CN109087023B - Multi-satellite observation layered scheduling method and system based on divide-and-conquer strategy - Google Patents
Multi-satellite observation layered scheduling method and system based on divide-and-conquer strategy Download PDFInfo
- Publication number
- CN109087023B CN109087023B CN201810967096.7A CN201810967096A CN109087023B CN 109087023 B CN109087023 B CN 109087023B CN 201810967096 A CN201810967096 A CN 201810967096A CN 109087023 B CN109087023 B CN 109087023B
- Authority
- CN
- China
- Prior art keywords
- task
- scheduling
- tasks
- scheduled
- scheduling sequence
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/063—Operations research, analysis or management
- G06Q10/0631—Resource planning, allocation, distributing or scheduling for enterprises or organisations
Abstract
The invention provides a multi-satellite observation layered scheduling method and a multi-satellite observation layered scheduling system based on a divide-and-conquer strategy, which specifically comprise the following steps: s1, distributing the tasks to each track cycle by using a distribution algorithm to form a task set of each track cycle; s2, solving the scheduling sequence of the task set on each orbit round by using an approximate optimization algorithm; s3, the allocation algorithm renews the allocation scheme of the allocation algorithm according to the tasks which are fed back by the scheduling sequence of each track circle and do not generate the scheduling sequence, and a new task set of each track circle is formed; s4, repeating the steps S1, S2 and S3 until reaching the termination condition of the multi-satellite observation layered scheduling. By effectively decomposing and simplifying a complex combined optimization problem into a double-layer plan, the complexity of problem solving is effectively reduced, and the method particularly shows excellent performance when solving a large-scale multi-satellite observation scheduling problem. The invention is applied to the technical field of satellite scheduling.
Description
Technical Field
The invention relates to the technical field of satellite scheduling, in particular to a multi-satellite observation layered scheduling method and system based on a divide-and-conquer strategy.
Background
Earth Observation Satellites (EOSs) fly around the Earth, can acquire image information of a designated area on the Earth surface to meet observation requirements of users, and play a key role in the fields of environmental monitoring, information reconnaissance and the like. In order to fully utilize the scarce satellite resources, an efficient observation scheduling method is researched to improve the application level of the satellite, and the method has important significance. Since the user's earth Observation requirement usually exceeds the Observation capability of the existing Satellite resources, the Satellite Observation Scheduling Problem (SOSP) is a type of over-subscription Problem.
Early studies related to SOSP focused on the scheduling of a single satellite. Vasquez and Hao convert the satellite observation scheduling problem into a classical knapsack model, and a tabu search algorithm is provided to solve the model. In another paper, Vasquez and Hao propose a decomposition-based approach to obtain the minimum upper bound of the optimization problem to evaluate the quality of the tabu search algorithm to solve the resulting satellite observation scheduling scheme. Bensana et al transform the scheduling of the Spot5 satellite into a constraint satisfaction problem and solve the problem separately using either exact or approximate algorithms. Gabrel and vanderpooent propose an acyclic graph model to describe the scheduling problem of the Spot5 satellite by first generating a plurality of valid paths and then selecting the best path. Lin et al use mathematical programming methods to obtain near-optimal satellite observation scheduling schemes. Baek et al use a new genetic algorithm to simulate the satellite actual observation scheduling problem. Miguel et al provides a planning solution for daily imaging scheduling of a Spot5 satellite based on the efficient inequality of node packing and 3-canonical independence systems.
The satellite observation scheduling is to determine the observation activity of each satellite on each orbit circle under the condition of meeting the requirements and constraint conditions of a series of observation tasks, so that the maximization of observation benefits is realized, and the satellite observation scheduling is the key for improving the use efficiency of the satellites. In the conventional multi-satellite scheduling algorithm, each satellite is taken as a scheduling unit, but if a scheduling time interval is large, one satellite can fly around the earth for a plurality of orbital circles (for example, an imaging reconnaissance satellite generally flies around 14 orbital circles every day), different satellites may observe the same target, and the same satellite may observe the same target in different orbital circles, so that an observation conflict is more complicated (generally, it is assumed that one target only needs to be observed once), and particularly, the solution degree is more complicated when a large-scale multi-satellite observation scheduling problem is solved.
Disclosure of Invention
Aiming at the defects in the prior art, the invention aims to provide a multi-satellite observation layered scheduling method and system based on a divide-and-conquer strategy, which effectively reduce the complexity of problem solving and particularly show excellent performance when solving a large-scale multi-satellite observation scheduling problem.
When multi-satellite observation layered scheduling is carried out, satellites are decomposed into orbit circles, each orbit circle is regarded as a resource with earth observation capacity, and the multi-satellite multi-orbit circle scheduling problem is converted into the multi-orbit circle scheduling problem.
Let set O contain a generic orbit coil resource with earth observation capability, O ═ OjI j 1, 2.. times.m }, where M is the base of the set, i.e., the number of orbital turn resources in the scheduling problem. Setting a set T as a target set to be observed, wherein the target set comprises point targets and decomposed strip targets, and T is { T { (T)iI 1, 2., N representing the number of tasks. 0-1 variable xijTo indicate the task t in the scheduling schemeiWhether in track circle ojThe method comprises the following steps: task tiIn track circle ojOn completion xij1, otherwise xij0. The goal of multi-star observation scheduling is to maximize the observation gain, which is the sum of the gains of all completed tasks, i.e.:
in the formula, piIs to complete task tiThe gain that can be achieved.
