CN108023637B - Isomorphic multi-satellite online collaboration method - Google Patents

Abstract

The invention discloses an isomorphic multi-satellite online collaboration method. In the isomorphic multi-satellite online collaboration method, a centralized-distributed collaboration architecture is adopted, a collaboration decision function is only configured on an earth stationary orbit communication satellite, and earth observation satellites are subjected to distributed computation, the isomorphic multi-satellite online collaboration method adopts a task collaborative allocation algorithm based on an auction mechanism, auctions are performed by earth stationary orbit communication satellite organizations serving as auctioneers, each earth observation satellite calculates a bid-out value corresponding to an emergency observation task according to local information of the earth observation satellite and bids according to the bid-out value, each earth observation satellite informs the auctioneer of the bid-out value, then the earth stationary orbit communication satellite determines a winner of each emergency observation task, and if a plurality of communication relay nodes are associated with each other, roles of the auctioneers can be transferred among the communication relay nodes.

Description

Isomorphic multi-satellite online collaboration method

Technical Field

The invention relates to the technical field of satellites, in particular to an isomorphic multi-satellite online cooperation method, and more particularly to an isomorphic multi-satellite online cooperation method adopting a centralized-distributed cooperation architecture.

Background

With the improvement of the on-satellite load detection capability and the image processing capability, the earth observation satellite can find a valuable observation target on the satellite and generate a further observation task request (generally, an emergency observation task with higher profit or priority level), and can also receive observation task requests (emergency observation task requests) transmitted by other satellite communications, wherein the observation task requests are dynamically and randomly arrived and have high timeliness requirements and need to be solved on line, and under the condition of multi-satellite (especially isomorphic multi-satellite), a multi-satellite on-line cooperative scheduling mechanism is needed for cooperative scheduling.

Disclosure of Invention

The invention aims to provide an isomorphic multi-satellite online cooperation method to realize isomorphic multi-satellite online cooperation, and particularly better deal with emergency observation tasks.

In order to achieve the above object, the present invention provides an isomorphic multi-satellite online collaboration method in which an earth observation satellite is used as a resource observation satellite, an earth geostationary orbit communication satellite is used as a communication relay node, the earth observation satellite and the earth geostationary orbit communication satellite communicate, and satellite communication cannot be directly performed between the earth observation satellites, the isomorphic multi-satellite online collaboration method employs a centralized-distributed collaboration architecture, a collaboration decision function is only configured on the earth geostationary orbit communication satellite, and distributed computation is performed on the earth observation satellite, the isomorphic multi-satellite online collaboration method employs a task collaborative allocation algorithm based on an auction mechanism, auctions are performed by an earth geostationary orbit communication satellite organization serving as an auctioneer, each earth observation satellite calculates a bid-out value corresponding to an emergency observation task according to its own local information, and bidding is carried out according to the bidding, each earth observation satellite informs an auctioneer of the bidding, then the earth stationary orbit communication satellite determines a winner of each emergency observation task, and if a plurality of communication relay nodes are associated with each other, the role of the auctioneer can be transferred among the communication relay nodes.

Preferably, the geostationary orbit communication satellite adopts the following bid winning decision mode:

1) if a plurality of earth observation satellites can complete the emergency observation task before the cut-off time, selecting the earth observation satellite which enables the total income increment of the system to be maximum as a winner; if the bidding value of the bidding price of a plurality of satellites corresponds to the highest total income increment, the earth observation satellite with the earliest actual observation starting time has the winning priority;

2) if only one earth observation satellite can complete the emergency observation task before the deadline and the increment of the total income of the system is a positive value, selecting the satellite as a winner-winning person;

3) the geostationary orbit communication satellite abandons the emergency observation mission if the system gross revenue increments generated when the geostationary observation satellite performs the emergency observation mission are both negative or no geostationary observation satellite can complete the emergency observation mission by the deadline time.

Preferably, the time availability w of the emergency observation task is calculated by the following equation_iojThe calculation is as follows:

wherein ols is_iojIs the latest observation start time of the jth emergency observation task in the ith batch, and ols_ioj＝max{olf_ioj-p_oj-(2·maxθ)/sp_roll,oes_ioj}。

Preferably, each earth observation satellite has six information vectors, which are as follows:

1.) the task bundle,

1≤n≤|BU_ioand the task bundle represents a task set which is selected from the o-th batch of emergency observation tasks and is successfully scheduled for the earth observation satellite i, and the tasks are sequenced according to the time of adding the tasks into the bundle. Length of current task bundle is | BU_ioL, and less than the number u of emergency observation tasks in a batch, | BU_ioU is less than or equal to | l; when task bundle is empty, use BU_ioPhi and BU_ioThe expression | ═ 0 means that,

2.) a corresponding planning sequence,

1≤n≤|PA_iothe length of the planning sequence is the same as the length of the task beam, | BU_io|＝|PA_io|≤u，

3.) the execution time vector is calculated,

1≤n≤|TI_ioi, the execution time vector represents the actual starting time of each task in the earth observation satellite i execution plan sequence, and the length of the vector is the same as that of the plan sequence,

4.) winner list WA_io@{wa_io1,...,wa_iouU, length, where wa_iojRepresenting the winner currently considered by the earth observation satellite i for the jth task in the ith emergency observation task, wherein the specific value corresponds to the number of the winner, and when wa_iojWhen phi, the earth observation satellite i considers that there is currently no winner for the task,

5.) bid-winning bid list WB_io@{wb_io1,...,wb_iouH length u, where wb_iojIndicates the corresponding winWhen the value of the bid given by the winner is 0, the task is not winner currently,

6.) timestamp vector

Length n₁Wherein ts is_ioi′And the timestamp indicating the latest information update of the earth observation satellite i for the o batch of emergency observation tasks, namely the time point when the updated information from the earth observation satellite i' is received.

Preferably, the task co-allocation algorithm comprises two stages of iterations: the method comprises a bundle construction phase and a consistency construction phase, wherein the former generates ordered task bundles in a greedy search mode corresponding to each earth observation satellite, the latter generates task distribution conflicts corresponding to the identification task and carries out conflict resolution through local communication between the adjacent earth observation satellites, and the two phases are repeated iteratively until convergence is achieved.

Preferably, the communication between the earth observation satellite and the earth stationary orbit communication satellite is intermittent, and the task cooperative allocation algorithm further comprises a new phase: a synchronous communication loop prediction phase, wherein a synchronous communication loop is a bidirectional communication link for two or more earth observation satellites (satellites) to share information vectors to and receive feedback information from other earth observation satellites from the loop, one synchronous communication loop is established on the basis of overlapping time between communication time windows of an initiator and one or more responders,

the description form is a multi-tuple<ID,I,R,ES,LF>_SCL：

-the ID is an identifier;

i is the communication initiator number, i.e. the first one shares its information;

-R is the number of the communication responder receiving and sharing the information;

ES is the earliest start time of the communication loop;

LF is the latest end time of the communication loop,

wherein the communication initiator I and the responder R are earth observation satellites, and ES and LF are determined by overlapping time intervals over communication time windows of the initiator and all responders, a synchronous communication loop prediction phase for predicting the presence of a synchronous communication loop, and a tuple of the synchronous communication loop is provided<ID,I,R,ES,LF>_SCLThe synchronous communication loop prediction phase is a precondition for the bundle construction phase.

Drawings

Fig. 1 is a schematic diagram of a synchronous communication loop.

Fig. 2 is a schematic diagram of two cases of an asynchronous communication loop.

Fig. 3 is a general conceptual diagram of online co-scheduling solution over the entire scheduling interval.

Fig. 4 is a schematic diagram of communication between a relay node and a geostationary satellite.

Fig. 5 is a schematic diagram of information flow in a contractual network agreement.

Fig. 6 shows the communication available between the space and ground segments.

FIG. 7 shows the m-CBBA algorithm results for the non-interconnected and interconnected cases between GEOs.

Fig. 8 shows the results of the algorithms with spatial distribution of tasks and different rolling scheduling periods.

Detailed Description

Embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

The satellite (earth observation satellite LEO) in the present invention is an autonomous agent (autonomous satellite) and has the following characteristics: autonomous perception, autonomous planning scheduling and autonomous execution.

