and have a relatively high risk. Compared with centralized methods, distributed methods have gradually become a research hotspot due to their reliability, flexibility, and decentralized architecture design, and are widely applied to distributed satellite systems. As shown above, this problem scenario can be described as that emergency tasks arrive at the distributed satellite system in real time and randomly. Before the tasks arrive, their specific task information and arrival time are unknown to the satellite system. In addition, since the satellite system is a distributed system, the satellites are also in an unknown state about each other's information and need to communicate and negotiate through inter-satellite links to obtain the status information of other satellites in order to finally achieve the collaborative allocation of the online arriving tasks. At this time, a utility-based online auction (UOA) algorithm that considers maximizing the current efficiency is usually adopted as the solution algorithm for the SOS-ET problem. As Figure 1 shown, the UOA algorithm simulates the bidding process in the market economy through modes such as tender invitation, tender submission, and bid competition, providing an effective information interaction and self-organization mechanism for numerous agents in the system. Usually, a round of auction includes four key steps: demand release, tender specification, tender evaluation, and final transaction confirmation.

[0004] Communication 1: Requirement Release. In the constellation system, the cluster head satellite acts as the tendering satellite. After collecting the tasks arriving online, it releases the relevant task requirements to all other member satellites in the cluster, and the remaining member satellites automatically act as bidding satellites to initiate the bidding process.

[0005] Communication 2: Bid Document Formulation. After receiving the information related to the task requirements, the bidding satellite starts the bidding procedure, evaluates according to its current own status information, and formulates a bid document.

[0006] Communication 3: Bid Evaluation. After receiving the bid documents feedback from all bidding satellites or after exceeding the predetermined deadline, the tendering satellite will end the tender and initiate the bid evaluation procedure: according to the received bid documents, determine and notify the winning satellite, and assign the task to the winning satellite.

[0007] Communication 4: Transaction Confirmation. After receiving the winning notice, the winning satellite notifies the tendering satellite to remove the task, and the two parties reach an agreement to confirm the task assignment. At the same time, the winning satellite incorporates the task into its task pool and updates its ongoing task plan.

[0008] The UOA algorithm can provide a feasible satellite online scheduling scheme for the SOS-ET problem. This algorithm plans all the tasks arriving currently based on the greedy criterion. Specifically, for the SOS-ET problem under the UOA algorithm, the tendering satellite starts the tender for all the received tasks, sends the bid documents to other satellites in the distributed satellite system, and the other satellites act as bidding satellites. On the premise of meeting the current constraints, they return the task planning scheme as the bid document to the tendering satellite. In this process, the satellite will only consider the current task revenue during the bid evaluation or bidding process, and try to complete the current task as much as possible without considering the revenue of the tasks arriving in the future. Therefore, in the resource-limited scenario such as the distributed satellite system, the UOA algorithm still has the following defects:

[0009] (1) There is a lack of a satellite online scheduling problem model with high applicability for emergency task arrivals. It is necessary to analyze diverse task requirements and differentiated satellite resources, and then guide the design of the algorithm. How to uniformly process complex objectives, and standardly give the mathematical expressions of decision variables, constraints, and revenues to help understand the combinatorial optimization essential characteristics of multi-complex objective scheduling, and then better guide the design of algorithms and operators is an important topic.

[0010] (2) When the UOA algorithm performs online task allocation, it only needs to consider the immediate benefits of the current task, without being forward-looking and unable to comprehensively evaluate the potential value of future tasks. In the SOS-ET problem, before the online task arrives, its specific task information and arrival time are unknown to the satellite system. The satellite immediately coordinates, allocates, and plans to execute the task after it arrives, or abandons the task. However, due to the limitation of the number of tasks that the satellite can execute, the UOA algorithm only considers the benefits of the currently arriving task and ignores the arrival of future tasks and the benefits that these tasks may bring. In a satellite scenario where resources are limited and tasks arrive online, this strategy may cause low-profit tasks to consume a large amount of early resources, resulting in high-profit tasks in the later stage being unable to be executed due to resource constraints. Therefore, there is an urgent need for a future-beneficial distributed satellite online scheduling algorithm to avoid low-benefit tasks that arrive early in the SOS-ET problem from occupying valuable satellite imaging resources, so as to reserve resources for high-benefit tasks in the later stage.

[0011] Therefore, there is an urgent need for a method to solve the problem of establishing the SOS-ET model and, at the same time, to solve the problem of designing a future-beneficial distributed satellite online scheduling scheme for emergency task arrivals. Summary of the Invention

[0012] Objective of the Invention The objective of the present invention is to address the above technical problems and propose a future-beneficial distributed satellite online scheduling method for emergency task arrivals. After the ground station obtains the satellite set S and the online task set T, a satellite online scheduling model for emergency task arrivals is established. The specific steps are as follows:

[0013] S1. Any satellite k ∈ S in the satellite set S in the satellite cluster is represented as a binary tuple <n k , c k >:

[0014] Where, n k represents the maximum number of tasks that satellite k can execute in the current scheduling period; c k represents the payload resolution capability of satellite k, which is determined by the minimum resolution of the on-board payload. c k is normalized to a constant between 0 and 1. The larger the resolution capability value c k , the better the satellite imaging quality;

[0015] S2. The task attribute of any task i ∈ T in the line task set T is a five-tuple

[0016] Where, d i represents the task execution duration of task i; p irepresents the overall benefit value that can be obtained by executing task i, which is pre-defined by the user or the onboard system; the final execution benefit of the task will be determined by the overall benefit r of the task i and satellite payload resolution capability c k Joint decision making; represents the arrival time of task i, represents the expiration deadline of task i; T i Batch represents the set of tasks that arrive in the same batch as task i, W i Represents the set of observation visibility time windows of the mission under all satellite sets S;

[0017] S3. Set the offline version of the SOS-ET problem and determine the decision variables of the offline version of the SOS-ET problem;

[0018] S4. Configure on-board resource capacity constraints, visibility constraints, and consistency constraints;

[0019] S5. Establish a satellite online scheduling model for emergency mission arrival.

[0020] Furthermore, the observed visible time window set W i It is expressed as:

[0021]

[0022] Among them, w ijk Represents the observation visibility time window; it is defined as the jth continuous observable time period for any satellite k for any mission i, starting from the time window start time and the time window end time composition.