In the scheduling process, complex constraint conditions including uniqueness constraint, energy constraint, storage capacity constraint and pendulum measurement time constraint need to be met, and the method specifically comprises the following steps:
the uniqueness constraint is as follows: each task only needs to be completed at most once, namely:
energy restraint: track circle ojThe energy consumed by the observation activity and the sensor rolling activity cannot be more than the maximum energy allowed to be consumed on the track circle:
in the formula, eojIndicated in track turns ojThe upper satellite and sensors observe the rate at which activities consume energy; te (te)iRepresenting a task tiThe end time of (d); ts isiRepresenting a task tiThe start time of (c); y isihIs a variable from 1 to 0, representing tiWhether or not to precede thIf the completion is finished, 1 is taken before the completion, otherwise, 0 is taken; esjIndicated in track turns ojRate of energy consumption by the upper sensor yaw activity; thetaiIndicated on track ojUpper observation task tiThe angle of the pendulum is required to be measured by the time sensor; thetahIndicated on track ojUpper observation task thThe angle of the pendulum is required to be measured by the time sensor; ejIndicating track turns ojIs allowed to consume the maximum stored energy.
Storage capacity constraints: in track circle ojIn the above, the observation activity acquires earth observation data and performs on-satellite storage, but the consumed storage resource cannot exceed the maximum storage capacity allowed in the track circle:
in the formula, wjIndicating track turns ojThe rate at which the last observed activity consumes storage resources; wjIndicating track turns ojIs allowed to consume the maximum storage capacity.
And (3) side sway time constraint: when the satellite observes different targets, the satellite needs to be restarted, and meanwhile, due to the change of the relative space geometric relationship, the side swing angle of the sensor needs to be adjusted to align the observation target. Therefore, in the same orbit, sufficient time is required for the satellite sensor to start and calibrate (adjust the yaw angle) between two adjacent observation tasks:
tsh-tei≥aj+|θh-θi|/vj,i,h=1,2,...,N,yih=1,j=1,2,...M
in the formula, tshRepresenting a task thThe start time of (c); a isjIndicated on track ojThe sensor turn-on time before observing a certain target.
Assuming that a multi-satellite scheduling problem considers N tasks and M orbit turns, the variable number is N.M, and the number of constraint conditions is N + 3M. The satellite scheduling problem in a real scene is generally large in scale, and the scheduling problem in the general scale is considered: for example, the number of targets to be observed (including the decomposed observation strips) is 500, and considering 5 satellites, the scheduling period is one day, each satellite flies 14 circles around the ground, and there are about 70 orbit turns, the size of the variable is 500x 70-35000, and the size of the constraint is 500+3x 70-710. The multi-satellite observation scheduling problem is a complex discrete optimization problem, and how to solve the problem quickly and efficiently needs to design an effective scheduling framework and an effective scheduling algorithm.
In summary, the invention adopts the technical scheme that: a multi-satellite observation layered scheduling method based on a divide-and-conquer strategy comprises the following steps:
s1, distributing the tasks to each track cycle by using a distribution algorithm to form a task set of each track cycle;
s2, solving the scheduling sequence of the task set on each orbit round by using an approximate optimization algorithm;
s3, the allocation algorithm renews the allocation scheme of the allocation algorithm according to the tasks which are fed back by the scheduling sequence of each track circle and do not generate the scheduling sequence, and a new task set of each track circle is formed;
s4, repeating the steps S1, S2 and S3 until reaching the termination condition of the multi-satellite observation layered scheduling.
As a further improvement of the above technical solution, in step S2, the approximate optimization algorithm is a simulated annealing algorithm.
As a further improvement of the above technical solution, the step of obtaining the initial scheduling sequence of the current track turn by the simulated annealing algorithm includes:
s211, arranging the tasks to be planned and scheduled in the single-orbit circle task set in sequence from large to small according to the weight to form an initial task set U of the single-orbit circle which is not scheduled;
s212, selecting the task t with the maximum weight from the UkApplying insert and delete neighborhood structure transformations to transform tkAdding the solution into the local solution S to generate a new local solution S ', if the fitness of S ' is higher than that of S, updating S, making S equal to S ', otherwise, not updating S, and simultaneously updating tkMove from UExcept to UjIn which U isjRepresenting an initial set of tasks that are not scheduled on a track turn;
s213, repeating the step S212 until all the tasks in the U are traversed, and acquiring a scheduling sequence S on the current track and a task set U which is not scheduled on the current trackj。
As a further improvement of the above technical solution, the domain search structure of the simulated annealing algorithm includes a neighborhood search structure based on a greedy criterion and a neighborhood search structure based on a probability search, and the simulated annealing algorithm includes a neighborhood search structure dynamic selection strategy: and determining the selection of the domain search structures in the subsequent distribution scheme according to the optimization performance of the two neighborhood search structures in the previous distribution scheme.
As a further improvement of the above technical solution, the simulated annealing algorithm of the neighborhood search structure based on the greedy criterion specifically includes:
s221, acquiring a scheduling sequence S of the current track circle and a task set U which is not scheduled in the current track circlej;
S222, in UjTaking out a task t with the maximum weight which is not scheduled yetkLet Uj=Uj-tk;
S223, adding tkInserting the scheduling sequence S into the scheduling sequence S of the current track circle, judging whether the side sway time constraint is met, if so, directly forming a new scheduling sequence S ', and if not, deleting the scheduling sequence S' and tkAfter the conflicting task a new scheduling sequence S' is formed, in which the deleted tasks are stored SuIn, SuIs a set of tasks for which all orbital turns are not scheduled;
s224, judging whether the new scheduling sequence meets the energy constraint and the storage capacity constraint:
if yes, repeating steps S221, S222, S223 and S224 until UjAll the tasks are completed in a traversal way, the final scheduling sequence S' on the current track is obtained, and the tasks which are not scheduled on the current track are stored into a set SuPerforming the following steps;
if not, deleting the tasks with the minimum weight in sequence untilRepeating the steps S221, S222, S223 and S224 after the energy constraint and the storage capacity constraint are met until UjAll the tasks are completed in a traversal way, the final scheduling sequence S' on the current track is obtained, and the tasks which are not scheduled on the current track are stored into a set SuIn which S isuIs the set of tasks that have not been scheduled for all orbital turns.