The autonomous embodying aspects comprise the aspects of precise attitude determination, orbit determination, attitude sensor/driver and load calibration, attitude control, orbit maneuver, state monitoring and predictive analysis, fault detection, diagnosis, isolation and repair, forecasting modeling (in planning scheduling, state monitoring analysis, orbit determination and maintenance and the like, the support of forecasting models is required, such as information of ephemeris, solar intensity and the like), task planning and scheduling, load control and configuration, data storage and communication, image data processing and the like. For example, the conforming emergency task can be autonomously generated by awareness.

The distributed earth observation satellite system has unique orbital characteristics and intermittent communication characteristics, communication constraints are more, and communication connectivity in the system cannot be guaranteed in real time. Therefore, in a dynamic task environment, facing limited on-satellite computing power and time-varying communication constraints, how to design an efficient and applicable multi-satellite online cooperative scheduling mechanism and algorithm is a very challenging problem, and a balance is obtained between system efficiency and communication cost.

A distributed earth observation satellite system is a collection of satellites located in different orbits and is tasked with performing observation imaging of various hot spot areas on the earth's surface, such as hot spots corresponding to volcanic activity or forest fires. Each earth observation satellite is provided with a single imaging load, and the central view field line of the camera can swing left and right. The roll activity requires a duration limit, so that a task request to observe too close an activity cannot be fulfilled by the same satellite. In addition, the satellite is equipped with anomaly detection equipment, and can discover hot events and generate observation task requests on the satellite.

The online arrival of observation task requests comes primarily from three main sources: tasks generated by satellites in the constellation, tasks generated and transmitted by other satellites in the constellation and emergency upper notes of ground stations.

Earth observation satellites are in low orbit and inter-satellite links are not always available. For example, RapidEye is a commercial multispectral earth observation satellite project from RapidEye AG of blandenburg, germany, comprising a satellite constellation consisting of five small satellites. Five RapidEye geostationary satellites are deployed on the same sun-synchronized orbit at a height of 630 km and are placed equidistantly to ensure consistent imaging conditions and short revisit intervals. The satellites are spaced apart from each other in their orbital plane by approximately 19 minutes. It can be seen that there is no inter-satellite direct communication link between the five satellites, because the number and altitude of the satellites does not allow inter-satellite links to exist, so we can only operate and coordinate these satellites offline in the ground station. The present invention provides more communication opportunities than before, given that there are a limited number of communication relay nodes. The relay nodes can be located on the ground, such as ground communication stations or mobile communication vehicles, and can also be communication satellites deployed in space, and the like.

In the problem of on-line scheduling of satellite constellations, there are two types of tasks, the conventional observation task and the emergency observation task. And the ground control center receives and arranges the observation requests of the users and uploads the generated observation plans through the satellite-ground link. In the present invention, these observation requests as collected in advance are referred to as regular observation tasks. In contrast, observation requests that arrive randomly from satellites within and outside the constellation are referred to as emergency observation tasks, and unforeseen emergency observation tasks require online scheduling without ground intervention. The value of the return for emergency observation tasks is generally higher than for conventional observation tasks, since such tasks have high value and require a quick response. For the more general case, emergency observation tasks often arrive randomly in batches.

From the perspective of the wider multi-earth observation satellite system cooperation, the earth observation satellite can only image one target on the ground at a time. Secondly, the tasks are directed to different observation targets and are independent of each other, and each task can be completed by one satellite. Third, the satellite scheduling problem is an oversubscription problem where the number of mission requests is significantly greater than the number of earth observation satellites available, thus building a mission plan for each satellite.

For uniform and simple expression, characters used in the following of the invention are uniformly defined:

subscript

i, i', i ″, low orbit leo (low Earth orbit) satellite number, i ═ 1,2, K, n₁

g, communication relay node number, h 1,2, K, n₂

j, j' emergency observation task number, j is 1,2, K, u

k is the conventional observation task number, k is 1,2_i

q is communication time window number, q is 1,2, K, m_ig

o, o', o ″, task lot number, o ═ 1,2, K, l

Amount of ginseng

H, whole scheduling interval

tl is the total duration of the whole scheduling interval H

n₁Total number of earth observation satellites in the system

n₂Number of available communication relay nodes

u number of emergency observation tasks in a batch of tasks

l total number of emergency observation task batches in whole dispatching interval H

v_iNumber of regular tasks uploaded on Earth Observation satellite i

r_oTime for relay node to issue o batch of emergency observation tasks to earth observation satellite

tr_ioThe time of the o-th batch of emergency observation tasks reaches the time of the earth observation satellite i

sp_rollSide-swinging angular velocity of earth observation satellite

max theta maximum yaw angle of earth observation satellite

oes_iojAiming at the earliest observation starting time of the jth emergency observation task in the ith batch of the geostationary observation satellite i

olf_iojThe latest observation end time of the jth emergency observation task in the ith batch of the geostationary observation satellite i

ols_iojThe latest observation start time of the jth emergency observation task in the ith batch of the geostationary observation satellite i

otw_iojObservation time window for j-th emergency observation task in o-th batch of earth observation satellites i

p_ojImaging duration of jth emergency observation task in ith batch

e_ojThe yield of the jth emergency observation task in the ith batch is given by a manager or an on-board decision

θ_iojAiming at the j emergency observation task in the o batch of the earth observation satellite iAngle of sidesway

b_iojActual start time of jth emergency observation task in the o-th batch for earth observation satellite i

c_iojActual end time of jth emergency observation task in the ith batch for geostationary observation satellite i

oes_ikThe earliest starting time of the kth conventional task on the earth observation satellite i

olf_ikThe latest ending time of the kth conventional task on the earth observation satellite i

p_ikImaging duration of kth conventional task on earth observation satellite i

e_ikThe profit for the k-th conventional mission on earth observation satellite i is given by the manager

b_ikActual start time of kth conventional task on earth observation satellite i

c_ikActual end time of kth conventional task on earth observation satellite i

s_iojkOn earth observation satellite i, the order-dependent transition time when the kth regular task is executed next to the jth emergency observation task in the o lot

s_ikojOn earth observation satellite i, the order-dependent switching time when the jth emergency observation task in the o-th batch is executed next to the kth conventional task

s_iojo′j′On the geostationary satellite i, the sequence of the j-th emergency observation task in the o ' -th batch when executed immediately after the j-th emergency observation task in the o-th batch depends on the switching time, wherein (o ≠ o ') | (j ≠ j ') 1

ctw_igqQ communication time window between earth observation satellite i and relay node g

m_igNumber of communication time windows between earth observation satellite i and relay node g

ut_iojRepresentation form of j-th emergency observation task in o-th batch aiming at earth observation satellite i

w_iojWhen the jth batch in the o batch is neededThe hard-watch task may be equal to 1 when scheduled to execute on earth observation satellite i, and equal to 0 otherwise.

The communication time window set between the communication relay node and all earth observation satellites in the system is defined as follows:

specifically, a set of communication time windows between the communication relay node g and the earth observation satellite i is defined as follows:

wherein a communication time window ctw_igqIs defined as:

ctw_igq＝[ces_igq,clf_igq]

ces therein_igqIs the earliest communication start time, clf_igqIs the latest communication end time.

The o-th batch of emergency observation tasks may be issued from the relay node to the earth observation satellite i if and only if one or more communication time windows exist between the earth observation satellite i and the relay node. Time availability to batch tasks w_ioThe calculation is as follows:

if and only if the o-th batch of emergency observation tasks starts at the latest observation start time ols_iojThe observation satellite i arrives before, and the jth emergency observation task in the ith batch can be scheduled to be executed on the observation satellite i. Time availability w for urgent observation tasks_iojThe calculation is as follows:

wherein ols_ioj＝max{olf_ioj-p_oj-(2·maxθ)/sp_roll,oes_ioj}。

The whole system-oriented assumption according to the practical application is as follows:

1. the emergency observation tasks from different sources are collected at the communication relay node and are issued by the communication relay node;

2. the load on the satellite is the optical imaging load, and the influence of the ground shadow needs to be considered;

3. the communication delay time between the earth observation satellite and the relay node is short and can be ignored;

4. on-board dead space and sustained operating time are sufficient throughout the scheduling interval.