[0023] Furthermore, due to the limited onboard resources, including onboard power and onboard storage capacity, the maximum number of tasks that a satellite can perform in one scheduling cycle is n. k , which is the on-board resource capacity constraint, expressed as:

[0024]

[0025] Furthermore, the visibility constraint means that the task must be executed before the task deadline and within the visible time window with the satellite; the execution start time variable of task i is l i , then the task execution start time must be no less than its arrival time and the task time window start time The maximum value of the task execution end time must not be greater than the visible time window end time. and task deadlines

[0026] Furthermore, the consistency constraint means that each task in the satellite can only be executed once and will not be repeatedly observed by multiple satellites.

[0027] Furthermore, the model of the SOS-ET problem can be expressed as:

[0028]

[0029] Furthermore, the decision variables of the offline version of the SOS-ET problem include: the 0-1 decision variable x for task execution ijk , where the decision variable value of 1 indicates that task i is selected to be executed in the j-th time window on satellite k, otherwise it is 0;

[0030] Furthermore, the decision variables of the offline version of the SOS-ET problem include: the integer decision variable l for the start time of the task i ; representing the start execution time of task i.

[0031] Furthermore, the final execution revenue value r of the task ik is jointly determined by the predefined overall revenue value p of the task i and the payload resolution of the executed satellite c k , that is, r ik = p i ·c k .

[0032] Furthermore, the optimization objective of the SOS-ET problem is expressed as:

[0033] BRIEF DESCRIPTION OF THE DRAWINGS

[0034] In order to more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the following will briefly introduce the drawings required for the description of the specific embodiments or the prior art. Obviously, the following drawings are some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0035] Figure 1 is a schematic diagram of an online auction algorithm (UOA) considering maximizing the current efficiency;

[0036] Figure 2 is a flowchart of an online collaborative algorithm (OA-TE) based on the expected tender efficiency;

[0037] Figure 3 is a flowchart of an online collaborative (OA-BE) algorithm based on the expected tender efficiency;

[0038] Figure 4 It is a schematic diagram of the satellite online collaboration process for online arrival of tasks;

[0039] Figure 5 It is a schematic diagram of the maximum pitch angle parameter of the satellite;

[0040] Figure 6 It is the result of the algorithm comparison experiment under different numbers of tasks;

[0041] Figure 7 It is the result of the algorithm comparison experiment under different numbers of executable tasks of the satellite;

[0042] Figure 8 It is the result of the algorithm comparison under different maximum pitch angle parameters of the satellite;

[0043] Figure 9 It is the result of the algorithm comparison experiment when the task execution benefit is exponentially distributed;

[0044] Figure 10 It is the result of the algorithm comparison experiment when the task execution benefit is uniformly distributed;

[0045] Figure 11 It is the result of the algorithm comparison experiment when the task execution benefit is normally distributed. Detailed implementation manners

[0046] The following describes exemplary embodiments of the present application with reference to the accompanying drawings. Various details of the embodiments of the present application are included to facilitate understanding, and they should be considered merely exemplary. Therefore, those of ordinary skill in the art should recognize that various changes and modifications can be made to the embodiments described herein without departing from the scope and spirit of the present application. Similarly, descriptions of well-known functions and structures are omitted below for clarity and conciseness.

[0047] Aiming at the problems of the SOS-ET problem lacking a standardized mathematical model and a high-performance distributed online task planning algorithm, 1) a satellite online scheduling model for emergency task arrival is proposed, and 2) a future beneficial distributed satellite online scheduling algorithm for emergency task arrival is designed, including an online auction based on tendering utility expectation (OA-TE) and an online auction based on bidding utility expectation (OA-BE).

[0048] Establish a satellite online scheduling model for emergency task arrival:

[0049] The problem of Satellite Online Scheduling for Emergency Task Arrival (SOS-ET) studied in this invention involves a set of satellites in a satellite cluster and a set of a series of online tasks, aiming to maximize the cumulative value of task benefits under the satisfaction of various constraints such as logical constraints, capacity constraints, and consistency constraints. This invention mainly focuses on the scientific problems related to the SOS-ET problem and makes the following reasonable assumptions before problem modeling:

[0050] Satellites and their controlling parties have a certain predictive ability for the arrival pattern of tasks before the tasks arrive, and can switch appropriate collaborative algorithms according to the predicted arrival pattern. In addition, the benefit distribution of online tasks can be obtained through the analysis of the task execution data of the on-board system and the historical database of the ground satellite control system. This invention uses three common data distributions in online problems to replace the task benefit data distribution to achieve the theoretical analysis and derivation of the competitive ratio of online algorithms.

[0051] Tasks are divided into meta-tasks, which are the smallest observation units, to ensure that tasks can be completed in one observation and cannot be interrupted during the execution process. At the same time, the influence of external factors such as weather and sea conditions on imaging activities is not considered.

[0052] The on-board energy limit and storage capacity limit are comprehensively considered as the maximum number of tasks that a satellite can execute within a scheduling period. After the end of each scheduling period, the satellite has enough time to complete the clearing of solid storage and the charging of the payload to ensure the sustainable operation of the satellite system.

[0053] In the SOS-ET problem, a set of satellites S and a set of online tasks T are involved. First, a formal description of them will be given. For any satellite k ∈ S in the set of satellites S in the satellite cluster, it can be represented as the following binary tuple:

[0054] n k ,c k (1)

[0055] Among them, n k represents the maximum number of tasks that satellite k can execute within the current scheduling period; c k represents the payload resolution ability carried by satellite k, which is determined by the minimum resolution of the on-board payload. For the convenience of calculation, c k is normalized to a constant between 0 and 1. Generally speaking, the larger the resolution ability value c k , the better the imaging quality of the satellite and the higher the benefit of the finally executed task.