As a further improvement of the above technical solution, the simulated annealing algorithm of the neighborhood search structure based on probability search specifically includes:
s231, acquiring a scheduling sequence S of the current track circle and a task set U of which the current track circle is not scheduledj;
S232 at UjTake out any task tkHas a probability ofWherein indexkAs task tkPriority index, order Uj=Uj-tk;
S233, mixing tkInserting the scheduling sequence S into the scheduling sequence S of the current track circle, judging whether the side sway time constraint is met, if so, directly forming a new scheduling sequence S ', and if not, deleting the scheduling sequence S' and tkAfter the conflicting task a new scheduling sequence S' is formed, in which the deleted tasks are stored SuIn, SuIs a set of tasks for which all orbital turns are not scheduled;
s234, judging whether the new scheduling sequence meets the energy constraint and the storage capacity constraint:
if yes, repeating steps S231, S232, S233 and S234 until UjAll the tasks are completed in a traversal way, the final scheduling sequence S' on the current track is obtained, and the tasks which are not scheduled on the current track are stored into a set SuPerforming the following steps;
if not, deleting the task tjHas a probability ofWherein indexjAs task tjPriority index until energy constraint is satisfiedAnd repeating steps S231, S232, S233 and S234 until U after the storage capacity is limitedjAll the tasks are completed in a traversal way, the final scheduling sequence S' on the current track is obtained, and the tasks which are not scheduled on the current track are stored into a set SuIn which S isuIs the set of tasks that have not been scheduled for all orbital turns.
As a further improvement of the above technical solution, the allocation algorithm is an ant colony optimization algorithm.
As a further improvement of the above technical solution, step S3 specifically includes:
s31, acquiring task set S with unscheduled all orbit turns according to each orbit turn scheduling sequenceu;
S32, Slave SuSelecting a task with the highest weight, inserting the task into the orbit with the smallest conflict weight, if the conflict weight of the task is smaller than the weight of the task, inserting the task successfully and updating the distribution scheme, otherwise, not updating the distribution scheme;
s33, repeating the steps S32 to SuAll unscheduled tasks in (a) are traversed to completion.
As a further improvement of the above technical solution, in step S4, the termination condition of the multi-satellite observation hierarchical scheduling is that the calculation time reaches a preset time value or the number of iterations reaches a preset number of times.
The invention also discloses a multi-satellite observation layered scheduling system based on the divide-and-conquer strategy, which adopts the technical scheme that:
a multi-satellite observation hierarchical scheduling system based on a divide-and-conquer strategy comprises a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of the method when executing the computer program.
The invention has the beneficial technical effects that:
the invention divides the tasks into different orbit turns through the distribution algorithm, each orbit turn corresponds to a task set, which is a task distribution layer of the method, then the observation scheme of the satellite on each orbit turn is obtained through the approximate optimization algorithm, which is a task scheduling layer of the method, and finally the distribution scheme of the distribution algorithm is renewed according to the feedback of the scheduling sequence of each orbit turn, thereby forming a new task set of each new orbit turn. And iteration is repeated until the algorithm termination condition is met, a complex combined optimization problem is effectively decomposed and simplified to form a double-layer plan, the problem solving complexity is effectively reduced, and the method particularly shows excellent performance when a large-scale multi-satellite observation scheduling problem is solved.
Drawings
Fig. 1 is a frame configuration diagram of the present embodiment.
Detailed Description
In order to facilitate the practice of the invention, further description is provided below with reference to specific examples.
As shown in fig. 1, a multi-satellite observation hierarchical scheduling method based on a divide-and-conquer strategy includes the following steps:
s1, distributing the tasks to each track cycle by using a distribution algorithm to form a task set of each track cycle;
s2, solving the scheduling sequence of the task set on each orbit round by using an approximate optimization algorithm;
s3, the allocation algorithm renews the allocation scheme of the allocation algorithm according to the tasks which are fed back by the scheduling sequence of each track circle and do not generate the scheduling sequence, and a new task set of each track circle is formed;
and S4, repeating the steps S1, S2 and S3 until the end condition of the multi-satellite observation layered scheduling is reached, wherein the end condition of the multi-satellite observation layered scheduling is that the calculation time reaches a preset time value or the iteration times reaches a preset number.
In step S2, the approximate optimization algorithm is a simulated annealing algorithm, and the step of obtaining the initial scheduling sequence of the current track turn by the simulated annealing algorithm includes:
s211, arranging the tasks to be planned and scheduled in the single-orbit circle task set in sequence from large to small according to the weight to form an initial task set U of the single-orbit circle which is not scheduled;
s212, selecting the task t with the maximum weight from the UkApplying insert and delete neighborhood structure transformations to transform tkAdding the solution into the local solution S to generate a new local solution S ', if the fitness of S ' is higher than that of S, updating S, making S equal to S ', otherwise, not updating S, and simultaneously updating tkIs removed from U to UjIn which U isjRepresenting an initial task set which is not scheduled in a track circle, wherein the insertion and deletion of neighborhood transformation represents a neighborhood searching mode of firstly inserting a task and then deleting a task which conflicts with the newly inserted task, a local solution refers to a solution with potential to continue adding the task, and the fitness measures the quality of the solution, wherein the solution is an initial scheduling sequence required to be obtained;
s213, repeating the step S212 until all the tasks in the U are traversed, and acquiring a scheduling sequence S of the current track circle and a task set U of which the current track circle is not scheduledj。
The domain search structure of the simulated annealing algorithm in this embodiment includes a neighborhood search structure based on a greedy criterion and a neighborhood search structure based on a probability search, and the simulated annealing algorithm includes a neighborhood search structure dynamic selection strategy: and according to the optimization performance of the two neighborhood searching structures in the previous distribution scheme, determining the selection of the neighborhood searching structures in the subsequent distribution scheme:
in the process of solving the problem in an optimized manner, different neighborhood structure search strategies may have different advantages, so this embodiment proposes an adaptive mechanism to realize dynamic adjustment of two different neighborhood search structures, and determines the probability of their selection in the subsequent iteration process according to the optimization performance of each neighborhood search structure in the previous distribution scheme, that is, the optimization performance in the previous given iteration number, which is an idea based on reinforcement learning. Suppose pro1And pro2Respectively expressed as probabilities of selecting a greedy criterion-based neighborhood search structure and a probability search-based neighborhood search structure, pro is set upon initialization of the algorithmiEvery certain number of iterations Itr, the selection probability of each neighborhood structure is updated by the following rule:
proi=proi′/∑i=1,2proi′
in the formula, proi' represents an intermediate variable; eta is an inertia weight factor and represents the proportion of the selection probability before the neighborhood structure; (1- η) represents the weight of the current latest historical search experience to the update selection probability; seliThe number of times the ith neighborhood structure is selected in the iteration process of the latest Itr times; suciIndicating the number of times a higher quality solution was produced using the ith neighborhood structure.