When the o "th batch of emergency observation tasks reaches the satellite constellation, the online co-scheduling problem can be represented by a mixed integer linear programming model milp (mixed integer linear programming):

decision variables

c_ioj+(s_iojk+p_ik)y_iojk+olf_ioj(y_iojk-1)≤c_ik(4.2)

c_ik+(s_ikoj+p_oj)y_ikoj+olf_ik(y_ikoj-1)≤c_ioj(4.3)

c_io'j'+(s_io'j'oj+p_ioj)y_io'j'oj+olf_io'j'(y_io'j'oj-1)≤c_ioj(4.4)

(tr_io+p_oj)x_ioj+s_ikojy_ikoj+s_io'j'ojy_io'j'oj≤c_ioj(4.5)

p_ikz_ik+s_iojky_iojk≤c_ik(4.6)

(oes_ioj+p_oj)x_ioj≤c_ioj(4.7)

(oes_ik+p_ik)z_ik≤c_ik(4.8)

c_ioj≤olf_iojx_ioj(4.9)

c_ik≤olf_ikz_ik(4.10)

b_ioj+p_oj＝c_ioj(4.11)

b_ik+p_ik＝c_ik(4.12)

x_ioj＝{0,1},y_iojk＝{0,1},y_ikoj＝{0,1},y_io'j'oj＝{0,1},z_ik＝{0,1},w_ioj＝{0,1} (4.15)

In the centralized-distributed collaborative architecture, the collaborative decision function is only configured on the central node, while other nodes can perform distributed computation, and the corresponding method is a task collaborative allocation method based on a market mechanism, including an auction mechanism. The auction is conducted by the auctioneer organization, wherein each earth observation satellite calculates a bid price value corresponding to the task according to its local information and makes a bid accordingly. Each earth observation satellite informs the auctioneer of its bid and the auctioneer then decides the winner of each task. The algorithmic mechanism may ensure that a conflict-free solution is generated because the auctioneer selects only one earth observation satellite as the winner. Since the algorithmic mechanism is oriented to a centralized-distributed architecture, bid calculations are performed in a distributed fashion.

In a multi-satellite cooperative application scene, the communication relay node plays the role of an auctioneer and is responsible for bidding on emergency observation tasks, evaluating bids of each bidding satellite and feeding back bid results. If a plurality of communication relay nodes are associated with each other, the role of the auctioneer can be transferred among the communication relay nodes to improve the reliability of the system. As a specific observation resource earth observation satellite, satellite communication cannot be directly performed between earth observation satellites, and a communication link between an auctioneer and the resource earth observation satellite is intermittent.

Aiming at a bid-winning decision stage in the algorithm mechanism, the following decision rules are proposed:

1. and if a plurality of earth observation satellites can complete the emergency observation task before the deadline, selecting the earth observation satellite with the largest increment of the total system income as the winner. In this case, if there are a plurality of satellites whose bid values correspond to the highest total revenue increment, the earliest earth observation satellite actually starts to have a winning priority.

2. If only one earth observation satellite can complete the emergency observation task before the deadline and the increment of the total income of the system is positive, the satellite is selected as the successful winner.

3. If the total revenue increment of the system generated when the earth observation satellite executes the emergency observation task is negative, or no earth observation satellite can complete the emergency observation task before the expiration date, the auction dealer abandons the emergency observation task.

In fig. 5, relay node 1 acts as an auctioneer, relay node 2 acts as a communication relay, and LEO _3 wins the bid. The relay node 1 performs auction distribution, bid collection and bid feedback through available communication links.

The pseudo code of the online collaborative algorithm based on the contract network agreement under the centralized-distributed architecture is as follows:

inputting:

LEO satellite set

Relay node set

Set of emergency observation tasks Ut ═ Ut¹,ut²,...,ut^(u·l)}

Current set of dispatch plans

Communication time window set CTW

The total number u of emergency observation tasks in each batch

And (3) outputting:

updated set of dispatch plans

Aiming at the realization of a specific contract network agreement, the invention provides two modes: the bidding under the single task and the bidding under the batch task are respectively carried out, wherein the former refers to the bidding of only a single task and the determination of a winner of the task in each iteration, and the latter refers to the bidding of batch tasks and the determination of winners of a plurality of tasks in each iteration.

Single-task contract network agreement Algorithm SingleItem-CNP-based on Online organizing scheduling Algorithm

The bubble sequencing computation complexity of each batch of emergency observation tasks is O (u)²) The dichotomy is adopted to carry out time availability check on a given batch task and a given earth observation satellite, and the corresponding computational complexity is

For a given earth observation satellite, the online scheduling computation complexity of each emergency observation task is

The global computation complexity of the contract network agreement algorithm under the single task is

Contract network agreement Algorithm under Batch task, digest-CNP-based online order Scheduling Algorithm 2

Function BatchCNP(Lsat,Gsat,

Sc^～t,r_o,u,o)

Bid phase under%% batch mission

Selection and winning stage under% batch task

The global computation complexity of the contract network agreement algorithm under the batch task is as same as that under the single task

Facing to the distributed architecture, the distributed algorithm, particularly the auction algorithm and the consistency protocol are adopted to carry out conflict-free distribution on tasks, and the method is suitable for the unreliable condition of the communication environment which is difficult to realize global consistency situation perception.

Improved consistency bundle algorithm based on synchronous communication

Aiming at the complex combination optimization problem of multi-satellite online cooperative task scheduling under communication constraint and oriented to a distributed architecture, the invention provides an improved consistency beam algorithm m-CBBA algorithm based on synchronous communication.

The algorithm for synchronous communication is adopted to have strict rule definition on when communication is carried out, meanwhile, the synchronous strategy can queue the communication situation and can be executed only under the triggering of a specific event, and the algorithm is ensured to have an expected state in the program execution process. In most iterative algorithms, a synchronization algorithm is employed. In the synchronous iterative algorithm, each module executes parallel computation, shares state variables, and then waits for the next trigger event to perform the next iterative process of the algorithm. In this process, the state of each earth observation satellite is guaranteed.

The cost of using the synchronization algorithm is that synchronization behavior needs to be maintained, and particularly in distributed and decentralized algorithms, if synchronization cooperative behavior needs to be realized, simultaneous communication of cooperative parties needs to be visible, and cooperative synchronization can be performed.

The CBBA is a distributed auction algorithm, complex constraints are processed in a real-time distributed mode, a task allocation scheme is provided for multi-task allocation problems of multi-pair earth observation satellites, and the CBBA is very efficient in practical application.

The consistency bundle algorithm has a plurality of characteristics and can be used for realizing the cooperative task scheduling of the heterogeneous system, and N is assumed_tFor the total number of tasks to be allocated, L_tFor the maximum number of tasks per bundle, N_aD is the number of the earth observation satellites in the earth observation satellite network, and the specific details are as follows:

(1) the coherent bundle algorithm is distributed architecture oriented;

(2) the coherent bundle algorithm belongs to a polynomial time algorithm, and the computational complexity of the bundle construction stage is O (N)_tL_t)；

(3) The coherent bundle algorithm will be at max N_t,L_tN_aD iterations are internally converged, so that the consistent beam algorithm can well adapt to the changes of the network scale and the task number;

(4) different solving targets, earth observation satellite internal models and related constraints can be embodied by designing appropriate scoring functions. If the scoring function meets the marginal profit attenuation characteristic, a good feasible solution can be guaranteed.

For a distributed multi-satellite system, to execute a coherent beam algorithm, each earth observation satellite needs to have six information vectors, which is as follows:

the task is to be bundled,

1≤n≤|BU_iol. The task bundle represents a task set which is selected from the o-th emergency observation task and is successfully scheduled for the earth observation satellite i, and the task set is sequenced according to the time of adding each task into the bundle. Length of current task bundle is | BU_ioI, andless than the number u, | BU of emergency observation tasks in a batch_ioU is less than or equal to | l; when task bundle is empty, use BU_ioPhi and BU_ioAnd | ═ 0.