[0056] For any task i ∈ T in the set of online tasks T in the problem, its relevant online task (hereinafter referred to as "task") attributes can be represented as the following five-tuple:

[0057]

[0058] Among them, d i represents the task execution duration of task i, and p i represents the overall benefit value that can be obtained by executing task i, which is predefined by the user or the on-satellite system. Different application scenarios will adopt different ways to define the overall benefit value of tasks according to the characteristics of the scenarios. It should be noted that the final execution benefit of the task will be determined by the overall benefit r i of the task and the satellite payload resolution ability c k jointly; represents the arrival time of task i, represents the expiration deadline of task i; T i Batch represents the set of tasks that arrive in the same batch as task i, and W i represents the set of observation visible time windows of the task under all satellite sets S, that is

[0059]

[0060] Among them, w ijk represents the observation visible time window. The satellite platform has a relatively fixed flight orbit, which results in that the ground imaging task target is only visible during a period when the satellite flies over, and only within the visible time window can the satellite complete the task. Its specific definition is the j-th consecutive observable time period of any satellite k for any task i, which consists of the start time of the time window and the end time

[0061] This section first gives the offline version of the SOS-ET problem, aiming to construct a mixed integer programming model to capture the mutual influence between tasks and the finiteness of resources, and defines the SOS-ET mathematical model. For the offline version, it is assumed that the arrival order and arrival time of all tasks are predetermined and fully known, so as to provide an upper bound of the optimal solution for the online problem. The decision variables of the offline version of the SOS-ET problem include:

[0062] The 0-1 decision variable x ijk for task execution, whose value is 1 when task i is selected to be executed on the j-th time window of satellite k, otherwise 0;

[0063] The integer decision variable l i for the start time of the task: its value represents the start execution time of task i. In the actual engineering application of the satellite, both the attitude data and the final command plan are given in seconds, so the start time of the task defined in the present invention is also a discrete variable in seconds.

[0064] The optimization objective of the present invention is to maximize the sum of task cumulative benefits, and the final execution benefit value rik (hereinafter referred to as "revenue") is determined by the pre-defined overall revenue value p of the mission i and the payload resolution of the satellite being executed c k , that is, r ik = p i ·c k . In summary, the optimization objective of the SOS-ET problem can be expressed by the following formula:

[0065]

[0066] Based on optimizing the above objective function, the feasible task allocation and planning scheme in the SOS-ET problem also need to satisfy a series of constraint conditions, including on-board resource capacity constraints, task visibility constraints, logical constraints between tasks, attitude conversion time constraints, and consistency constraints.

[0067] On-board resource capacity constraints: Due to limited on-board resources, including on-board power and on-board storage capacity, jointly limit the maximum number of tasks n that the satellite can execute within a scheduling period k , that is

[0068]

[0069] Visibility constraints: It means that the task execution must be before the task deadline and within the visible time window of the satellite. The execution start time variable of task i is l i , then the execution start time of this task must not be less than the maximum value of its arrival time and the start time of the task time window , and its task execution end time must not be greater than the end time of the visible time window and the task deadline

[0070] Consistency constraints: Restrict that each task in the satellite can only be executed once and will not be repeatedly observed by multiple satellites:

[0071]

[0072] In summary, the mathematical model of the SOS-ET problem can be expressed as:

[0073]

[0074] In formulas (5)-(8), x ijk ∈{0,1}, l i ∈N,

[0075] Without loss of generality, the offline version of the SOS-ET problem can provide a theoretical upper bound for the optimal solution of the online version of the SOS-ET problem. This upper bound is obtained by statically analyzing all possible task sets and satellite resources, calculating the optimal allocation scheme, and thus providing a performance benchmark for the online problem. However, the core challenge of the SOS-ET problem lies in its inherent online nature, that is, tasks arrive dynamically in an unknown order and must be responded to immediately. In the space environment where satellites operate, the limitations of communication and storage resources make traditional centralized solution methods impractical. Due to the limited communication bandwidth of inter-satellite links and the strict limitations on the storage capacity of satellites, these factors jointly restrict the ability of a central decision-making satellite node to collect the status information of all satellites in real time and perform complex time window calculations and task planning. To overcome these limitations, a feasible solution is to adopt a distributed online scheduling algorithm to maximize the computing power of each satellite. At this time, each satellite acts as an autonomous participant and independently calculates the task plan related to itself. Subsequently, through a "question-and-answer" communication and interaction mechanism, each satellite makes online decisions on all task allocation and planning schemes while considering the balance between the benefits of individual rationality and the overall system benefits.

[0076] However, the distributed satellite online scheduling methods represented by the UOA algorithm focus on the current task benefits and do not consider the arrival of future tasks and the possible benefits they may bring. In the scenario of a distributed satellite system with limited resources and online task arrivals, such a strategy may lead to low-profit tasks consuming a large amount of early resources, resulting in high-profit tasks in the later stage being unable to be executed due to resource constraints, further reducing the overall utility value of the system. Therefore, to address the above limitations, this paper proposes a future-beneficial distributed online collaborative algorithm, including the distributed satellite online scheduling algorithm with beneficial tendering efficiency expectation (OA-TE) and the distributed satellite online scheduling algorithm with beneficial bidding efficiency expectation (OA-BE).

[0077] Future-beneficial distributed satellite online scheduling method for emergency task arrivals:

[0078] This section proposes a future-beneficial distributed satellite online scheduling algorithm for SOS-ET problems facing emergency task arrivals. As emergency tasks arrive sequentially with a high arrival frequency and tight deadlines, strict requirements are imposed on the solution efficiency of the algorithm. In addition, considering the limitation of the number of tasks that satellites can execute, when the algorithm performs online task allocation, it not only needs to consider the immediate benefits of the current task but also be forward-looking and comprehensively evaluate the potential value of future tasks. Therefore, this section first adopts a Utility-based Online Auction (UOA) algorithm that maximizes the current efficiency as the baseline algorithm, which allocates all currently arrived tasks based on the greedy criterion. Based on UOA, a task screening mechanism is set up respectively based on the expected efficiency of solicitation and the expected efficiency of bidding, and an online distributed satellite scheduling algorithm beneficial to the expected efficiency of solicitation (OA-TE) and an online distributed satellite scheduling algorithm beneficial to the expected efficiency of bidding (OA-BE) are designed to avoid early-arriving low-benefit tasks occupying valuable satellite imaging resources and thus reserve resources for high-benefit tasks in the later stage.