And finally, standardizing the selection probability of different neighborhood structures. The update rule of the neighborhood structure selection probability indicates that a neighborhood structure which can generate a better solution can obtain a higher selection probability. Therefore, dynamic self-adaptive selection of the neighborhood structure in the running process of the optimization algorithm is realized.
In neighborhood structure search, an insert/delete neighborhood is a classic and efficient neighborhood structure, also commonly referred to as a swap neighborhood. Selecting the task with the largest weight from the task set which is not scheduled and not traversed each time, inserting the task into the current observation scheduling sequence, and if the task which conflicts with the inserted task exists in the scheduling sequence, deleting all the conflicting tasks, so that the simulated annealing algorithm of the neighborhood search structure based on the greedy criterion specifically comprises the following steps:
s221, acquiring a scheduling sequence S of the current track circle and a task set U which is not scheduled in the current track circlej;
S222, in UjTaking out a task t with the maximum weight which is not scheduled yetk,tkIs not contained in the tabu list, order Uj=Uj-tkWherein t iskThe fact that the tasks are not contained in the tabu table indicates an algorithm mechanism, namely, the tasks placed in the tabu table do not participate in neighborhood structure search;
s223, adding tkInserting the scheduling sequence S into the scheduling sequence S of the current track circle, judging whether the side sway time constraint is met, if so, directly forming a new scheduling sequence S ', and if not, deleting the scheduling sequence S' and tkConflictAfter the task (S) a new scheduling sequence S' is formed, wherein the deleted tasks are stored SuIn, SuIs a set of tasks for which all orbital turns are not scheduled;
s224, judging whether the new scheduling sequence meets the energy constraint and the storage capacity constraint:
if yes, repeating steps S221, S222, S223 and S224 until UjAll the tasks are completed in a traversal way, the final scheduling sequence S' on the current track is obtained, and the tasks which are not scheduled on the current track are stored into a set SuPerforming the following steps;
if not, deleting the task with the minimum weight in sequence until the energy constraint and the storage capacity constraint are met, and repeating the steps S221, S222, S223 and S224 until UjAll the tasks are completed in a traversal way, the final scheduling sequence S' on the current track is obtained, and the tasks which are not scheduled on the current track are stored into a set SuIn which S isuIs the set of tasks that have not been scheduled for all orbital turns.
A neighborhood search structure based on a greedy criterion facilitates local searching. In order to improve the diversity of the solution, a neighborhood searching structure based on probability searching is also designed, and in the neighborhood structure, not only the weight of the task but also the side pendulum resource and the time window resource which are potentially consumed for completing each task are considered. Integrating weights, yaw angles and time windows for each task t in the process of inserting and replacing the neighborhood structure transformationkCalculating a priority indexkFirstly, respectively calculating a weight index, a time window index and a yaw angle index:
the weight index is:
in the formula, iWkIs a weight index; w is akIs the weight of the kth task; w is aiIs the weight of the ith task; n isjIs UjThe number of tasks.
The time window index is:
in the formula, iTkIs a time window index; spankIs tkThe length of the time window of (a); spaniIs tiThe length of the time window of (a).
The side sway angle indexes are as follows:
in the formula, iTkIs a lateral swing angle index; thetaiOn the track ojUpper observation task tiThe angle of the pendulum is required to be measured by the time sensor; thetakOn the track ojUpper observation task tkThe sensor needs to measure the swing angle.
Priority indexkThe final expression is:
indexk=iWk α·iTk β·iθk γ
in the formula, α, β, and γ represent influence factors of different elements, respectively, and are specifically set by a user for a specific problem.
The simulated annealing algorithm of the neighborhood search structure based on probability search specifically comprises the following steps:
s231, acquiring a scheduling sequence S of the current track circle and a task set U of which the current track circle is not scheduledj;
S232 at UjTake out any task tkHas a probability ofWherein indexkAs task tkPriority index, tkIs not contained in the tabu list, order Uj=Uj-tk;
S233, mixing tkInserting the data into a scheduling sequence S of the current track circle, judging whether the side sway time constraint is met, and if the data meets the side sway time constraint, judging whether the data meets the side sway time constraintIf the new scheduling sequence S 'is satisfied, a new scheduling sequence S' is directly formed, and if the new scheduling sequence S 'is not satisfied, the new scheduling sequence S' is deleted and tkAfter the conflicting task a new scheduling sequence S' is formed, in which the deleted tasks are stored SuIn, SuIs a set of tasks for which all orbital turns are not scheduled;
s234, judging whether the new scheduling sequence meets the energy constraint and the storage capacity constraint:
if yes, repeating steps S231, S232, S233 and S234 until UjAll the tasks are completed in a traversal way, the final scheduling sequence S' on the current track is obtained, and the tasks which are not scheduled on the current track are stored into a set SuPerforming the following steps;
if not, deleting the task tjHas a probability ofWherein indexjAs task tjThe priority index is obtained until the energy constraint and the storage capacity constraint are met, and the steps S231, S232, S233 and S234 are repeated until UjAll the tasks are completed in a traversal way, the final scheduling sequence S' on the current track is obtained, and the tasks which are not scheduled on the current track are stored into a set SuIn which S isuIs the set of tasks that have not been scheduled for all orbital turns.