The corresponding planning sequence is set to be in accordance with the plan,

1≤n≤|PA_iol. The tasks in the planning sequence are the same as the task bundle and are used to indicate the specific order in which each task in the task bundle is performed on earth observation satellite i. The length of the planned sequence is the same as the length of the task bundle, | BU_io|＝|PA_io|≤u。

The execution of the time vector is performed,

1≤n≤|TI_iol. The execution time vector represents the actual start time for executing each task in the planning sequence to the earth observation satellite i, and the length of the vector is the same as that of the planning sequence.

Winner list WA_io@{wa_io1,...,wa_iouU, length, where wa_iojAnd the number of the winner which is currently considered by the earth observation satellite i and aims at the jth task in the ith batch of emergency observation tasks is represented, and the specific value corresponds to the number of the winner. When wa_iojWhen phi, earth observation satellite i considers that it is currently not winner for the task.

Bid and bid amount listing WB_io@{wb_io1,...,wb_iouH length u, where wb_iojAnd the bids given by the corresponding winners are represented, and when the value is 0, the task is represented as the current winners.

Timestamp vector

Length n₁Wherein ts is_ioi′And the timestamp indicating the latest information update of the earth observation satellite i for the o batch of emergency observation tasks, namely the time point when the updated information from the earth observation satellite i' is received.

The consistent beam algorithm CBBA consists of two stages of iteration: a bundle construction phase and a consistency construction phase, wherein the former generates ordered task bundles in a greedy search mode corresponding to each earth observation satellite, and the latter corresponds to identifying task allocation conflicts and carrying out conflict resolution through local communication with adjacent earth observation satellites s. These two phases are repeated iteratively until convergence is reached.

On the basis of a consistency beam algorithm CBBA, a new stage is added before the original two stages by improving the consistency beam algorithm m-CBBA, and a communication loop prediction stage is synchronized.

Because of the strict communication constraint in practical application, strong connectivity between the earth observation satellites s is impossible to achieve, so the communication loop prediction can balance between improving the global performance of the system and ensuring conflict-free scheduling of emergency observation tasks, rather than realize global convergence on a winner list.

Stage 1: synchronous communication loop prediction phase

The synchronous communication loop prediction stage is used to predict the existence of a synchronous communication loop. A synchronous communication loop is a two-way communication link for two or more earth observation satellites (satellites) to share information vectors to and receive feedback information from other earth observation satellites from the loop. A synchronous communication loop is constructed based on the existence of overlapping times between communication time windows of an initiator and one or more responders. The synchronous communication loop being described as a tuple<ID,I,R,ES,LF>_SCL：

-the ID is an identifier;

i is the communication initiator number, i.e. the first one shares its information;

-R is the number of the communication responder receiving and sharing the information;

ES is the earliest start time of the communication loop;

LF is the latest end time of the communication loop.

Where communication initiator I and responder R are both earth observation satellites and ES and LF are determined by overlapping time intervals over the communication time windows of the initiator and all responders.

Each synchronous sharing feedback interaction is formed by triggering on an available synchronous communication loop, wherein the sharing uplink timestamp is the same as the sharing downlink timestamp, and the feedback uplink timestamp is the same as the feedback downlink timestamp. Corresponding description form<id,i,i′,t_US-DS,t_UF-DF>From synchronous communication loops<ID,I,R,ES,LF>_SCLWherein I ∈ I, I' ∈ R, t_US-DS,t_UF-DF∈[ES,LF]，t_US-DS<t_UF-DF。

For each earth observation satellite, when a batch of emergency observation tasks arrives, it is necessary to predict the available synchronous communication loop that itself acts as the initiator. Furthermore, when a set of shared information vectors arrives, it is necessary to record the synchronous communication loop that itself is the responder. Through both of the above scenarios, each satellite will establish and update a synchronous communication loop from its own perspective. For each satellite, all synchronous communication loops are updated and ordered and executed in order of earliest start time.

To achieve local convergence, whenever a task bundle constructed by a loop initiator or responder changes, a synchronous shared feedback interaction is triggered on the loop while generating a timestamp of the last information update, within the time interval in which the loop is available.

One shared feedback interaction in the synchronous communication loop is considered as one communication, and at most un exists in the online cooperative scheduling of the m-CBBA algorithm₁(n₁-1) a communication. In each shared feedback interaction, the following two phases are performed.

And (2) stage: beam construction

In contrast to enumerating all possible task bundles, in an m-CBBA, each earth observation satellite constructs a respective task bundle and updates as the scheduling process progresses. In this phase of the algorithm, each earth observation satellite continually adds tasks to its task bundle in a sequential greedy manner until no more tasks can be added. Tasks in the task bundle are arranged according to the sequence of sequential addition, and tasks in the execution plan are ordered according to their corresponding actual start times.

The bundle construction process is based on a single-star oriented online scheduling algorithm. For each available task that is not currently in the task bundle, the earth observation satellite compares its score value to the task score in the current bid-winning vector, and if greater, retains the new task bid-winning value. For the unscheduled task set, the earth observation satellite selects the task with the highest score and adds it to the task bundle, and updates the task bundle, the execution plan, the timestamp, the winner and the landmark vector to include the newly added task.

The recursive process of bundle construction continues until the task bundle has reached its capacity limit, or no more tasks can be added to the task bundle, i.e., the satellite can no longer perform the remaining tasks over the other satellites.

And the single-star online scheduling algorithm schedules the single-star task at the LEO. And the single-star online scheduling algorithm schedules the conventional task of the star and the emergency observation task distributed to the star by the cooperative distribution algorithm. The single-star online scheduling algorithm comprises the following steps:

(1) at the T-driven scheduling time point, generating a new task plan in the next period time interval by adopting a complete rescheduling strategy in a progressive method, wherein the T-driven scheduling time point determines a specific scheduling time point lT according to a given time interval T, L is more than or equal to 0 and less than or equal to L, LT is more than or equal to H < (L +1) T, and when the scheduling time point lT is reached, calculating and generating the task plan of the next scheduling interval [ lT, (L +1) T ], wherein L is a positive integer, T is the given time interval, L is the maximum T-driven scheduling times, H is the total scheduling interval, and

(2) at C^*-driving rescheduling time points, using a schedule repair strategy in a modified manner, if at a certain time t (0) when the satellite is operating within a given scheduling interval<t<H) On-board emergency observation task cumulant C_tExceeding a given threshold C^*Then a rescheduling calculation is performed, wherein the threshold value C^*For emergency observation tasksThe critical cumulative number of the phase-change material,

except for the two scheduling time points, no scheduling is performed at any other time point.

More specifically, in the single-star online scheduling algorithm, a specific scheduling algorithm at a T-driven scheduling time point is as follows:

inputting:

-a set of emergency observation tasks that have arrived and that have not been scheduled before the T-driven scheduling time point;

-a set of regular observation tasks that have been received and that have not been scheduled before a T-driven scheduling time point;

and (3) outputting:

-a scheduling plan for the next time period T;

the method comprises the following specific steps:

step 11 is respectively from

And

selecting the conventional observation task and the emergency observation task whether the time window falls into the next time period T, and generating a conventional observation task set to be scheduled and solved

And set of emergency observation tasks

Step 12 will

And

integrating into an observation task set;

step 13, according to a set heuristic rule, sequencing the tasks in the integrated observation task set;

step 14, scheduling the tasks in the integrated observation task set one by one according to the sequence to determine whether to add the tasks into the observation task set

Until no more tasks in the integrated observation task set can be added

In (1),

step 15 outputs the scheduling plan in the next time period T

At C^*The scheduling algorithm for the driven rescheduling time points is as follows:

inputting:

within the present time period T and later than C^*-a scheduling plan driving a scheduling time point t;

-a set of emergency observation tasks that have arrived before the scheduling time point t and that are not scheduled;

and (3) outputting:

-a revised dispatch plan at time t,

the method comprises the following specific steps:

step 21 is to collect the slave tasks according to the condition that the observation time window is in the time interval from the time T to the next T-drive scheduling time point

Selecting emergency observation task to generate new task set

Step 22, according to the set heuristic rule, for

Sequencing the emergency observation tasks in the step (2);

step 23, according to the new task sequence, selecting one by one

In emergency observation task pair

Revising until

No emergency observation task can be added

In (1),

step 24 outputs the revised dispatch plan

And (3) stage: consistency construction

Once the earth observation satellite s has built its own mission beam, it needs to communicate with other earth observation satellites to resolve the allocation conflict. After receiving winners and corresponding winning bid information from neighboring geo-observation satellites s, each geo-observation satellite can determine whether any of its tasks in the task bundle have been won by other geo-observation satellites and winning the winning bid.