[0079] A. Online Distributed Satellite Scheduling Algorithm Beneficial to the Expected Efficiency of Solicitation:

[0080] In this section, a screening mechanism based on the expected efficiency of solicitation is introduced into the UOA algorithm process, and the improved algorithm is named the online distributed satellite scheduling algorithm beneficial to the expected efficiency of solicitation (OA-TE). The task screening mechanism of the OA-TE algorithm is activated when a task arrives at the soliciting satellite. Combining the benefits of the arriving task and the expected benefits of future soliciting tasks, the value of this task is evaluated and screened. The specific algorithm process of OA-TE is as Figure 2 shown. The soliciting satellite no longer allocates all online-arriving tasks without discrimination, but filters out low-value tasks through the task screening mechanism before the bidding process of UOA and rejects them.

[0081] Although existing research tends to use static thresholds for screening decisions of online tasks, as the resource state changes over time, static thresholds are difficult to adapt to dynamic changes, thus compromising the overall efficiency. To make up for this deficiency, OA-TE designs a dynamic threshold strategy for the expected efficiency of solicitation, which comprehensively considers the future task distribution information predicted based on historical data and the state information of the current satellite system. Its pseudocode is as follows:

[0082]

[0083] Algorithm 1 first calculates the average benefit value R of tasks according to the current time period Exp(Line 1 of the algorithm), combined with the task information that has arrived currently, evaluate the possible number of tasks that may arrive in the future and the average expected future revenue (Line 2 of the algorithm). As shown in Line 3 of the algorithm, when a task arrives online, if the task is to be rejected, two conditions need to be met simultaneously: 1) the revenue value of the current task is less than the expected revenue of future tasks; 2) the remaining number of task executions in the current entire satellite system is greater than the expected total number of future arriving tasks.

[0084] Compared with the UOA algorithm (baseline algorithm), the advantages and characteristics of the OA-TE algorithm lie in the following three aspects:

[0085] (1) Forward-looking task screening mechanism: The OA-TE algorithm evaluates the executable value and expected benefits of each task by introducing a task screening decision mechanism before task allocation. Different from the baseline algorithm, OA-TE does not accept all tasks without discrimination. Instead, it screens tasks based on the expected revenue of the tasks and their impact on the future resource utilization of the system. This method helps to avoid short-sighted behavior, that is, to avoid sacrificing greater benefits that may be obtained in the future for the sake of small current benefits, ensuring the efficient use of system resources, especially suitable for conditions with limited resources.

[0086] (2) Dynamic threshold setting: The OA-TE algorithm uses a dynamic threshold to determine whether to accept a task, rather than a static threshold. The dynamic threshold can be adjusted according to the system resource status and the predicted future task distribution, and can more accurately grasp the task acceptance ability and system resource status at each time period. This can respond more flexibly and precisely to the dynamic changes of tasks and system requirements, improving the adaptability of the algorithm and the operation efficiency of the overall system.

[0087] (3) Reducing communication costs: In the baseline algorithm, a comprehensive bidding process is carried out every time a task arrives, and each process requires multiple data exchanges between satellites. In the OA-TE algorithm, through the prior task screening mechanism, some tasks with lower revenues can be screened out before the collaborative process starts. This means that only tasks with an expected revenue higher than a certain dynamic threshold will trigger the inter-satellite bidding communication process, avoiding ineffective communication for tasks with lower revenues, thus directly reducing the number of communications and related costs.

[0088] B. A Distributed Satellite Online Scheduling Algorithm with Beneficial Expected Bidding Efficiency

[0089] This section presents an online collaborative algorithm beneficial to the expected efficiency of bidding (OA-BE), which is an improvement based on the UOA and OA-TE in the previous section. OA-TE makes a decision on whether to execute a task before bidding without considering the feasibility of the task on the satellite. This may lead to the rejection of some tasks with lower benefits but higher feasibility, while retaining some tasks with higher benefits but impossible to complete on the same satellite, thus wasting resources. Therefore, this section improves the OA-TE algorithm by adding a task screening mechanism to the bidding satellites when they receive the tender tasks. Compared with OA-TE, OA-BE considers the specific conditions of each satellite, making the task screening more accurate.

[0090]

[0091]

[0092] As Figure 3 shown, the task screening mechanism is integrated into the bidding satellites instead of the tendering satellites. The bidding satellites can conduct more accurate screening based on their own situations, filtering out some tasks with low execution value and giving up bidding. The pseudocode of the task screening mechanism based on the expected efficiency of bidding is shown in Algorithm 2: First, according to the total number of tasks of the bidding satellite at the current time and the average benefit value of the tasks (Line 1 of the algorithm), combined with the task information that has arrived currently, evaluate the number of tasks that may arrive in the future (Line 2 of the algorithm). As shown in Line 3 of the algorithm, when a task arrives online, if the task is to be rejected, two conditions need to be met simultaneously: 1) the benefit value of the current task is less than the average benefit value of the tasks of the bidding satellite; 2) the remaining number of tasks to be executed by the current bidding satellite is greater than the expected number of tasks to arrive in the future.

[0093] The analysis of the algorithm example is as follows:

[0094] From the perspective of the tendering satellite, the problem of satellite online scheduling (SOS-ET) for emergency task arrival can be equivalent to an online bipartite graph matching problem. Among them, tasks and satellites correspond to two disjoint vertex sets in the bipartite graph. If the corresponding satellite submits a bid for the tender task, there is an edge between the two vertices of the two disjoint sets, and its weight is the product of the task benefit and the satellite bid benefit value. Assigning online tasks to satellites is regarded as a matching in the bipartite graph, and the weight of the matching is the weight value of the edge in the bipartite graph. Figure 4Shows an example of the SOS-ET problem. Tasks i1, i2, i3, and i4 arrive at times t1, t2, t3, and t4 respectively. The newly arrived tasks receive bids delivered by satellites k1 and k2, and their satellite capacity values c1 = 0.8 and c2 = 0.6 are obtained through the bid evaluation method of the multi-attribute evaluation criterion. Mapping it to the online bipartite graph matching problem, the weight values r of the edges can be calculated respectively 1,1 = p1·c1 = 1.6, r 1,2 = 1.2. It is set that the remaining number of executable tasks for satellites k1 and k2 at this time is 1 each.