The allocation algorithm is an ant colony optimization algorithm, and the step S3 specifically includes:
s31, acquiring task set S with unscheduled all orbit turns according to each orbit turn scheduling sequenceu;
S32, Slave SuSelecting a task with the highest weight, inserting the task into the orbit with the smallest conflict weight, if the conflict weight of the task is smaller than the weight of the task, inserting the task successfully and updating the distribution scheme, otherwise, not updating the distribution scheme;
s33, repeating the steps S32 to SuAll unscheduled tasks in (a) are traversed to completion.
When the task is first assigned to each track turn by using the ant colony optimization algorithm in step S1, the ant colony optimization algorithm performs initial assignment according to the parameter settings, and a first assignment scheme is directly formed.
As a further improvement of the above technical solution, in step S4, the algorithm termination condition is that the calculation time reaches a preset time value or the iteration number reaches a preset number of times.
The embodiment also discloses a multi-satellite observation layered scheduling system based on the divide-and-conquer strategy, which adopts the technical scheme that:
a multi-satellite observation layered scheduling system based on a divide-and-conquer strategy comprises a memory and a processor, wherein a computer program is stored in the memory, and the steps of the method are realized when the processor executes the computer program.
To evaluate the performance of the ACO-SA algorithm based on the divide and conquer strategy in this example, it was compared with the Tabu Search (TS), Genetic Algorithm (GA), Simulated Annealing (SA) highest priority scheduling algorithm (HPFS), and ant colony optimization algorithm with local search (ACO-LS). The ACO-LS utilizes the ant colony optimization algorithm and the local search strategy to combine to generate a solution of the multi-star planning scheduling problem, but a divide-and-conquer strategy is not used in the algorithm, and the scheduling problem is regarded as a whole to be solved.
In order to comprehensively and more comprehensively evaluate the performance of the algorithm, the embodiment provides six simulation scenarios. The satellites used comprise two groups, the first group being 8 resources and a keepsake series of scout satellites and the second group being 16 scout satellites. Each satellite travels around the earth for about 100 minutes for 14 orbital turns a day. The sensor has a lateral yaw angle range of-330, 330. The scheduling period is 24 hours. And calculating by using special software STK to obtain a visible time window and a yaw angle of the satellite and the target. The six scenes are respectively: observing a Daxing-an mountain forest area by 8 satellites in a first scene; the second scene is the observation of 8 satellites on the Changbai mountain forest area; the third scene is the observation of 8 satellites on a great Khingan forest area and a Changbai mountain forest area; observing the Daxing-an mountain forest area by 16 satellites in the fourth scene; the fifth scene is the observation of 16 satellites on the Changbai mountain forest area; the sixth scenario is the observation of 16 satellites on the great Xingan mountain forest area and the Changbai mountain forest area.
The algorithm was implemented using C + + language programming and the experiments were run on a computer with a Windows 7 system with Intel Core (TM) i7-4810MQ @2.8GHZ CPU, 16.0GB RAM.
The calculation results of the first to sixth sets of simulation experiment scenarios are shown in the data in tables 1 to 6. Through observation of experimental data, the following phenomena can be easily found out:
the ACO-SA stably shows the best performance in each group of experiments, and shows that the multi-satellite observation and scheduling problem can be effectively solved by adopting a divide-and-conquer strategy to integrate an ant colony algorithm and a simulated annealing algorithm. Particularly, the ACO-SA algorithm always generates a better observation scheduling scheme than the ACO-LS algorithm, and the dividing and controlling strategy has obvious advantages for solving the large-scale satellite observation scheduling problem;
2. under the simulation scene of 8 satellites, the observation coverage rate of the forest region is obviously smaller than that of the forest region under the simulation scene of 16 satellites, and the fact that the proper increase of satellite observation resources is an important means for improving the forest observation coverage capability is shown;
3. through the data of the third group of simulation experiments, it can be found that effective observation coverage of the greater Khingan mountains and the Changbai mountains is difficult to effectively complete within 24 hours for 8 satellites, and the coverage rate of about 50% can only be achieved by adopting the most efficient ACO-SA algorithm. Forest resources covering the Changbai mountain area can be basically and effectively observed within 24 hours by 16 satellites, and the observation completion rate reaches more than 97 percent;
4. according to timeliness of multi-satellite observation planning and scheduling, the timeliness of the ACO-SA is in an intermediate level, the time for generating the satellite observation scheme by the ACO-SA is approximately linearly increased along with the increase of the scale of the number of tasks and the number of satellites, and under the maximum scale (the sixth group of simulation scenes), a high-quality planning and scheduling scheme can be obtained within 900 seconds, so that the applicability of the ACO-SA algorithm is shown.