In the consistency construction phase, each pair of adjacent geostationary satellites s share the following information vectors simultaneously: winner list WA_ioBid-winning bid list WB_ioAnd a timestamp vector TS_ioAnd corresponds to a timestamp indicating that the latest information update was received by the other earth observation satellite s.

For each message passing between the sender i and the responder i ', the earth observation satellite i' performs a series of operations according to the received information, so as to update the information vector of the earth observation satellite. These specific operations include WA of their own information_i′o，WB_i′oAnd TS_i′oAnd compared to the information corresponding to the earth observation satellite i to determine which is the most up-to-date information for each task. The earth observation satellite i' may take three possible actions for each task j:

1. update of wa_i′oj＝wa_ioj,wb_i′oj＝wb_ioj；

2. Reset wa_i′oj＝φ,wb_i′oj＝0；

3. Abandon wa_i′oj＝wa_i′oj,wb_i′oj＝wb_i′oj；.

In response to the message sent from earth observation satellite i, the operational decision rule adopted for earth observation satellite i' for emergency observation task oj is as shown in table 4.1.

TABLE 1 operation decision rule when receiving message under synchronous communication

If during the communication a winner list WA is available_ioOr bid-winning bid list WB_ioIf there is a change, the earth observation satellite checks whether the updated or reset task is in its own task bundle. If so, thenThese tasks and their successors in the task bundle need to be cleared and released, then

The first winning task position sequence number in the task bundle, and then all the subsequent tasks in the task bundle

And the task footer thereof, will be updated as follows:

the task bundle is thus cut to remove the subsequent tasks:

at the same time, the corresponding task entry is also removed from the execution plan and timestamp vector.

If the subsequent scheduled tasks are not released, the result is too conservative, which leads to the reduction of the algorithm income, and the subsequent scheduled tasks are released and the bundle construction is carried out again, which is determined by the time relevance before and after the task scheduling. At this point, the algorithm re-enters the second phase, adding a new task to the task bundle. The m-CBBA algorithm iterates in the last two stages until the information variables are no longer queued. In practical applications, the number of synchronous sharing feedback on the same synchronous communication loop is limited by the cost of communication cost.

The computational complexity of the synchronous communication loop prediction is

Given a specific earth observation satellite, the computational complexity of the online scheduling algorithm for a batch of emergency observation tasks is

The overall computational complexity of the m-CBBA algorithmIs composed of

Improved asynchronous consistency bundle algorithm based on asynchronous communication

Asynchronous communication is applied to the situation where the modules in the algorithm operate independently of each other. Therefore, asynchronous communication is widely used in decentralized algorithms, i.e. it is possible to utilize the available information at any time, and not according to strict scheduling of nodes.

In the m-CBBA algorithm, the beam construction phase can be performed independently and asynchronously, since only local information is required for the earth observation satellite s at this phase. However, the adoption of the synchronous communication strategy in the consistency construction stage of the m-CBBA algorithm influences the response capability and performance of the whole system, because the conflict resolution strategy requires that each earth observation satellite know the latest information of each adjacent earth observation satellite in each communication iteration. It can be seen that this is a method of global consistency, in which each earth observation satellite shares its current state, waits to receive feedback messages from all neighboring earth observation satellites s (or waits for a certain time), and then enters the next algorithm stage.

For a real-time system, such a forced algorithm delay is not practical and affects performance and convergence rate of the algorithm, so that the m-ACBBA algorithm is provided in this section and adopts an asynchronous communication loop and an asynchronous conflict resolution rule.

Compared with the m-CBBA algorithm, the m-ACBBA not only treats the relay node as a communication node, but also adopts the relay node to play the role of a coordinator. Thus, the difference between m-CBBA and m-ACBBA is that m-ACBBA uses asynchronous communication loop prediction and asynchronous local decision rules.

Stage 1: asynchronous communication loop prediction

When the earth observation satellite schedules a new batch of emergency observation tasks, the information vector of the earth observation satellite is immediately shared to the relay node, and the relay node serves as a coordinator and is responsible for predicting the asynchronous communication loop.

Asynchronous sub-communication loop (ASCLs, Asynchronous)sub-communication loops) are determined by the scheduling period of the single satellite itself, the deadline of the emergency observation task and the communication time window between the relay node and the earth observation satellite. The specific description form is a multi-element group<ID,I,R,ES,LF>_ASCLThe method is the same as the description form of the synchronous communication loop, but the difference is that I or R is a relay node, and when the earth observation satellite is determined to be I, R is set as the relay node. Otherwise, when the I is the relay node, the R is set as the earth observation satellite. As can be seen, communication time window ctw_igqEquivalent to asynchronous sub-communication loop<id,i,g,ces_igq,clf_igq>Or<id,g,i,ces_igq,clf_igq>。

Asynchronous Communication Loop (ACL)<ID,I,R,ES_US,LF_US,ES_DS-UF,LF_DS-UF,ES_DF,LF_DF>_ACLIs composed of two or three independent asynchronous sub-communication loops.

FIG. 5(a) shows two asynchronous sub-communication loops<ID₁,I₁,R₁,ES₁,LF₁>_ASCLAnd<ID₂,I₂,R₂,ES₂,LF₂>_ASCLthe above-described asynchronous communication loop is formed. In the two asynchronous sub-communication loops, five correlation relationships exist, specifically as follows:

(1) both I and R are earth observation satellites. ACL initiator I is the initiator I of the previous ASCL₁While ACL responder R is responder R of the next ASCL₂。

I＝I₁

R＝R₂

(2) Responder R of the previous ASCL₁And initiator I of the latter ASCL₂Are all communication relay nodes. When the communication relay nodes are connected with each other, the relay nodes corresponding to the two ASCLs may be different. Otherwise, the corresponding relay nodes are the same.

(3)ES_USAnd LF_USRepresenting a shared upstream communication time window of the ACL.

ES_US＝ES₁

LF_US＝min{LF₁,LF₂}

ES_US≤LF_US

(4)ES_DS-UFAnd LF_DS-UFA shared downlink-feedback uplink communication time window representing the ACL.

ES_DS-UF＝max{ES₁,ES₂}

LF_DS-UF＝LF₂

ES_DS-UF≤LF_DS-UF

(5)ES_DFAnd LF_DFA feedback downlink communication time window representing the ACL.

ES_DF＝max{ES₁,ES₂}

LF_DF＝LF₁

ES_DF≤LF_DF

FIG. 5(b) shows three asynchronous sub-communication loops ASCLs<ID₁,I₁,R₁,ES₁,LF₁>_ASCL，<ID₂,I₂,R₂,ES₂,LF₂>_ASCLAnd<ID₃,I₃,R₃,ES₃,LF₃>_ASCLan asynchronous communication loop ACL is formed in the order of the earliest start time. Likewise, there are five correlations among the three ASCLs, as follows:

(1) both I and R are earth observation satellites. The initiator I of the ACL is the initiator I of the first ASCL₁While the responder R is the responder R of the last ASCL₃。

I＝I₁

R＝R₃

(2) Initiator I of the first ASCL₁Responder R with the last ASCL₃And the same to form a complete communication loop.

I₁＝R₃

(3) Responder R of first ASCL₁And initiator I of the last two ASCL₂And I₃Are all communication relay nodes. When the relay nodes are communicated with each other, the corresponding relay nodes can be different, otherwise, the relay nodes are the same relay node.

(4)ES_USAnd LF_USRepresenting a shared upstream communication time window of the ACL.

ES_US＝ES₁

LF_US＝min{LF₁,LF₂}

ES_US≤LF_US

(5)ES_DS-UFAnd LF_DS-UFA shared downlink-feedback uplink communication time window representing the ACL.

ES_DS-UF＝max{ES₁,ES₂}

LF_DS-UF＝LF₂

ES_DS-UF≤LF_DS-UF

(6)ES_DFAnd LF_DFA feedback downlink communication time window representing the ACL.