[0095] UOA algorithm

[0096] Based on Figure 4 the example, at time t1, the UOA algorithm assigns task i1 to satellite k1 when comparing the effectiveness values of satellites k1 and k2. At this time, the remaining number of executable tasks for satellite k1 is zero, and the remaining number of executable tasks for satellite k2 is 1. Similarly, the UOA algorithm assigns task t2 to satellite k2 at t2. When tasks i3 and i4 arrive, even though both satellites k1 and k2 have submitted bids, since the remaining number of executable tasks for both satellites k1 and k2 is zero, the allocation of tasks i3 and i4 cannot be realized. According to the above analysis, the task revenue value finally achieved by the UOA algorithm is r 1,1 + r 2,2 = 1.6 + 1.8 = 3.4.

[0097] OA-TE algorithm

[0098] Based on Figure 4 the example, it can be known that: after task i1 arrives at time t1, the OA-TE algorithm starts the task screening mechanism. First, the expected number of tasks and the expected revenue in this period are obtained as N Exp = 4 and R Exp = 4.5 respectively. Then, according to the second line of Algorithm 4.1, N Fut = 4 and R Fut = 4.5 are calculated respectively. Given that the revenue R of task i1 i1 = 2 and the remaining number of executable tasks N of the satellite system Ar = 2, it can be seen that the condition in the third line of Algorithm 4.1 is satisfied, and task i1 is rejected. Similarly, for task i2, N Fut = 3 and R Fut = (4 * 4.5 - 2) / 3 = 5.3, while the revenue of i2 is only 2. Therefore, task i2 is rejected. After task i3 arrives at time t3, the future number of tasks N Fut = 2 and the revenue R Fut=(4 * 4.5 - 5) / 2 = 6.5. At this time, the benefit value of task i3 is 8, and the remaining number of executable tasks is 2, which does not meet the task screening conditions. Then, the bidding process for task i3 is launched and it is matched to satellite k2. Similarly, for task i4, N Fut = 1 and R Fut =(4 * 4.5 - 13) / 1 = 5, while the benefit value of task i4 is 5. Therefore, the bidding process for task i4 is launched. However, at this time, the remaining number of executable tasks of satellite k2 is 0 and it cannot execute task i4, so the allocation of task i4 fails.

[0099] To sum up, after introducing the task screening mechanism, the OA-TE algorithm effectively screens tasks i1 and i2 with relatively low task benefits, avoiding their occupation of satellite resources, and successfully matches task i3 with relatively high task benefits to satellite k2. According to the above analysis, the final task benefit value achieved by the OA-TE algorithm is r 3,2 = p3·c2 = 5.6.

[0100] OA-BE algorithm

[0101] Based on Figure 4 the example, it can be seen that:

[0102] 1) After task i1 arrives at time t1, the task screening mechanisms of satellites k1 and k2 in the OA-BE algorithm are launched respectively. First, for satellite k1, its expected number of tasks N1 = 2, and the expected average benefit At this time, the remaining executable number of satellite k1 The benefit of task i1 is 2. Then, according to the task screening conditions in Algorithm 4.2, satellite s1 will reject task i1 and thus give up bidding; similarly, for satellite k2, its expected number of tasks Expected average benefit Then, according to the task screening conditions in Algorithm 5, satellite s2 also rejects task i1 and thus gives up bidding. At this time, task i1 is not allocated to any satellite.

[0103] 2) After task i2 arrives at time t2, for satellite k1, its future expected number of tasks N1 = 1, and the expected average benefit At this time, the remaining executable number of satellite k1 The benefit of task i1 is 3, which does not meet the task screening conditions. Satellite k1 will bid for task i2, and at this time, task i2 is matched to satellite k1.

[0104] 3) After task i3 arrives at time t3, according to the OA-BE algorithm, the future expected number of tasks of satellite k2 is obtained Expected average benefit At this time, the remaining executable number of satellite k2 The benefit of task i3 is 8, which does not meet the task screening criteria. Satellite k2 will bid on task i3, and at this time, the task will be matched to satellite k2.

[0105] 4) After task i4 arrives at time t4, the number of executable tasks of satellite k1 and satellite k2 is both 0 at this time, and task i4 allocation fails.

[0106] In summary, after introducing the task screening mechanism, the OA-BE algorithm effectively screens out task i1 with a relatively low task benefit, preventing it from occupying satellite resources. Moreover, it successfully retains task i2 that satellite k1 can execute and successfully allocates task i3 with a higher benefit, achieving precise screening of tasks. According to the above analysis, the task benefit value finally achieved by the OA-TE algorithm is r 2,1 +r 3,2 = 1.8 + 5.6 = 7.4.

[0107] Experimental settings

[0108] The code for this solution is written in Matlab 2022b, and all experiments are run on an i7-9700k CPU with 64GB RAM.

[0109] (1) Satellite simulation scenario design

[0110] In terms of satellite simulation scenario design, the intra-cluster cooperation scenario involved in this section includes 20 heterogeneous remote sensing satellites, which are interconnected through real-time inter-satellite links to achieve rapid exchange of task information and command information. These inter-satellite links can enable rapid transmission and reception of task information and command information between satellites, and ignore the delay interference caused by message sending and receiving. The planned scheduling period is set to 3 hours. Table 1 presents the basic parameters of the satellite system, including information such as the orbital altitude, orbital inclination, and right ascension of the ascending node of the satellites.

[0111] Table 1 Basic parameters of the satellite system

[0112]

[0113] (2) Design of experimental related parameters

[0114] In the present invention, the generation of imaging tasks follows a certain geographical distribution law, ensuring the rationality of the experimental design and the feasibility of practical applications. Specifically, without special instructions, the longitude and latitude coordinates of the tasks are randomly distributed in the area between 60°S and 60°N latitude. This setting aims to simulate the global observation requirements while considering the satellite's observation capabilities and coverage. According to the satellite imaging requirements, the imaging duration of a single task is set between 40 and 60 seconds, and the maximum conversion time between tasks for the satellite is 5 seconds. Considering the heterogeneity of the satellite system, the satellite constellation constructed in this paper includes remote sensing satellites with different imaging capabilities, equipped with optical payloads of different resolutions. To quantify the benefits of imaging tasks, the final effectiveness value of the imaging tasks is directly correlated with their imaging quality in this paper. Specifically, for satellites equipped with high-resolution imaging payloads, the satellite resolution ability value is set to 1; while for satellites equipped with ordinary imaging payloads, the satellite resolution ability value is set to 0.8. In the simulation scenario of this paper, 8 satellites are set to carry high-resolution optical imaging payloads, and the remaining satellites are all equipped with different imaging payloads. The experimental related parameter settings are shown in Table 2, and the visible relationship between satellites and tasks as well as the orbital relationship of satellites are generated using a dedicated satellite simulation software, where the default parameters are marked in bold.