TABLE 1 results of different satellite observation scheduling algorithms in a first set of test scenarios
Algorithm | CT | NT | Cov(%) | Time(s) |
TS | 1862 | 1215 | 65.22% | 321.52 |
GA | 1862 | 1244 | 66.52% | 425.62 |
SA | 1862 | 1198 | 62.12% | 564.24 |
ACO-LS | 1862 | 1254 | 67.25% | 368.45 |
ACO-SA | 1862 | 1351 | 72.65% | 388.78 |
TABLE 2 results of different satellite observation scheduling algorithms in a second set of test scenarios
Algorithm | CT | NT | Cov(%) | Time(s) |
TS | 1456 | 1025 | 70.40% | 280.25 |
GA | 1456 | 1085 | 74.52% | 355.32 |
SA | 1456 | 987 | 67.79% | 451.52 |
ACO-LS | 1456 | 1102 | 75.69% | 298.12 |
ACO-SA | 1456 | 1203 | 83.24% | 326.45 |
TABLE 3 results of different satellite observation scheduling algorithms in third set of test scenarios
Algorithm | CT | NT | Cov(%) | Time(s) |
TS | 3318 | 1425 | 42.95% | 582.35 |
GA | 3318 | 1524 | 45.93% | 565.15 |
SA | 3318 | 1324 | 39.90% | 684.60 |
ACO-LS | 3318 | 1588 | 46.85% | 535.40 |
ACO-SA | 3318 | 1658 | 50.02% | 606.66 |
TABLE 4 calculation results of different satellite observation scheduling algorithms in a fourth group of test scenarios
Algorithm | CT | NT | Cov(%) | Time(s) |
TS | 1862 | 1625 | 87.27% | 541.12 |
GA | 1862 | 1678 | 90.12% | 655.45 |
SA | 1862 | 1598 | 85.82% | 743.15 |
ACO-LS | 1862 | 1699 | 91.25% | 524.36 |
ACO-SA | 1862 | 1742 | 93.56% | 560.65 |
TABLE 5 results of different satellite observation scheduling algorithms in the fifth set of test scenarios
Algorithm | CT | NT | Cov(%) | Time(s) |
TS | 1456 | 1355 | 93.06% | 482.32 |
GA | 1456 | 1382 | 94.92% | 505.24 |
SA | 1456 | 1275 | 87.57% | 661.32 |
ACO-LS | 1456 | 1398 | 96.02% | 458.55 |
ACO-SA | 1456 | 1418 | 97.39% | 500.48 |
TABLE 6 results of different satellite observation and scheduling algorithms in the sixth set of test scenarios
In the above table, the technical term ACO-SA is the algorithm of the present invention, TS is tabu search, GA is genetic algorithm, SA is simulated annealing, and ACO-LS is an algorithm for generating a solution to the multi-star scheduling problem by combining the ant colony optimization algorithm with a local search strategy.
The divide-and-conquer framework provided by the embodiment has universality and is suitable for many types of planning and scheduling, such as unmanned plane scheduling, vehicle path planning and the like. Subsequent work was to further investigate the scheduling problem of agile and hypersensitive satellites, and the integrated scheduling problem of imaging and data transmission.
The foregoing description of the preferred embodiments of the present invention has been included to describe the features of the invention in detail, and is not intended to limit the inventive concepts to the particular forms of the embodiments described, as other modifications and variations within the spirit of the inventive concepts will be protected by this patent. The subject matter of the present disclosure is defined by the claims, not by the detailed description of the embodiments.
Claims (5)
1.A multi-satellite observation layered scheduling method based on a divide-and-conquer strategy is characterized by comprising the following steps:
s1, distributing the tasks to each track cycle by using a distribution algorithm to form a task set of each track cycle;
s2, solving the scheduling sequence of the task set on each orbit round by using an approximate optimization algorithm;
s3, updating the allocation scheme of the allocation algorithm again according to the tasks which are fed back by the scheduling sequences of each track circle and do not generate the scheduling sequences, and further forming a new task set of each track circle;
s4, repeating the steps S1, S2 and S3 until reaching the termination condition of the multi-satellite observation layered scheduling;
in step S2, the approximate optimization algorithm is a simulated annealing algorithm;
the step of acquiring the initial scheduling sequence of the current track turn by the simulated annealing algorithm comprises the following steps:
s211, arranging the tasks to be planned and scheduled in the single-orbit circle task set in sequence from large to small according to the weight to form an initial task set U of the single-orbit circle which is not scheduled;
s212, selecting the task t with the maximum weight from the UkApplying insert and delete neighborhood structure transformations to transform tkAdding the solution into the local solution S to generate a new local solution S ', if the fitness of S ' is higher than that of S, updating S, making S equal to S ', otherwise, not updating S, and simultaneously updating tkMove from U to UjIn which U isjRepresenting a set of tasks that are not scheduled on a track turn;
s213, repeating the step S212 until all the tasks in the U are traversed, and acquiring a scheduling sequence S of the current track circle and a task set U of which the current track circle is not scheduledj;
The field search structure of the simulated annealing algorithm comprises a neighborhood search structure based on a greedy criterion and a neighborhood search structure based on probability search, and the simulated annealing algorithm comprises a neighborhood search structure dynamic selection strategy: according to the optimization performance of the two neighborhood search structures in the previous distribution scheme, the selection of the neighborhood search structure in the subsequent distribution scheme is determined, and the method specifically comprises the following steps:
suppose pro1And pro2Respectively expressed as a summary for selecting a greedy criterion-based neighborhood search structure and a probability search-based neighborhood search structureRate, setting pro at algorithm initializationiEvery certain number of iterations Itr, the selection probability of each neighborhood structure is updated by the following rule:
in the formula (I), the compound is shown in the specification,representing an intermediate variable; eta is an inertia weight factor and represents the proportion of the selection probability before the neighborhood structure; (1- η) represents the weight of the current latest historical search experience to the update selection probability; seliThe number of times the ith neighborhood structure is selected in the iteration process of the latest Itr times; suciIndicating the number of times a higher quality solution was produced using the ith neighborhood structure;