ES_DF＝max{ES₂,ES₃}

LF_DF＝LF₃

ES_DF≤LF_DF

The asynchronous shared feedback interaction is triggered to be generated on an available asynchronous communication loop. In the interaction, there are four timestamps including sharing uplink, sharing downlink, feedback uplink and feedback downlink. Description of the form<id,i,i′,t_US,t_DS,t_UF,t_DF>From asynchronous communication loops<ID,I,R,ES_US,LF_US,ES_DS-UF,LF_DS-UF,ES_DF,LF_DF>_ACLWherein I ∈ I, I' ∈ R, t_US∈[ES_US,LF_US]，t_DS,t_UF∈[ES_DS-UF,LF_DS-UF]，t_DF∈[ES_DF,LF_DF]，t_US≤t_DS，t_DS<t_UF，t_UF≤t_DF。

For each earth observation satellite, an asynchronous shared feedback interaction is performed on the loop when the task bundle of the initiator or responder changes during the available time interval of the ACL. Because the time occupied by communication delay and scheduling calculation is relatively short, the time stamps of the shared downlink and the feedback uplink are equal, t_DS≈t_UF。

The one asynchronous sharing feedback interaction on the ACL includes three communications in total, namely sharing uplink, sharing downlink-feedback uplink and feedback downlink. For one batch of emergency observation tasks, performing online cooperative scheduling by adopting an m-ACBBA algorithm, wherein 3un is needed at most₁(n₁-1) a communication.

In each asynchronous shared feedback interaction, both bundle construction and consistency construction are performed. Compared with the m-CBBA algorithm, the difference lies in that the m-ACBBA algorithm adopts an asynchronous local decision rule in the consistency construction stage.

Messages arrive out of order in an asynchronous communication protocol, e.g., messages generated at an earlier time may arrive later than messages generated at a later time. Therefore, the timestamp update mechanism based on the message reception time for the earth observation satellite in the m-CBBA algorithm is not suitable for the asynchronous case. Then the m-ACBBA algorithm is configured such that the actual time of winning the earth observation satellite s, TS, is indicated by a timestamp on the message_io@{ts_io1,...,ts_iouU, length, where ts_iojIndicating the time of generation of this winning bid on earth observation satellite i for emergency observation task j in lot o. Rather than the time of update of the geostationary satellite s message as used in the m-CBBA algorithm, i.e.

Length n₁Wherein ts is_ioi′Representing the location of the emergency observation task of the o-th batchThe observation satellite i makes a timestamp of the latest information update.

In the m-ACBBA algorithm, a series of local resolution rules are employed without requiring global state information. Meanwhile, the resolution protocol can carry out the consistency processing on the disordered messages and the redundant information.

The conflict resolution rules in the macba algorithm are not just the recipient updates the winner list and bid price list, but determines the messages that need to be shared again.

The reason for sharing again is that: firstly, the communication burden in the network is reduced, and the redundant information is prevented from being shared again; the second is to deal with the confusion of time stamps in asynchronous systems.

The recipient may take the following five possible actions:

1. update & re-share: the receiver updates the winner list, the bid-winning list and the winning time vector according to the information from the sender, and then shares the update;

2. discard & re-share: the receiver does not change the information state of the receiver, and then shares the winner information determined by the receiver;

3. discard & no more share: the receiver does not change the information state of the receiver and does not share the information, and the method is applied to the condition that the received information is consistent with the existing information.

4. Reset & re-share: the receiver clears the winner and the winner value, and then shares the originally received message again, so as to eliminate confusion;

5. update time & re-share: the recipient is the winner and when the message is received, the winning generation timestamp is updated to the current time to confirm that the recipient still believes to be the winner.

In response to the message sent from earth observation satellite i, the operational decision rule employed by earth observation satellite i' for emergency observation task oj is as shown in table 2.

TABLE 2 local decision rules when receiving messages in asynchronous communication

The computational complexity of the prediction of the asynchronous communication loop is

For a given earth observation satellite, the computational complexity of online dispatch for a batch of emergency observation tasks is

The overall computational complexity of the m-ACBBA algorithm is then

The objective of the earth observation task is to monitor and find, detail and track forest fire or volcanic eruption activity. More specifically, forest fires and volcanic eruptions require automatic monitoring for discovery, localization and identification. In the case of emergency events, the system is focused on improving the quick response capability of the system and improving the overall benefit of the system through online cooperation.

The geostationary orbit (GEO) communication satellite is used as a communication relay node, the situation of communication shortage is relieved to a certain extent, but the communication condition is still limited, and the constraint is still harsh. First, the communication time window between the GEO and the LEO is much larger than the communication time window between the ground station and the earth observation satellite; secondly, the GEO is connected with ground stations in the coverage range of the GEO all the time, so that the real-time availability of task requirement uploading and data downloading is ensured, and therefore, multi-satellite online collaboration becomes a reality in the communication mode, and disaster management on a higher level can be supported. Fig. 6 shows the communication available between the space segment and the ground segment, in particular as follows,

(1) communications between the earth observation satellite and the geostationary orbit communication satellite are available when the LEO passes through the coverage area of the GEO. The communication link is used for realizing bidirectional transmission of emergency observation tasks between the LEO and the ground station through the GEO, and downloading low-speed data from the LEO to the Ground Station (GS).

(2) The communication interaction between the GEO and the GS is possible from time to time because the GS is located in the coverage area of the GEO. The communication link is used for emergency observation task uploading and low-rate data downloading.

(3) Two-way communication interaction between the GEO is possible, enabling the transfer of task-related information between GEO.

(4) Direct communication between earth observation satellites is not possible.

In an application scene, 3 geostationary orbit communication satellites are arranged on the same orbit, and 3 geostationary orbit communication satellites are arranged above the equator, wherein specific satellite parameters are shown in a table 4.3. The time length of the simulation was set to 6 h.

TABLE 3 orbital parameter settings for three geostationary satellites and three geostationary orbiting satellites

The configuration scenario is characterized as follows:

1. when three earth observation satellites in the same orbit perform observation imaging on the same target on the ground, the time interval between LEO _ a and LEO _ B is about 12 minutes, and the time interval between LEO _ B and LEO _ C is also 12 minutes.

2. The communication coverage area of a geostationary orbital communication satellite is a conical area with a cone angle of 7.5 °.

3. The observation imaging area of the earth observation satellite is a rectangle. If the same target is observed and imaged on the same rail, the change value of the imaging angle between LEO _ a and LEO _ B is about 20 °, and the change value of the imaging angle between LEO _ B and LEO _ C is also about 20 °.

The communication conditions of the current distributed satellite system are analyzed, including the communication conditions between high and low orbit satellites facing a centralized-distributed architecture, as shown in table 4, and the communication conditions between low orbits facing a decentralized architecture, as shown in table 5.

TABLE 4 communication constraints between high and low orbit satellites in current distributed satellite systems

TABLE 5 communication constraints between Earth observing satellites in Current distributed satellite systems

Four online cooperative scheduling algorithms facing the distributed satellite system are compared, and the four online cooperative scheduling algorithms comprise a contract network agreement algorithm SI-CNP under a single task, a contract network agreement algorithm BA-CNP under a batch task, an improved consistency bundle algorithm m-CBBA and an improved asynchronous consistency bundle algorithm m-ACBBA.

In the m-CBBA and m-ACBBA algorithms, sharing uplink and sharing downlink of each sharing feedback interaction are respectively generated randomly in a time interval between the earliest starting time and the latest ending time corresponding to a communication loop. On each full communication loop, at most one shared feedback interaction is performed.

Three performance indexes are adopted, namely total income, the proportion of the scheduled successful emergency observation tasks and the corresponding total communication times. The three indexes are specifically described as follows:

(1) the total system gain refers to the sum of all scheduled successful task gains of the distributed satellite system in the whole scheduling interval.

(2) The proportion of the successfully scheduled emergency observation tasks refers to the percentage of the successfully scheduled emergency observation tasks in the distributed satellite system in the emergency observation tasks entering the scheduling solution.