[0115] Table 2 Basic Parameters of Experimental Design

[0116]

[0117] Among them, the total number of tasks reflects the total number of online arrival nodes in the experimental scenario. On the premise of a fixed number of satellites, different total numbers of tasks can reflect the scarcity of online resources. Similarly, restricting the maximum number of tasks that a satellite can execute can also reflect the resource scarcity, thereby further considering the performance of the online allocation algorithm under different resource-task ratios. As Figure 5 (a) shows, the satellite maximum pitch angle parameter affects the size of the satellite observation area: specifically, a task can only be observed if it is within the satellite's observation area. As Figure 5(As shown in (b), the tasks in the overlapping area of two satellites can be jointly observed by the two satellites. Therefore, the maximum pitch angle parameter of the satellite actually limits the observation range of the satellite. The larger the value of this parameter, the more tasks the satellite can execute, and at the same time, the more satellites can be selected for the tasks. The maximum pitch angle of the satellite actually limits the observation range of the satellite and affects the number of tasks that the satellite can execute and the number of satellites that can be selected for the tasks. However, the other two angle parameters of the satellite, the maximum roll angle and the maximum yaw angle, do not have such an impact. Therefore, these two parameters are directly set to 15° and 0°. In addition, in order to simulate the situation where the utility value of task-satellite matching follows a certain distribution, values of task benefits following normal distribution, uniform distribution, and exponential distribution are generated, and task benefit parameters with different values are generated to test the algorithm performance. The task arrival order follows a uniform distribution under the random order model. This setting aims to simulate the uncertainty of task arrival in practical applications, as well as the adaptability and stability of the online algorithm when dealing with random task flows.)

[0118] (3) Design of experimental comparison metrics

[0119] The experimental comparison metrics mainly include the following three, namely benefit, the CPU time consumed by the algorithm execution, and the communication cost required by the satellite online collaboration algorithm.)

[0120] Benefit refers to the cumulative benefit of the planned and executed tasks obtained from the algorithm scheduling result after the task allocation within the satellite cluster is completed. This metric reflects the performance of the algorithm in terms of solution quality, that is, the algorithm can achieve the objective function value of the offline OIAC-OT problem model.)

[0121] The running time of the algorithm is measured in seconds, which is the CPU time consumed from the start of the simulation to the end of the algorithm execution. This metric is directly related to the solution efficiency of the algorithm.)

[0122] The communication cost reflects the total number of information exchanges required by the satellites within the cluster during the algorithm execution. In the present invention, one communication between any two satellites is regarded as adding one to the communication cost. This metric is a unique measurement standard for the satellite online collaboration algorithm and has recently been widely used to evaluate the performance of such algorithms.)

[0123] The above algorithm revenue and algorithm running time, as common algorithm performance evaluation metrics, evaluate the algorithm from two dimensions: solution quality and solution efficiency. The communication cost, on the other hand, is a special measurement metric for satellite online collaborative algorithms and has recently been widely used to measure the performance of satellite online collaborative algorithms. Using the communication cost as a measurement metric mainly takes into account the drawbacks such as the impact of frequent communication on ① the life of the antenna payload, ② the on-board energy, and ③ the risk of packet loss. During the scheduling process, the number of communications should be minimized as much as possible. The online collaborative algorithm of the satellite cluster is required to save the communication link and the energy consumption caused by sending information as much as possible, so that the algorithm can minimize the communication cost while optimizing the problem itself.

[0124] Comparison results with other algorithms

[0125] Two online scheduling algorithms were used as baseline algorithms in the experiment, namely the online scheduling algorithm based on logarithmic random threshold (E-GRT) and the global online scheduling algorithm based on two phases (Two-phase-based Global Online Allocation, TGOA). E-GRT and TGOA achieved competitive ratios of 1 / (lnU + 1) and 1 / 4 respectively in the online matching problem of bipartite graphs, and they are one of the currently known most advanced bipartite graph online scheduling algorithms. Since the problems solved by the above two algorithms are different from the problems in this chapter, in order to compare under unified conditions in this chapter's experiment, the above methods were respectively extended and improved based on the reverse auction framework:

[0126] The basic idea of E-GRT is to randomly select a value related to the edge weight as the threshold for measuring whether to accept the current match as an alternative match. When E-GRT is applied to the problems in this chapter, its bid generation algorithm and bid evaluation algorithm are both generated by the algorithms proposed in this chapter. For newly arrived online tasks, the cluster head satellite will screen out the bids with bid effectiveness values greater than a given threshold from all eligible bids to form an optional bid set, and then randomly select the winning satellite from it.

[0127] The basic idea of TGOA is to first divide all nodes into two equal groups according to the arrival order and adopt different strategies for them. For the nodes arriving in the first half, TGOA adopts a greedy strategy, assigning each newly arrived node to the matching edge with the highest edge weight that satisfies all constraints; for the vertices in the second half, TGOA adopts an optimization strategy based on the Hungarian algorithm. Applying TGOA to the problems in this chapter, for the newly arrived tasks that arrive online, the cluster head satellite extracts the bid effectiveness values of all task bids at historical moments and the corresponding satellites to form the matching edge weights and matching edges, and uses the Hungarian algorithm to find the global optimal matching. If the currently newly arrived task is in this global optimal matching, the corresponding satellite in the global optimal matching is selected as the winning satellite; otherwise, the task is abandoned.

[0128] Figure 6 shows the comparison results of the proposed algorithm and the baseline algorithm under different numbers of tasks. In Figure 6 , when the number of tasks is less than 800, the revenue values of each algorithm will increase with the increase in the number of tasks. However, when the number of tasks exceeds 800, the growth trend of the revenue value begins to slow down. This is because the number of tasks that satellites can execute is limited, and the revenue obtained by satellites for executing tasks does not increase linearly with the increase in the number of system tasks. The proposed UOA, OA-TE, and OA-BE algorithms have achieved greater improvements than the baseline algorithms E-GRT and TGOA in all examples with different numbers of satellites. When the number of tasks is less than 400, the differences in the total revenue performance of UOA, OA-TE, and OA-BE are relatively small, with a difference amplitude within 5%. However, as the number of tasks increases, the performance differences between these algorithms gradually expand. Especially when the number of tasks exceeds 800, OA-BE has an average revenue increase of 17.55% and 43.86% compared to OA-TE and UOA, respectively.

[0129] In terms of the running time of the algorithms, the running time of all algorithms shows an upward trend with the increase in the number of tasks. Since the TGOA algorithm needs to run the Hungarian algorithm in each iteration, its running time is much higher than that of other algorithms and shows a linear increase with the increase in the number of satellites. The running times of the proposed algorithms and the baseline algorithm E-GRT can both be controlled within seconds. Among them, the running time of the baseline algorithm E-GRT is slightly higher than that of the proposed algorithms, and the UOA algorithm has the shortest running time and the highest efficiency, which is consistent with the results of its time complexity analysis. Finally, by observing Figure 6From the comparison results of the number of algorithm communications, it can be found that OA-TE and OA-BE achieve the lowest communication cost among all algorithms. This is mainly due to the task screening mechanism of OA-TE and OA-BE in the bidding stage, which effectively avoids unnecessary inter-satellite communication negotiations. Especially for the OA-TE algorithm, it evaluates the task before initiating the bidding, greatly saving the communication times consumed by publishing task information for bidding, and this part of the communication volume accounts for a relatively large proportion in all collaborative processes.

[0130] The experimental results of algorithm comparison in the scenario of example group 2 are as follows Figure 7 shown. When the number of tasks that the satellite can execute increases, the revenue value obtained by the algorithm also increases, because the satellite can execute more tasks, thereby improving the overall efficiency value. Under the conditions of different numbers of tasks that the satellite can execute, the proposed algorithm shows significant advantages over the baseline algorithm in terms of the total revenue value and the algorithm running time. Specifically, similar to the experimental results of other example groups, the average total revenue value of the proposed algorithm is at least 150% higher than that of the baseline algorithm.

[0131] It should be noted that the revenue gap between UOA and OA-TE, OA-BE decreases as the number of executable tasks increases. Until the number of tasks that the satellite can execute reaches 40, the performance of UOA is basically the same as that of OA-TE and OA-BE. This is because as the number of executable tasks increases, the advantages of the task screening mechanism of OA-TE and OA-BE gradually weaken. At this time, the number of tasks that the satellite can execute is relatively large and the satellite resources are sufficient. Therefore, the UOA algorithm without any task screening and filtering mechanism can achieve better performance.

[0132] The conclusion regarding the algorithm running time is the same as that of example group 1. The experimental results of algorithm comparison regarding the communication cost are as follows Figure 7 shown. It can be seen from the figure that the communication times of the baseline algorithm and the UOA algorithm are less affected by the change in the number of executable tasks, while OA-TE and OA-BE are more affected. Among them, when the number of tasks that the satellite can execute increases from 30 to 40, the communication times of the OA-TE algorithm increase sharply from 13358.0 to 17261.4, with an increase rate of approximately 30%. This change can also be attributed to the fact that as the number of tasks that the satellite can execute increases, the number of tasks that can be screened out by the task screening mechanism of the OA-TE algorithm gradually decreases, resulting in an increase in its communication times.

[0133] Figure 8 shows the performance results of the algorithm under different maximum pitch angle parameters of the satellite, that is, example group 3 in the corresponding experimental scenario design. As Figure 8As shown, the change in the algorithm's benefit value caused by the change in the maximum pitch angle parameter of the satellite is very slight, and the total benefit value fluctuations of all algorithms are within 5%. Under all satellite maximum pitch angle parameters, the proposed algorithms UOA, OA-TE, and OA-BE achieved better results than the baseline algorithms E-GRT and TGOA. In terms of algorithm efficiency and communication cost, the algorithm comparison conclusions in case group 3 are consistent with those obtained in case groups 1-2. It is worth noting that as the satellite maximum pitch angle parameter increases, the running time and communication times of all algorithms show a slight upward trend. The reason for this phenomenon is that the increase in the satellite maximum pitch angle parameter leads to an increase in the overlapping area of the imaging tasks between satellites, thus increasing the number of feasible bids in the original problem and resulting in an increase in the algorithm running time and communication cost. Specifically, when the satellite maximum pitch angle parameter increases, the observation range of the satellite expands, enabling more tasks to be observed simultaneously by multiple satellites. This not only increases the number of candidate solutions that the algorithm needs to consider during task allocation but also increases the frequency of inter-satellite communication negotiation to resolve potential conflicts in task allocation. Therefore, the running time and communication cost of the algorithm increase with the increase in the satellite maximum pitch angle parameter.

[0134] In Figure 9 、 Figure 10 and Figure 11 , the effects of different benefit expectations on the algorithm performance are respectively shown under the exponential distribution, uniform distribution, and normal distribution. In terms of the total benefit value, the OA-BE algorithm performs excellently under all benefit distributions and benefit expectation values, achieving the highest total benefit value, and the proposed algorithms as a whole are superior to the baseline algorithms.

[0135] From the vertical comparison of different benefit distributions, the OA-TE algorithm performs relatively prominently under the exponential distribution, and the average benefit gap with OA-BE remains at 5.13%. The UOA algorithm has the highest average benefit value of 3523.2 under the normal distribution, but its average benefit value under the exponential distribution is relatively low, at 3406.6. This indicates that the performance of the UOA algorithm fluctuates greatly under different benefit distributions, while the OA-BE algorithm is more stable. In terms of the algorithm running time, the running time of all algorithms shows a slight downward trend with the increase in the task expected benefit. Except for the baseline algorithm TGOA, the running time of all algorithms can be controlled within seconds. Among them, the UOA algorithm achieves the minimum running time under all parameter configurations, demonstrating its efficiency advantage. In terms of communication cost, the communication times of all algorithms do not change significantly with the change in the benefit expectation value. The OA-TE algorithm achieves the minimum communication cost under all benefit distributions and benefit expectation values. Especially under the exponential distribution, its average communication times are reduced by 17.11% and 26.09% compared with the uniform distribution and normal distribution respectively.

[0136] Therefore, from the perspective of algorithm selection for different revenue distributions, for the online task allocation problem with an exponential revenue distribution, it is recommended to choose the OA-TE algorithm. This is because while maintaining a high revenue level, the OA-TE algorithm can significantly reduce the communication cost. Compared with the OA-BE algorithm, the OA-TE algorithm reduces the communication cost by approximately 36.7%, and the revenue loss is controlled within about 5%.

[0137] After the above analysis, this section summarizes the main experimental conclusions of the future beneficial online allocation algorithm for frequent task arrivals as follows:

[0138] 1) Comprehensive performance of all algorithms: Generally speaking, OA-BE performs the best in terms of the quality of the final result and algorithm efficiency. In terms of the total efficiency obtained by the algorithm, OA-BE achieves the highest revenue value among all algorithms in most case groups and parameters. Taking case group 1 as an example, compared with the OA-TE algorithm and the UOA algorithm, OA-BE improves the revenue value by 17.55% and 43.86% respectively. In terms of the running time of the algorithm, even as the number of tasks and satellites increases, OA-BE still controls the running time within seconds for all cases.

[0139] 2) Communication cost efficiency of the OA-TE algorithm: The OA-TE algorithm mainly has a significant advantage in the number of communications required by the algorithm. In the default scenario, the OA-TE algorithm reduces the number of communications by an average of 30% compared with other algorithms. At the same time, its performance in the total revenue of the algorithm is second only to OA-BE, and the average revenue gap with OA-BE remains at about 10%. In addition, the OA-TE algorithm has obvious advantages under certain specific distributions. For example, Figure 4 under the exponential distribution shown, the OA-TE algorithm reduces the communication cost by approximately 36.7% compared with the OA-BE algorithm, but only loses about 5% of the revenue value.

[0140] 3) Correlation between algorithm performance and parameters: The total revenue value of the online task algorithm has an obvious positive correlation with the number of tasks, the number of satellites, the number of tasks that satellites can execute, and the expected task revenue. The maximum pitch angle of the satellite has little impact on the total revenue value of the algorithm. Except for TGOA, the running time of other algorithms only increases significantly when the number of satellites and tasks increases, shows a small increase when the maximum pitch angle of the satellite increases, and the impact of other parameters is small, which is also consistent with the analysis of the algorithm time complexity. The influencing factors of the algorithm communication cost are the same as those of the running time. Among them, the number of communications of the OA-TE algorithm is most affected by the number of tasks that satellites can execute.

[0141] The above specific embodiments do not constitute a limitation to the protection scope of this application. Those skilled in the art should understand that various modifications, combinations, sub-combinations and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions and improvements made within the spirit and principle of this application shall be included within the protection scope of this application.

Claims

1. A future beneficial distributed satellite online scheduling method for emergency mission arrival, characterized in that: After the ground station obtains the satellite set S and the online task set T, it establishes a satellite online scheduling model for emergency task arrival. The specific steps are as follows: S1. Any satellite k∈S in the satellite set S in the cluster is represented as a binary <n k ,c k 〉; Among them, n k represents the maximum number of tasks that satellite k can perform in the current scheduling cycle; c k represents the resolution capability of the payload carried by satellite k, which is determined by the minimum resolution of the onboard payload, c k Normalized to a constant between 0 and 1, the resolution capability value c k The larger it is, the better the satellite imaging quality is; S2. The task attribute of any task i∈T in the line task set T is a five-tuple Among them, d i represents the execution time of task i; p i represents the overall benefit value that can be obtained by executing task i, which is pre-defined by the user or the onboard system; the final execution benefit of the task will be determined by the overall benefit r of the task i and satellite payload resolution capability c k Joint decision making; represents the arrival time of task i, represents the expiration deadline of task i; T i Batch represents the set of tasks that arrive in the same batch as task i, W i Represents the set of observation visibility time windows of the mission under all satellite sets S; S3. Set the offline version of the SOS-ET problem and determine the decision variables of the offline version of the SOS-ET problem; S4. Configure on-board resource capacity constraints, visibility constraints, and consistency constraints; S5. Establish a satellite online scheduling model for emergency mission arrival.

2. The method for online scheduling of future beneficial distributed satellites for emergency mission arrival according to claim 1, characterized in that: The observation visible time window set W i It is expressed as: Among them, w ijk Represents the observation visibility time window; it is defined as the jth continuous observable time period for any satellite k for any mission i, starting from the time window start time and the time window end time composition.

3. The method for online scheduling of future beneficial distributed satellites for emergency mission arrival according to claim 2, characterized in that: Due to the limited on-board resources, including on-board power and on-board storage capacity, the maximum number of tasks that a satellite can perform in one scheduling cycle is n. k , which is the on-board resource capacity constraint, expressed as:

4. The method for online scheduling of future beneficial distributed satellites for emergency mission arrival according to claim 3, characterized in that: The visibility constraint means that the task must be executed before the task deadline and within the visible time window with the satellite; the execution start time variable of task i is l i , then the task execution start time must be no less than its arrival time and the task time window start time The maximum value of the task execution end time must not be greater than the visible time window end time. and task deadlines 5. The method for online scheduling of future beneficial distributed satellites for emergency mission arrival according to claim 4, characterized in that: The consistency constraint means that each task in the satellite can only be executed once and will not be observed repeatedly by multiple satellites.

6. The method for online scheduling of future beneficial distributed satellites for emergency mission arrival according to claim 5, characterized in that: The model of the SOS-ET problem can be expressed as:

7. The method for online scheduling of future beneficial distributed satellites for emergency mission arrival according to claim 6, characterized in that: The decision variables of the offline version of the SOS-ET problem include: 0-1 decision variables x for task execution ijk When the decision variable value is 1, it means that task i is executed in the jth time window on satellite k, otherwise it is 0.

8. The method for online scheduling of future beneficial distributed satellites for emergency mission arrival according to claim 7, characterized in that: The decision variables of the offline version of the SOS-ET problem include: an integer decision variable l for the task start time i ; represents the start execution time of task i.

9. The method for online scheduling of future beneficial distributed satellites for emergency mission arrival according to claim 8, characterized in that: The final execution benefit value of the task r ik The overall benefit value p defined in advance by the task i c is determined by the payload resolution of the satellite being executed. k , that is, r ik =p i ·c k .

10. The method for online scheduling of future beneficial distributed satellites for emergency mission arrival according to claim 9, characterized in that: The optimization objective of the SOS-ET problem is expressed as:

Images