the simulated annealing algorithm of the neighborhood search structure based on probability search specifically comprises the following steps:
s231, acquiring a scheduling sequence S of the current track circle and a task set U of which the current track circle is not scheduledj;
S232 at UjTake out any task tkHas a probability ofWherein indexkAs task tkPriority index, order Uj=Uj-tk;
S233, mixing tkInserting the scheduling sequence S into the scheduling sequence S of the current track circle, judging whether the side sway time constraint is met, if so, directly forming a new scheduling sequence S ', and if not, deleting the scheduling sequence S' and tkAfter the conflicting task a new scheduling sequence S' is formed, in which the deleted tasks are stored SuIn, SuIs a set of tasks for which all orbital turns are not scheduled;
s234, judging whether the new scheduling sequence meets the energy constraint and the storage capacity constraint:
if yes, repeating steps S231, S232, S233 and S234 until UjAll the tasks are completed in a traversal way, the final scheduling sequence S' on the current track is obtained, and the tasks which are not scheduled on the current track are stored into a set SuPerforming the following steps;
if not, deleting the task tjHas a probability ofWherein indexjAs task tjThe priority index is obtained until the energy constraint and the storage capacity constraint are met, and the steps S231, S232, S233 and S234 are repeated until UjAll the tasks are completed in a traversal way, the final scheduling sequence S' on the current track is obtained, and the tasks which are not scheduled on the current track are stored into a set SuIn which S isuIs a set of tasks for which all orbital turns are not scheduled;
in the neighborhood searching structure based on probability search, not only the weight of the task, but also the side swing resource and the time window resource which are potentially consumed for completing each task, and the weight, the side swing angle and the time window are integrated, and in the process of inserting and replacing the neighborhood structure transformation, each task t is subjected tokCalculating a priority indexkFirstly, respectively calculating a weight index, a time window index and a yaw angle index:
the weight index is:
in the formula, iWkIs a weight index; w is akIs the weight of the kth task; w is aiIs the weight of the ith task; n isjIs UjThe number of tasks;
the time window index is:
in the formula, iTkIs a time window index; spankIs tkThe length of the time window of (a); spaniIs tiThe length of the time window of (a);
the side sway angle indexes are as follows:
in the formula, iTkIs a lateral swing angle index; thetaiOn the track ojUpper observation task tiThe angle of the pendulum is required to be measured by the time sensor; thetakOn the track ojUpper observation task tkThe angle of the pendulum is required to be measured by the time sensor;
priority indexkThe final expression is:
indexk=iWk α·iTk β·iθk γ
in the formula, alpha, beta and gamma respectively represent influence factors of different elements and are specifically set by a user aiming at specific problems;
the simulated annealing algorithm of the neighborhood search structure based on the greedy criterion specifically comprises the following steps:
s221, acquiring a scheduling sequence S of the current track circle and a task set U which is not scheduled in the current track circlej;
S222, in UjTaking out a task t with the maximum weight which is not scheduled yetkLet Uj=Uj-tk;
S223, adding tkInserting the scheduling sequence S into the scheduling sequence S of the current track circle, judging whether the side sway time constraint is met, if so, directly forming a new scheduling sequence S ', and if not, deleting the scheduling sequence S' and tkAfter the conflicting task a new scheduling sequence S' is formed, in which the deleted tasks are stored SuIn, SuIs a set of tasks for which all orbital turns are not scheduled;
s224, judging whether the new scheduling sequence S' meets the energy constraint and the storage capacity constraint:
if yes, repeating steps S221, S222, S223 and S224 until UjAll the tasks are completed in a traversal way, the final scheduling sequence S' on the current track is obtained, and the tasks which are not scheduled on the current track are stored into a set SuPerforming the following steps;
if not, deleting the task with the minimum weight in sequence until the energy constraint and the storage capacity constraint are met, and repeating the steps S221, S222, S223 and S224 until UjAll the tasks are completed in a traversal way, the final scheduling sequence S' on the current track is obtained, and the tasks which are not scheduled on the current track are stored into a set SuIn which S isuIs the set of tasks that have not been scheduled for all orbital turns.
2. The multi-satellite observation layered scheduling method based on the divide-and-conquer strategy according to claim 1, wherein the distribution algorithm is an ant colony optimization algorithm.
3. The multi-satellite observation layered scheduling method based on the divide-and-conquer strategy as claimed in claim 2, wherein the step S3 specifically comprises:
s31, acquiring task set S with unscheduled all orbit turns according to each orbit turn scheduling sequenceu;
S32, Slave SuSelecting a task with the highest weight, inserting the task into the orbit with the smallest conflict weight, if the conflict weight of the task is smaller than the weight of the task, inserting the task successfully and updating the distribution scheme, otherwise, not updating the distribution scheme;
s33, repeating the steps S32 to SuAll unscheduled tasks in (a) are traversed to completion.
4. The multi-satellite observation layered scheduling method based on the divide-and-conquer strategy as claimed in claim 1, wherein in step S4, the termination condition of the multi-satellite observation layered scheduling is that the calculation time reaches a preset time value or the number of iterations reaches a preset number of times.