(3) The total number of communications. The total number of communication is the sum of the number of communication times of three unidirectional communication situations, namely LEO initiates communication to GEO, GEO initiates communication to LEO and unidirectional communication between GEOs.

In order to evaluate the performance of four online cooperative scheduling algorithms, we focus on the responsiveness of a distributed satellite system to the arrival of an emergency observation task, so that two parameters related to the task, the arrival rate and the camera pointing angle of the emergency observation task, one parameter related to a scheduling mechanism, and a rolling scheduling period are very important for the generation of a test example. It can be seen that the two task-related parameter quantities represent the time distribution characteristic and the space distribution characteristic of the emergency observation task, respectively, and the rolling scheduling period determines the scale of the online scheduling calculation. Furthermore, all other parameter quantities are generated by a given uniform distribution. The parameter settings associated with the satellite, the task and the scheduling mechanism are shown in table 6.

TABLE 6 relevant parameter settings

FIG. 7 shows the results of the m-CBBA algorithm for the case of no interconnection and interconnection between geostationary orbit communication satellites respectively under different rolling scheduling periods. The network of interconnected geostationary-orbit communication satellites facilitates a higher overall throughput and a higher percentage of emergency observation tasks than would be possible if the geostationary-orbit communication satellites were not interconnected. Because interworking causes an increase in the number of communication time windows or an extension of the time intervals compared to non-interworking. Meanwhile, interconnection and interworking require more communication, especially communication interaction between GEO.

Table 7 shows that as the emergency observation task arrival rate increases, the total revenue achieved by each algorithm also increases. Under the condition that the arrival rates of the emergency observation tasks are the same, the wider the spatial distribution range of the emergency observation tasks is, the lower the total income is obtained, and the lower the proportion of the emergency observation tasks which are successfully scheduled is.

When the rolling scheduling period T is 6min, the m-ACBBA and the m-CBBA both obtain higher total income than the SI-CNP and the BA-CNP when the spatial and temporal distribution of the emergency observation task is the same, and the total income of the m-ACBBA is generally higher than that of the m-CBBA. And when the rolling scheduling period T is 12min, the m-ACBBA obtains the highest total benefit in the four algorithms, and the benefit obtained by the SI-CNP algorithm is higher than that obtained by the BA-CNP and m-CBBA algorithms. It can be seen that, for the BA-CNP and m-CBBA algorithms, the longer the rolling scheduling period is, the lower the total profit is, because the BA-CNP algorithm schedules the batch tasks, and each scheduling period corresponds to one emergency observation task batch, and at the same time, the longer the rolling scheduling period is, the fewer the number of sharing feedback times is, since the m-CBBA algorithm performs sharing feedback interaction at most once on the synchronous communication loop in each rolling scheduling period, the more the rolling scheduling period is, thereby affecting the timely response scheduling of the emergency observation tasks.

For the SI-CNP and m-ACBBA algorithms, the longer the rolling scheduling period is, the higher the total profit is, especially in the case that the number of communication times of the m-CBBA algorithm is reduced, the total profit is still increased, and the asynchronous communication strategy is excellent.

TABLE 7 Total revenue obtained by the algorithms under different parameters

Table 8 shows that m-ACBBA has the highest proportion of emergency observation tasks successfully scheduled in the four algorithms under the same parameter quantity. Sequencing the algorithms according to the mode that the proportion of the successfully scheduled emergency observation tasks is from high to low on the whole, and sequentially: m-ACBBA > m-CBBA > BA-CNP > SI-CNP.

TABLE 8 Emergency Observation task scheduling success ratio (%) obtained by each algorithm under different parameter conditions

Table 9 shows that as the arrival rate of the emergency observation task increases, the number of communications increases, while the SI-CNP traffic increases the most because it co-schedules the individual tasks in turn. In the four algorithms, the communication times of the m-CBBA are the minimum, and the communication times of the SI-CNP are the maximum, because the construction conditions of the synchronous communication loop are strict, the number of the loops is small, and the communication interaction times are further influenced. As the rolling scheduling period grows, the traffic corresponding to the SI-CNP is greatly increased, and the traffic corresponding to the other three algorithms is reduced.

TABLE 9 number of communications for each algorithm under different parameters

Table 10 shows that all three algorithms have improved overall yield over SI-CNP, with the greatest magnitude of gain increase for m-ACBBA. Similarly, Table 11 shows that the three algorithms have improved the proportion of successful emergency observation task scheduling compared with the SI-CNP algorithm, wherein the improvement of m-ACBBA is the largest.

Table 12 shows that as the arrival rate of emergency observation tasks increases, the average calculation time increases. Under the same parameter condition, the calculation time required by the four algorithms is almost the same and is matched with the actual calculation capacity on the satellite.

TABLE 10 gain (%), for each algorithm, compared to SI-CNP, when T is 6min (%)

TABLE 11 Emergency Observation task scheduling success increment (%), for each algorithm, compared to SI-CNP when T is 6 min%

TABLE 12 average calculation time (/ s) for scheduling in last scheduling period of one earth observation satellite

In summary, when the communication cost in the system is high, the m-CBBA algorithm can balance the total system benefit and the number of times of communication, and when the communication cost of the system is low, the m-ACBBA algorithm is the best choice for obtaining the ratio of the total system benefit to the success rate of scheduling the high emergency observation task.

The invention researches the problem of on-line cooperative scheduling of an isomorphic distributed earth observation satellite system to a simple task under strict communication constraint. The multi-satellite cooperative scheduling problem oriented to simple tasks belongs to the ID [ ST-SR-TA ] problem. Because a plurality of tasks can be allocated and scheduled on a single satellite, and due to time window constraints and resource (electricity, solid) limitations, single-satellite scheduling for a certain task is affected by other tasks, and there is a dependency relationship inside the scheduling plan. Firstly, defining batch availability and time effectiveness of emergency observation tasks aiming at a communication time window and an observation time window, and constructing a subproblem MILP model when each batch of emergency observation tasks arrives; then, based on a single-satellite online scheduling mechanism proposed by chapter three, two online collaborative task scheduling algorithms are respectively proposed for a centralized-distributed architecture and a distributed architecture. Specifically, the two algorithms facing the centralized-distributed architecture are a contract network agreement algorithm SI-CNP under a single task and a contract network agreement algorithm BA-CNP under a batch task; two algorithms facing decentralized architecture are the improved coherence bundle algorithm m-CBBA and the improved asynchronous coherence bundle algorithm m-ACBBA, which are based on synchronous and asynchronous communication, respectively. The calculation experiment result shows that compared with the situation that the communication is not interconnected, the interconnection and the intercommunication of the relay nodes can increase and expand the communication time window, and the increase of the ratio of the total system income to the success of the emergency observation task scheduling is facilitated. When the communication cost in the system is high, the m-CBBA algorithm dominates in balancing profit and the number of communications required. When the communication cost in the system is high, the m-CBBA algorithm can balance the total income of the system and the communication times, and when the communication cost of the system is low, the m-ACBBA algorithm is the best choice for obtaining the ratio of the total income of the system to the success rate of the scheduling of the emergency observation task.

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 (6)

1. An isomorphic multi-satellite online cooperation method is characterized in that an earth observation satellite is used as a resource observation satellite, an earth geostationary orbit communication satellite is used as a communication relay node, the earth observation satellite is communicated with the earth geostationary orbit communication satellite, satellite communication can not be directly carried out between the earth observation satellites, the isomorphic multi-satellite online cooperation method adopts a centralized-distributed cooperation architecture, a cooperation decision function is only configured on the earth geostationary orbit communication satellite, distributed calculation is carried out on the earth observation satellite, the isomorphic multi-satellite online cooperation method adopts a task cooperation distribution algorithm based on an auction mechanism, auctions are carried out by an earth geostationary orbit communication satellite organization used as an auctioneer, each earth observation satellite calculates a bid-out value corresponding to an emergency observation task according to local information of the earth observation satellite and carries out bidding according to the bid-distributed calculation value, each earth observation satellite informs an auctioneer of its bid, and then the earth stationary orbit communication satellite decides the winner of each emergency observation task, and if a plurality of communication relay nodes are associated with each other, the role of the auctioneer can be transferred between the communication relay nodes,

the task co-allocation algorithm comprises two stages of iteration: a bundle construction phase and a consistency construction phase, wherein the former generates ordered task bundles in a greedy search mode corresponding to each earth observation satellite, the latter identifies task allocation conflicts and carries out conflict resolution through local communication between the adjacent earth observation satellites, the two phases are repeated iteratively until convergence is achieved,

the communication between the earth observation satellite and the earth stationary orbit communication satellite is intermittent, and the task cooperative allocation algorithm further comprisesComprises a synchronous communication loop prediction phase, wherein the synchronous communication loop is a bidirectional communication link used for two or more earth observation satellites to share information vectors to other earth observation satellites and receive feedback information from other earth observation satellites, a synchronous communication loop is established on the basis of overlapping time between communication time windows of an initiator and one or more responders, and the description form of the synchronous communication loop is a multi-group<ID,I,R,ES,LF>_SCL：

-the ID is an identifier;

-I is the communication initiator number;

-R is the number of the communication responder receiving and sharing the information;

ES is the earliest start time of the communication loop;

LF is the latest end time of the communication loop,

wherein communication initiator I and responder R are both earth observation satellites, and ES and LF are determined by overlapping time intervals over the communication time windows of the initiator and all responders,

synchronous communication loop prediction phase for predicting the presence of a synchronous communication loop and providing a tuple of the synchronous communication loop<ID,I,R,ES,LF>_SCLThe synchronous communication loop prediction phase is a precondition for the bundle construction phase.

2. The homogeneous multi-satellite online collaboration method as claimed in claim 1 wherein said geostationary orbiting communication satellite employs the following bid winning decision-making approach:

1) if a plurality of earth observation satellites can complete the emergency observation task before the cut-off time, selecting the earth observation satellite which enables the total income increment of the system to be maximum as a winner; if the bidding value of the bidding price of a plurality of satellites corresponds to the highest total income increment, the earth observation satellite with the earliest actual observation starting time has the winning priority;

2) if only one earth observation satellite can complete the emergency observation task before the deadline and the increment of the total income of the system is a positive value, selecting the satellite as a winner-winning person;

3) the geostationary orbit communication satellite abandons the emergency observation mission if the system gross revenue increments generated when the geostationary observation satellite performs the emergency observation mission are both negative or no geostationary observation satellite can complete the emergency observation mission by the deadline time.

3. The homogeneous multi-satellite online collaboration method as claimed in claim 1 wherein the time availability w of the emergency observation task is calculated by the following equation_iojThe calculation is as follows:

wherein ols is_iojIs the latest observation start time of the jth emergency observation task in the ith batch, and ols_ioj＝max{olf_ioj-p_oj-(2·maxθ)/sp_roll,oes_ioj}，

Wherein the content of the first and second substances,

olf_iojfor earth observation satellite i, the latest observation end time of the jth emergency observation task in the ith lot,

p_ojfor the imaging duration of the jth emergency observation task in the ith batch,

max theta is the maximum yaw angle of the earth observation satellite,

sp_rollto observe the yaw angular velocity of the satellite over the ground,

oes_iojfor the earliest observation start time of the jth emergency observation task in the ith batch of the geostationary observation satellite i

tr_ioThe time of the o-th batch of emergency observation tasks reaches the time of the earth observation satellite i

w_ioTime availability w for the No. o batch of Emergency Observation tasks_io。

4. The isomorphic multi-satellite online collaboration method as defined in claim 1 wherein each earth observation satellite has six information vectors, specifically as follows:

1.) the task bundle,

1≤n≤|BU_iothe task bundle represents a task set which is selected from the o-th batch of emergency observation tasks and is successfully scheduled for the earth observation satellite i, and meanwhile, the tasks are sequenced according to the time of adding each task into the bundle, and the length of the current task bundle is | BU_ioL, and less than the number u of emergency observation tasks in a batch, | BU_ioU is less than or equal to | l; when task bundle is empty, use BU_ioPhi and BU_ioThe expression | ═ 0 means that,

2.) a corresponding planning sequence,

1≤n≤|PA_iothe length of the planning sequence is the same as the length of the task beam, | BU_io|＝|PA_io|≤u，

3.) the execution time vector is calculated,

1≤n≤|TI_ioi, the execution time vector represents the actual starting time of each task in the earth observation satellite i execution plan sequence, and the length of the vector is the same as that of the plan sequence,

4.) winner list WA_io@{wa_io1,...,wa_iouU, length, where wa_iojRepresenting the winner currently considered by the earth observation satellite i for the jth task in the ith emergency observation task, wherein the specific value corresponds to the number of the winner, and when wa_iojWhen phi, the earth observation satellite i considers that there is currently no winner for the task,

5.) bid-winning bid list WB_io@{wb_io1,...,wb_iouH length u, where wb_iojIndicating the bid offered by the corresponding winner when the corresponding winner is takenWhen the value is 0, the task is currently not winner,

6.) timestamp vector

Length n₁Wherein ts is_ioi′And the timestamp indicating the latest information update of the earth observation satellite i for the o batch of emergency observation tasks, namely the time point when the updated information from the earth observation satellite i' is received.

5. The homogeneous multi-satellite online collaboration method as claimed in claim 1, wherein a single-satellite online scheduling algorithm is executed on the earth observation satellite, the regular task of the earth observation satellite is scheduled, and the cooperative allocation algorithm is allocated to the emergency observation task of the earth observation satellite, the single-satellite online scheduling algorithm comprises:

(1) at the T-driven scheduling time point, generating a new task plan in the next period time interval by adopting a complete rescheduling strategy in a progressive method, wherein the T-driven scheduling time point determines a specific scheduling time point lT according to a given time interval T, L is more than or equal to 0 and less than or equal to L, LT is more than or equal to H < (L +1) T, and when the scheduling time point lT is reached, calculating and generating the task plan of the next scheduling interval [ lT, (L +1) T ], wherein L is a positive integer, T is the given time interval, L is the maximum T-driven scheduling times, H is the total scheduling interval, and

(2) at C^*-driving rescheduling time points, using a schedule repair strategy in a modified manner, if at a certain time t (0) when the satellite is operating within a given scheduling interval<t<H) On-board emergency observation task cumulant C_tExceeding a given threshold C^*Then a rescheduling calculation is performed, wherein the threshold value C^*Is a critical cumulative number of emergency observation tasks,

except for the two scheduling time points, no scheduling is performed at any other time point.

6. The homogeneous multi-satellite online collaboration method as claimed in claim 5, wherein in the single-satellite online scheduling algorithm, a specific scheduling algorithm at a T-driven scheduling time point is as follows:

inputting:

-a set of emergency observation tasks that have arrived and that have not been scheduled before the T-driven scheduling time point;

-a set of regular observation tasks that have been received and that have not been scheduled before a T-driven scheduling time point;

and (3) outputting:

-a scheduling plan for the next time period T;

the method comprises the following specific steps:

step 11 is respectively from

And

selecting the conventional observation task and the emergency observation task whether the time window falls into the next time period T, and generating a conventional observation task set to be scheduled and solved

And set of emergency observation tasks

Step 12 will

And

integrating into an observation task set;

step 13, according to a set heuristic rule, sequencing the tasks in the integrated observation task set;

step 14, scheduling the tasks in the integrated observation task set one by one according to the sequence to determine whether to add the tasks into the observation task set

Until no more tasks in the integrated observation task set can be added

In (1),

step 15 outputs the scheduling plan in the next time period T

At C^*The scheduling algorithm for the driven rescheduling time points is as follows:

inputting:

within the present time period T and later than C^*-a scheduling plan driving a scheduling time point t;

-a set of emergency observation tasks that have arrived before the scheduling time point t and that are not scheduled;

and (3) outputting:

-a revised dispatch plan at time t,

the method comprises the following specific steps:

step 21 based on the observation timeCondition that the window is within a time interval from time T to the next T-driven scheduling point, from the task set

Selecting emergency observation task to generate new task set

Step 22, according to the set heuristic rule, for

Sequencing the emergency observation tasks in the step (2);

step 23, according to the new task sequence, selecting one by one

In emergency observation task pair

Revising until

No emergency observation task can be added

In (1),

step 24 outputs the revised dispatch plan

Images