5. A multi-satellite observation hierarchical scheduling system based on a divide-and-conquer strategy, comprising a memory and a processor, the memory storing a computer program, characterized in that the processor implements the steps of the method according to any one of claims 1 to 4 when executing the computer program.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810967096.7A CN109087023B (en) | 2018-08-23 | 2018-08-23 | Multi-satellite observation layered scheduling method and system based on divide-and-conquer strategy |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810967096.7A CN109087023B (en) | 2018-08-23 | 2018-08-23 | Multi-satellite observation layered scheduling method and system based on divide-and-conquer strategy |
Publications (2)
Publication Number | Publication Date |
---|---|
CN109087023A CN109087023A (en) | 2018-12-25 |
CN109087023B true CN109087023B (en) | 2021-03-02 |
Family
ID=64794304
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201810967096.7A Active CN109087023B (en) | 2018-08-23 | 2018-08-23 | Multi-satellite observation layered scheduling method and system based on divide-and-conquer strategy |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109087023B (en) |
Families Citing this family (12)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109767128B (en) * | 2019-01-15 | 2021-06-11 | 中国人民解放军国防科技大学 | Imaging satellite autonomous task planning method based on machine learning |
CN110766284B (en) * | 2019-09-24 | 2022-04-08 | 合肥工业大学 | Multi-star task synthesis method and system |
CN110728447B (en) * | 2019-10-10 | 2021-03-09 | 合肥工业大学 | Partitioned satellite task planning method for achieving regional target coverage at earliest |
CN111176807B (en) * | 2019-12-09 | 2023-12-12 | 中国空间技术研究院 | Multi-star collaborative task planning method |
CN112016812B (en) * | 2020-08-06 | 2022-07-12 | 中南大学 | Multi-unmanned aerial vehicle task scheduling method, system and storage medium |
CN112070370B (en) * | 2020-08-21 | 2024-03-08 | 中南大学 | Relay satellite task planning method, system and storage medium |
CN112613644B (en) * | 2020-12-09 | 2022-11-04 | 合肥工业大学 | Multi-satellite multi-station integrated scheduling method and system |
CN112737660B (en) * | 2020-12-09 | 2022-06-10 | 合肥工业大学 | Multi-satellite multi-station data downloading scheduling method and system |
CN113158384B (en) * | 2021-03-03 | 2021-10-08 | 东北石油大学 | Oil and gas pipeline route planning method and system |
CN113408063B (en) * | 2021-07-20 | 2022-06-21 | 中南大学 | Multi-satellite multi-reconnaissance target-oriented track maneuvering optimization method |
CN113487220B (en) * | 2021-07-23 | 2022-07-15 | 中南大学 | Static target observation-oriented space-sky heterogeneous earth observation resource cooperative scheduling method |
CN114604443B (en) * | 2022-02-17 | 2023-06-06 | 中国电子科技集团公司第十研究所 | Multi-star measurement and control scheduling method and system under real and complex constraint conditions |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104361234A (en) * | 2014-11-15 | 2015-02-18 | 北京理工大学 | Method for optimizing multi-star multitask observation dispatching under complicated constraint condition |
CN106951312A (en) * | 2016-08-31 | 2017-07-14 | 合肥工业大学 | The dispatching method and system of many star tasks |
CN108335012A (en) * | 2017-12-26 | 2018-07-27 | 佛山科学技术学院 | A kind of intelligence remote sensing satellite stratification distributed freedom cotasking planning system |
-
2018
- 2018-08-23 CN CN201810967096.7A patent/CN109087023B/en active Active
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104361234A (en) * | 2014-11-15 | 2015-02-18 | 北京理工大学 | Method for optimizing multi-star multitask observation dispatching under complicated constraint condition |
CN106951312A (en) * | 2016-08-31 | 2017-07-14 | 合肥工业大学 | The dispatching method and system of many star tasks |
CN108335012A (en) * | 2017-12-26 | 2018-07-27 | 佛山科学技术学院 | A kind of intelligence remote sensing satellite stratification distributed freedom cotasking planning system |
Non-Patent Citations (1)
Title |
---|
"An adaptive Simulated Annealing-based satellite observation scheduling method combined with a dynamic task clustering strategy";WuGuoHua等;《Computers&Industrial Engineering》;20171130;摘要,第3、5、6节 * |
Also Published As
Publication number | Publication date |
---|---|
CN109087023A (en) | 2018-12-25 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109087023B (en) | Multi-satellite observation layered scheduling method and system based on divide-and-conquer strategy | |
CN109829636B (en) | Satellite emergency task scheduling planning method based on dynamic priority | |
US10726368B2 (en) | Innovative satellite scheduling method based on genetic algorithms and simulated annealing and related mission planner | |
CN109948944B (en) | Satellite task scheduling method and system | |
CN113128828B (en) | Satellite observation distributed online planning method based on multi-agent reinforcement learning | |
He et al. | Scheduling multiple agile earth observation satellites with an edge computing framework and a constructive heuristic algorithm | |
CN103345504B (en) | The operator construction method of single star scheduling | |
CN105225003A (en) | A kind of cuckoo searching algorithm solves the method for UAV multitask investigation decision problem | |
CN112766813A (en) | Air-space cooperative observation complex task scheduling method and system | |
Wu et al. | Ensemble of metaheuristic and exact algorithm based on the divide-and-conquer framework for multisatellite observation scheduling | |
Song et al. | Improved genetic algorithm with local search for satellite range scheduling system and its application in environmental monitoring | |
CN109039428B (en) | Relay satellite single-address antenna scheduling random search method based on conflict resolution | |
CN109933842A (en) | A kind of mobile target list star mission planning method based on constraint satisfaction genetic algorithm | |
CN111612384B (en) | Multi-star relay task planning method with time resolution constraint | |
CN109034670B (en) | Satellite on-orbit activity planning method and system | |
Eddy et al. | Markov decision processes for multi-objective satellite task planning | |
CN113051815A (en) | Agile imaging satellite task planning method based on independent pointer network | |
Yang | Reevaluation and renegotiation of climate change coalitions—a sequential closed-loop game approach | |
CN110113212B (en) | Relay satellite single-address antenna scheduling method suitable for breakpoint continuous transmission | |
CN112990478B (en) | Federal learning data processing system | |
CN109299490A (en) | Mission planning method is stared over the ground based on band introduce taboo list ant group algorithm video satellite | |
CN116579513A (en) | Material distribution path optimization method and system based on improved whale optimization algorithm | |
CN113657718B (en) | Multi-robot dynamic alliance task allocation method and related device | |
Li et al. | Application of population based incremental learning algorithm in satellite mission planning | |
Long et al. | Improved particle swarm optimization with reverse learning and neighbor adjustment for space surveillance network task scheduling |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |