CN111651905B - Agile satellite scheduling method considering time-dependent conversion time - Google Patents
Agile satellite scheduling method considering time-dependent conversion time Download PDFInfo
- Publication number
- CN111651905B CN111651905B CN202010654910.7A CN202010654910A CN111651905B CN 111651905 B CN111651905 B CN 111651905B CN 202010654910 A CN202010654910 A CN 202010654910A CN 111651905 B CN111651905 B CN 111651905B
- Authority
- CN
- China
- Prior art keywords
- time
- observation
- target
- observed
- task
- 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
- 238000006243 chemical reaction Methods 0.000 title claims abstract description 45
- 238000000034 method Methods 0.000 title claims abstract description 43
- 230000036962 time dependent Effects 0.000 title claims abstract description 21
- 238000003780 insertion Methods 0.000 claims abstract description 98
- 230000037431 insertion Effects 0.000 claims abstract description 98
- 238000007781 pre-processing Methods 0.000 claims abstract description 7
- 230000007704 transition Effects 0.000 claims description 28
- 239000013598 vector Substances 0.000 claims description 22
- 230000008901 benefit Effects 0.000 claims description 5
- 230000006870 function Effects 0.000 claims description 5
- 238000010845 search algorithm Methods 0.000 claims description 5
- FGXWKSZFVQUSTL-UHFFFAOYSA-N domperidone Chemical compound C12=CC=CC=C2NC(=O)N1CCCN(CC1)CCC1N1C2=CC=C(Cl)C=C2NC1=O FGXWKSZFVQUSTL-UHFFFAOYSA-N 0.000 claims description 3
- 230000009191 jumping Effects 0.000 claims description 3
- 238000010276 construction Methods 0.000 claims description 2
- 238000004364 calculation method Methods 0.000 abstract description 20
- 230000008569 process Effects 0.000 abstract description 4
- 238000002474 experimental method Methods 0.000 description 10
- 230000008859 change Effects 0.000 description 7
- 238000010586 diagram Methods 0.000 description 5
- 230000000694 effects Effects 0.000 description 4
- 238000012966 insertion method Methods 0.000 description 4
- 230000003044 adaptive effect Effects 0.000 description 3
- 206010034719 Personality change Diseases 0.000 description 2
- 238000004458 analytical method Methods 0.000 description 2
- 238000009826 distribution Methods 0.000 description 2
- 238000003384 imaging method Methods 0.000 description 2
- 230000006978 adaptation Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000000052 comparative effect Effects 0.000 description 1
- 230000003111 delayed effect Effects 0.000 description 1
- 238000012217 deletion Methods 0.000 description 1
- 230000037430 deletion Effects 0.000 description 1
- 230000001419 dependent effect Effects 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000003875 gradient-accelerated spectroscopy Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000005096 rolling process Methods 0.000 description 1
- 238000003860 storage Methods 0.000 description 1
- 238000009827 uniform distribution Methods 0.000 description 1
- 238000012795 verification Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2111/00—Details relating to CAD techniques
- G06F2111/04—Constraint-based CAD
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2119/00—Details relating to the type or aim of the analysis or the optimisation
- G06F2119/12—Timing analysis or timing optimisation
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Computer Hardware Design (AREA)
- Evolutionary Computation (AREA)
- Geometry (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses an agile satellite scheduling method considering time dependence conversion time, which comprises the steps of obtaining a target sequence to be observed and preprocessing the target sequence; constructing an agile satellite scheduling model of time-dependent conversion time; inputting the preprocessed target sequence to be observed into an agile satellite scheduling model, and solving the agile scheduling model; and (3) outputting the solving result in the step (3) to obtain a scheduling scheme of the agile satellite. Because of considering the time dependence of the gesture conversion time between the continuous observation tasks, the constructed agile satellite scheduling model can better meet the scheduling requirement, and the scheduling efficiency of the solved scheduling scheme is higher. In the solving process, the unreachable earliest time and the unreachable latest time of each time window and the earliest starting time and the latest starting time between any pair of observation windows in the same circle are pre-calculated, so that repeated calculation for a plurality of times is reduced when the insertion operation is attempted, the calculation time is reduced, and the scheduling efficiency is improved.
Description
Technical Field
The invention relates to the technical field of satellite scheduling, in particular to an agile satellite scheduling method considering time dependence conversion time.
Background
Earth-looking satellites (EOSs) are space-platform systems that can acquire images of specific areas of the earth's surface according to various viewing needs. Along with the development of science and technology and economy, EOSs has been applied to various aspects of many economic societies, such as natural resource exploration, disaster early warning, climate change analysis, and the like. Considering the inconvenience cost of transmitting satellites, the efficient and reliable EOSs scheduling method plays a very critical role in improving the performance of a satellite observation system. The scheduling of earth observation satellites means that a group of targets to be observed with different profit values are given, partial targets are selected from the targets to be scheduled, and an imaging scheduling scheme is generated, so that observation profit is maximized, and meanwhile, the satellite capacity constraint is met. The imaging scheduling scheme is to determine the observation start time of each scheduled observation target. For distinction, one observation activity, in which the observation start time is determined, is defined as a task.
Generally, the traditional EOS only has the side sway capability, the satellite can only shoot images above the object to be observed, and the posture conversion time between any two objects can be calculated in advance, so that the problem of EOS scheduling is relatively simple, and the scheduling task can be efficiently completed. But this also means that the EOS system has limited observability for a plurality of closely distributed observables. In contrast, as a new generation earth observation system, the Agile Earth Observation Satellite (AEOSs) has the maneuvering capability of rolling, pitching and yawing at three angles, so that the satellite can shoot images before or after reaching the upper air of the targets, as shown in fig. 1, each target can be observed in a longer time window, therefore, the observation capability of the satellite for a plurality of targets which are closely distributed is greatly improved, and more tasks can be observed in a given time. At the same time, the agile motor capability causes the conversion time between targets and the observation starting time of the targets to be not fixed, which greatly improves the problem space and complexity of the scheduling. Furthermore, the required pose transition time between two successive observation tasks depends on the observation pose angles of the two tasks, i.e. the observation start time, also referred to as the time-dependent transition time. This time-dependent nature also places higher performance demands on the task scheduling algorithm.
Most of the existing agile satellite mission planning algorithms ignore the time dependence characteristic of the conversion time, or lack clear modeling and analysis on the time dependence characteristic, so that the problem of low scheduling efficiency is caused.
Disclosure of Invention
The invention aims to solve the technical problem of improving the scheduling efficiency of agile satellites, and provides an agile satellite scheduling method considering time-dependent conversion time.
In order to solve the problem, the invention adopts the following technical scheme:
An agile satellite scheduling method considering time-dependent transition time, comprising the steps of:
step 1: acquiring a target sequence to be observed and preprocessing the target sequence;
Step 2: constructing an agile satellite scheduling model of time-dependent conversion time;
Step 3: inputting the preprocessed target sequence to be observed into an agile satellite scheduling model, and solving the agile scheduling model;
step 4: and (3) outputting the solving result in the step (3) to obtain a scheduling scheme of the agile satellite.
Further, the construction method of the agile satellite scheduling model comprises the following steps:
The objective function is:
representing maximizing the overall benefit of scheduling tasks,
Wherein: for the decision variable, 1 represents that the ith target to be observed is scheduled on circle k, otherwise, 0, p i represents the profit of the ith target to be observed, T represents the set of the targets to be observed, and O represents the circle set in the scheduling period;
The constraint conditions are as follows:
Constraint (2) represents a multi-turn uniqueness constraint, namely each observation target can only be observed once at most in all turns;
constraint (3) represents a balance constraint, connecting decision variables And
For decision variables, 1 represents that i observation targets are observed and then j observation targets are observed, otherwise 0, i, j are observation target indexes, i, j epsilon T { s, e }, wherein { s, e } refers to virtual starting targets and virtual ending targets;
constraints (4) and (5) specify that the scheduling scheme starts at virtual target s and ends at virtual target e;
Constraint (6) specifies that the time interval between successive observations is not less than the transition time length, i k represents the observation task for observation target i at circle k, For an integer decision variable, representing the observation start time of the observation target i at the turn k, d i represents the observation duration of the observation target i,Representing the minimum pose transition time, i.e. the observation start time for a given observation target i at circle number kThe gesture conversion time corresponding to the earliest arrival time of the next observation target j, M represents the largest positive integer;
constraint (7) represents a visible time window constraint, i.e. the observation start time of each task must be within its visible time window; indicating the start time of object i in the visible time window of turn k, Indicating the end time of the visible time window of object i at turn k,Target i is in the visible time window time range of circle k;
Constraint (8) means that the observation target can only be scheduled on the circle with its visible time window; for binary parameters, when 1, the object i has a visible time window in circle k, otherwise, the object i is 0;
Constraint (9) represents the range of values of the decision variables in the model.
Further, the method for solving the agile satellite scheduling model in the step 3 is a greedy random iterative local search heuristic algorithm.
Further, the greedy random iterative local search heuristic algorithm comprises the following specific steps:
Step 3.1: any pair of visible time windows VTW j-1、VTWj of the same circle are pre-calculated, and each observation starting time on the previous visible time window is opposite to each observation starting time The earliest start time es j on the latter time window, and with respect to each observation start time on the latter visible time windowThe latest start time ls j-1 on the previous time window;
step 3.2: setting an initial value of a random factor Greed in a greedy random adaptive search algorithm to STARTGREED, and initializing a current solution to be null;
Step 3.3: initializing the iteration times of the inner loop, and initializing each element of disturbance factor parameter vectors S d and R d to be 1;
Step 3.4: invoking a disturbance operator, deleting R d (k) continuous scheduled tasks from a position S d (k) in a task sequence of a circle k from a current solution, wherein S d (k) represents a kth element of a parameter vector S d, namely, a starting position of a removed subsequence in the scheduled task sequence of the circle k, and R d (k) represents a kth element of the parameter vector R d, namely, a size of the removed subsequence of the circle k;
Step 3.5: invoking an insertion operator, sequentially inserting unscheduled objects to be observed into the current solution until the current solution has no feasible insertion, calculating all feasible insertion positions of all visible objects to be observed i on each circle k in the scheduled task sequence of the current circle k, and storing the positions In (1), ifOnly the insertion position where the cost is the smallest is reserved and stored in L k. The L k is used for storing feasible insertion of all unscheduled targets on the circle k, wherein each unscheduled target has at most one feasible insertion;
Step 3.6: sorting all the inserts in the L k from large to small according to the income of the corresponding observation target, only reserving the first (1-Greed) |L k | inserts, randomly selecting one insert by adopting a roulette manner, executing the insert into the current solution, updating the earliest starting time of each task after the insert position and the latest starting time of each task before the insert position, and returning to the step 3.5 until any feasible insert cannot be found in all circles;
step 3.7: calculating the total gain of the newly generated solution, if the total gain of the newly generated solution is larger than the current best solution, taking the newly generated solution as the new current best solution, setting the number of internal loop iteration to be zero, if the newly generated solution is worse than the current best solution, adding 1 to the number of internal loop iteration, if the number of the internal loop iteration exceeds the maximum continuous iteration step number, jumping to the step 3.8, otherwise returning to the step 3.4, and enabling the disturbance factor parameter vector S d＝Sd+Rd,Rd＝Rd +1;
Step 3.8: let the random factor Greed = Greed-GREEDDECREASE, GREEDDECREASE be the random factor step size, if Greed is still greater than or equal to the random factor minimum STARTGREED-GREEDRANGE, GREEDRANGE as the random factor fluctuation range, the current best solution is set as the current solution, and step 3.3 is returned, otherwise, the current best solution is output.
Further, the insertion operator is:
step 3.5.1: for each circle k of the current solution, if the current circle k has no scheduled task, when only one insertion position exists, setting the earliest starting time corresponding to the insertion position as the visible time window starting time of the unscheduled target i to be inserted, and setting the latest starting time as the visible time window ending time;
step 3.5.2: when attempting to insert an unscheduled object i into each position of the current sequence,
If the insertion position is the sequence head, making the earliest start time es i of the insertion be the visible time window start time of the unscheduled target i to be inserted;
If the insertion position is the sequence tail, making the latest start time ls i of the insertion be the visible time window end time of the unscheduled target i to be inserted;
If the insertion position is between j and j+1, calculating the earliest starting time es i corresponding to the insertion of the object i to be observed into the position according to the earliest starting time es j of the task j before the position to be inserted in the current sequence, and calculating the latest starting time ls i corresponding to the insertion of the object i to be observed into the position according to the latest starting time ls j+1 of the task j+1 after the position to be inserted in the current sequence, if es i＜lsi, the insertion is feasible, otherwise, the insertion is not feasible.
Further, before step 3.1, the method further comprises the following steps of
Step 3.1' precalculating any time window over any number of turns k"Unreachable" earliest time vector of (a)And "unreachable" latest time vectorWherein the method comprises the steps ofRepresenting slaveCan not arrive at the departureCorresponding to whenIs set to be the earliest observation start time of (c),Is shown inAfter observing object i internally, j cannot be scheduled atThe latest observation start time in the inner.
Then step 3.5.1 'is also included before step 3.5.1':
If an attempt is made to insert the observation target i between the scheduled tasks j and j +1,
If the observation start time t j of j is not earlier than the unreachable earliest time corresponding to i, i.e., t j＜uej (i), then the position is not pluggable and the position after j+1 is also not pluggable, without any further insertion attempt;
If the observation start time of j+1 is not later than the unreachable latest time corresponding to i, namely t j+1＜ulj+1 (i), the position is not insertable, and the position before j is also not insertable, and no insertion attempt is needed;
If the insertion position meets the requirement that the earliest and latest times are not reached, step 3.5.1 is shifted.
Further, the cost in step 3.3 means:
costi＝transji(esj,esi)+transi(j+1)(esi,es'j+1)-transj(j+1)(esj,esj+1)
where es j represents the earliest start time of task j, es i represents the earliest start time of object i to be observed, es' j+1 represents the earliest start time of task j+1 after updating, i is the object to be observed, and cost i represents the cost of inserting object i to be observed between positions j and j+1.
Further, step 3.5 further includes, after performing the insertion operation once for all the rounds, checking whether a certain target is scheduled to the task sequence of the plurality of rounds, if so, performing an allocation step, allocating the target to the task of one of the rounds according to a predetermined allocation policy, and deleting the scheduled tasks corresponding to the remaining rounds.
Further, the preset allocation policy refers to that the target to be observed is preferentially allocated to the circle with a smaller total time window number.
The invention also provides an agile satellite scheduling system considering time-dependent switching time, which comprises a processor and a memory connected with the processor, wherein the memory stores a program of an agile satellite scheduling method considering time-dependent switching time, and the program realizes the steps of the method when being executed.
Compared with the prior art, the invention has the beneficial effects that:
According to the agile satellite scheduling method considering time dependence and conversion time, the time dependence of gesture conversion time among continuous observation tasks is considered, so that the constructed agile satellite scheduling model can better meet scheduling requirements, and meanwhile, the scheduling scheme obtained by solving the model is higher in scheduling efficiency.
When the model is solved, after the conversion time meets the first-in first-out rule and the inequality rule, the observation starting time of each pair of observation windows in the same circle relative to the previous visible time window is calculated in advanceThe earliest start time es j on the latter time window, and with respect to each observation start time on the latter visible time windowThe latest starting time ls j-1 on the previous time window reduces repeated calculation in the dispatching solving process when the inserting operation is attempted, greatly reduces the calculation time and improves the dispatching efficiency.
By pre-calculating the unreachable earliest time and the unreachable latest time of each time window and comparing the starting time of the object to be observed with the unreachable earliest and latest time when the object to be observed tries to be inserted, a plurality of unreachable try to insert operations can be screened in advance, the calculation time is greatly reduced, and the scheduling efficiency is improved.
By using perturbation operators, the placement algorithm can be trapped in local optima, thus better searching the entire solution space.
Drawings
FIG. 1 is a schematic diagram of the operation of agile satellite scheduling;
FIG. 2 is a flow chart of the system of the present invention;
FIG. 3 shows the transition time first-in first-out rule schematic diagram;
FIG. 4 is a schematic diagram of minimum gesture transition times;
FIG. 5 is a flow chart of a greedy random iterative local search heuristic;
FIG. 6 is a schematic diagram of an insert operator;
fig. 7 is a schematic diagram of unreachable earliest and latest times.
Detailed Description
Fig. 1 to 7 show a specific embodiment of the agile satellite scheduling method according to the present invention, which considers time-dependent transition time, comprising the following steps, as shown in fig. 2:
step 1: acquiring a target sequence to be observed and preprocessing the target sequence;
Step 2: constructing an agile satellite scheduling model of time-dependent conversion time;
The agile satellite scheduling model constructed in this embodiment is based on the conversion time satisfying the first-in first-out rule and the triangle inequality rule.
(1) First-in first-out rule (FIFO rule)
The first-in first-out rule refers to that the arrival time of the next task corresponding to t j is not earlier than the arrival time corresponding to t i after the attitude conversion by assuming that the satellite has two different observation start times t i and t j(ti＜tj for the previous observation task. As shown in (a) and (b) of fig. 3, assuming that t i and t j are two observation start times of time windows i and j, respectively, the transition time is trans ij(ti,tj), the observation time Δt of the previous window is delayed to be equal to or greater than 0, the corresponding transition time is trans ij(ti+Δt,tj), and the calculation formula of the transition time is that
Wherein the amount of change in attitude
Representing the yaw, pitch and yaw angles, respectively, of window i at time t i.
Then the following proposition exists:
proposition 1: the time dependent transition time satisfies a first-in first-out rule, if and only if
And (3) proving: (. Fwdarw.necessity)
If trans ij(ti,tj) satisfies the first-in first-out rule, then
Furthermore, there are
(. Fwdarw.sufficiency)
According to the above conversion time and attitude change angle calculation formula, and the known attitude angle is continuously changed with time, it is known that trans ij(ti,tj) is continuously changed between the range [ t i,ti +Δt ]. From the median theorem, it is known that there is at least one point in [ t i,ti +Δt ]So that
If it isFor all [ t i,ti ] in +DeltatAll are true, there are
ti+transij(ti,tj)≤ti+Δt+transij(ti+Δt,tj)
I.e. the first-in first-out rule is satisfied.
According to the proposition, only the verification is neededThe transition time can be proved to be in accordance with the first-in first-out rule. Can not be provided with Which represents the change in conversion time per unit time for window i corresponding to time t i at a fixed t j. The conversion time and the attitude change amount are used for calculating a formula, and for the currently used semi-agile satellite, the yaw angle is unchanged when observation is performed, and the yaw angle is close to zero, so that the method can be obtained:
Where g i(ti) represents the instantaneous attitude angle of window i at time t i and v represents the instantaneous angular velocity. Since as a scheduling input the attitude angle is given in discrete form in seconds, the above equation can be translated into:
Where v min represents the minimum angular velocity in { v 1,v2,v3,v4 }, and κ| upper | represents the upper bound of |κ|. In practical applications, the value of i κ upper depends on the parameters of the satellite and the geographical location of the observation target. We find that given any actual satellite parameters and calculations, |κ upper | is always strictly less than 1. Therefore, the transition time always satisfies the first-in first-out rule.
Furthermore, fixing t i, varying t j, it can also be concluded that the latest start time of time window i must be later, deferring t j.
(2) Triangle inequality rule
The triangle inequality rule means that for time windows i and j, its observation start time is fixed, its transition time necessarily being less than or equal to the total transition time from i through another "detour" time window m to j. When the attitude conversion angular speed is a fixed value (not time-dependent), since |Δg ij|≤|Δgim|+|Δgmj|(Δgij represents the attitude angle variation amount from i to j, the triangle inequality rule is necessarily satisfied. However, in practical applications, the angular speed of the posture conversion depends on the amount of change of the posture angle, and the larger the amount of change, the larger the angular speed, because in general, the larger the posture angle, the faster the system will employ the maneuvering engine to implement the posture conversion. Thus, in general, v 1≤v2≤v3≤v4.
Assuming that the service time of the detour time window m is ignored, the triangle inequality rule is satisfied if the following inequality is satisfied:
trans(Δgim)+trans(Δgmj)≥trans(Δgij) (e)
Let v im,vmj,vij represent the angular velocity corresponding to |Δg ij |. Given that Δg ij＝Δgim+Δgmj is necessarily |Δg ij|≤|Δgim|+|Δgmj |, the following is divided into two values:
(1) If |Δg ij|＞|Δgmj |, there is
vim,vmj≤vij，
Substituted into formula (a), there must be
trans(Δgim)+trans(Δgmj)≥trans(Δgij)
(2) If |Δg ij|≤|Δgmj |, the conversion time function must be a monotonically increasing function, so that it is satisfied
The inequality (e) is satisfied.
Based on the first-in first-out rule and the triangle inequality rule, the agile satellite scheduling model is constructed by the following concrete steps:
The objective function is:
representing maximizing the overall benefit of scheduling tasks,
Wherein: For the decision variable, 1 represents that the ith target to be observed is scheduled on circle k, otherwise, 0, p i represents the profit of the ith target, T represents the set of the observed targets, and O represents the circle set in the scheduling period;
The constraint conditions are as follows:
Constraint (2) represents a multi-turn uniqueness constraint, namely each observation target can only be observed once at most in all turns;
constraint (3) represents a balance constraint, connecting decision variables And
For decision variables, 1 represents that i observation targets are observed and then j observation targets are observed, otherwise 0, i, j are observation target indexes, i, j epsilon T { s, e }, wherein { s, e } refers to virtual starting targets and virtual ending targets;
constraints (4) and (5) specify that the scheduling scheme starts at virtual target s and ends at virtual target e;
Constraint (6) specifies that the time interval between successive observations is not less than the transition time length, i k represents the observation task for observation target i at circle k, For an integer decision variable, representing the observation start time of the observation target i at the turn k, d i represents the observation duration of the observation target i,Representing the minimum pose transition time, i.e. the observation start time for a given observation target i at circle number kThe gesture conversion time corresponding to the earliest arrival time of the next observation target j, M represents the largest positive integer;
In this embodiment, when the time-dependent conversion time satisfies the first-in first-out rule, a "minimum gesture conversion time" is provided to replace the actual conversion time in model solution. As shown in fig. 4, given two time windows i and j, the observation start time t i of i is fixed, while the observation start time t j of j is variable, the earliest arrival time of the window j can be always found according to the first-in first-out rule, and the corresponding transition time is the minimum gesture transition time, and is marked as mintrans ij(ti). It follows that the minimum posture change time depends only on the start time of the previous observation task, and is independent of the observation start time of the subsequent task. The minimum gesture conversion time is adopted to replace the actual conversion time, so that the problem dimension is reduced, and the calculation complexity is simplified. In addition, in order to reduce the minimum posture transition time calculated repeatedly by the scheduling algorithm, the minimum posture transition time of each second observation start time relative to the previous window can be calculated by preprocessing for any two time windows.
Constraint (7) represents a visible time window constraint, i.e. the observation start time of each task must be within its visible time window; indicating the start time of object i in the visible time window of turn k, Indicating the end time of the visible time window of object i at turn k,Target i is in the visible time window time range of circle k;
Constraint (8) means that the observation target can only be scheduled on the circle with its visible time window; for binary parameters, when 1, the object i has a visible time window in circle k, otherwise, the object i is 0;
Constraint (9) represents the range of values of the decision variables in the model.
Step 3: inputting the preprocessed task sequence to be observed into an agile satellite scheduling model, and solving the agile scheduling model;
The method for solving the agile satellite scheduling model in this embodiment is a greedy random iterative local search heuristic (Greedy Randomized Iterated Local Search, GRILS). Is a combination of greedy random adaptive search algorithm (Greedy Randomized ADAPTIVE SEARCH process, GRASP) and iterative local search algorithm (Iterated Local Search, ILS). The method proves that the method can obtain good solving effect on the team oriented problem with time window of multi-resource constraint, and the problem has high similarity with the model of the agile satellite scheduling problem. The algorithm is modified to a certain extent when the algorithm is adopted, so that the algorithm is used for adapting to the problem characteristics of agile satellite scheduling. As shown in fig. 5:
The algorithm may be divided into two levels of loops. The inner loop corresponds to an ILS algorithm, and is used for specifically scheduling the scheduling access of the observation task, and is a core part of the overall algorithm. The outer loop controls the value of a random factor parameter Greed (0 < green < 1) as an input to the inner loop for controlling the randomness of the ILS algorithm. In the inner loop, ILS is composed of two operators: an INSERT operator (INSERT) and a perturbation operator (SHAKE). For each iteration of the ILS, the INSERT operator INSERTs the unscheduled task into the current solution (i.e., the scheduled task sequence) in turn until no more insertion is possible. The operator is controlled by Greed parameters, the larger Greed is, the more random the INSERT operator is. For example, assuming that there are 10 possible insertions (tasks to be inserted and positions of insertions are selected) for the current solution, the insertions are ordered in a profit-to-profit order, then only the previous (1-Greed). 10 insertions are considered, and one of the insertions is randomly selected for execution in the manner of roulette. The SHAKE operator removes a part of scheduling tasks from the current solution, so that the search is prevented from being trapped in local optimization, and the diversity of the search is improved. After the INSERT operator is performed, if the newly generated solution is better than the current best solution, it is recorded as the new current best solution. If the current best solution cannot be lifted within a certain number of continuous iteration steps, ending the inner ILS loop. During the outer layer cycle, the random factor Greed is stepped down (STARTGREED-GREEDRANGE) starting from parameter STARTGREED in GREEDDECREASE steps. At each iteration step of the outer loop, the ILS algorithm begins solving with the current best solution as the starting point. The detailed steps are as follows:
step 3.1' precalculating any time window over any number of turns k "Unreachable" earliest time vector of (a)And "unreachable" latest time vectorWherein the method comprises the steps ofRepresenting slaveCan not arrive at the departureCorresponding to whenIs set to be the earliest observation start time of (c),Is shown inAfter the internal observation target i, the j observation target cannot be scheduled atA latest observation start time within;
to reduce computation time, as shown in FIG. 7, for a certain time window over a certain number of turns k Calculating the "unreachable" earliest time vectorAnd "unreachable" latest time vectorBy employing such preprocessing, it is necessary to satisfy when inserting task j k after (before) attempting to scheduled task i k Therefore, a plurality of infeasible trial insertion operations can be screened out in advance, and the calculation time is greatly reduced.A visible time window representing the ith observation target over circle k,Representing the visible time window of the jth observed object over circle k.
Step 3.1: pre-calculating and storing each observation start time in the previous visible time window on any pair of visible time windows VTW j-1、VTWj of the same circleThe earliest start time es j on the latter time window, and with respect to each observation start time on the latter visible time windowThe latest start time ls j-1 on the previous time window. In the present embodiment, the earliest start time vector in the next visible time window is calculated and stored in advance with respect to the time of every second in the previous visible time window as the observation start time, in the case where the minimum posture change time is satisfied. By such preprocessing, when solving the scheduling model, the earliest start time of task observation in the next visible time window corresponding to the start time of observation can be obtained by querying as long as the start time of observation of the task in the previous visible time window is given.
The minimum gesture conversion time is obtained based on the first-in first-out rule, and is only dependent on the starting time of the previous observation task, and the calculation method of the minimum gesture conversion time is obtained by solving through a binary search method, specifically, the previous time window is fixedIs of (2)Detecting the latter time windowVisible time window start time of (2)If it isMeeting the conversion time constraint, orderOtherwise, detecting the time window ending timeIf the conversion time constraint is not satisfied, then representAfter observing the object i at the moment, the object j cannot be observed any more. If the constraint is satisfied, thenMust be within a feasible intervalWithin, using binary search checkingIf the midpoint of the (c) can meet the constraint, dividing the feasible interval into half of the original interval, and continuing the binary search until the earliest starting time is found, wherein the corresponding conversion time is the minimum gesture conversion time. This calculation process is marked asLikewise, the latter observation is fixedA similar method can also be used to calculate the latest start time of the previous observation, labeled asThe method comprises the following steps: fix the latter observationDetecting a previous time windowVisible time window start time of (2)If it isMeeting the conversion time constraint, orderOtherwise, detecting the time window ending timeIf the conversion time constraint is not satisfied, then representAfter observing the object i at the moment, the object j cannot be observed any more, and if the constraint is satisfied, thenMust be at the latest observation time of (2)Within, continue checkingIf the midpoint of (2) can meet the conversion time constraint and divide the feasible interval into half of the original, continue the binary search until the latest start time is found.
Step 3.2: setting an initial value of a random factor Greed in a greedy random iterative local search algorithm to STARTGREED, and initializing a current solution to be null;
Step 3.3: initializing the iteration times of the inner loop, and initializing disturbance factor parameter vectors S d and R d;
Step 3.4: invoking a disturbance operator, deleting R d (k) continuous scheduled tasks from a position S d (k) in a task sequence of a circle k from a current solution, wherein S d (k) represents a kth element of a parameter vector S d, namely, a starting position of a removed subsequence in the scheduled task sequence of the circle k, and R d (k) represents a kth element of the parameter vector R d, namely, a size of the removed subsequence of the circle k;
The perturbation operator is the scheduled task that removes a portion of the current solution, in this embodiment two parameter vectors S d and R d are defined, and at the beginning of the inner loop, the elements of S d and R d are initialized to 1. When the disturbance algorithm is called, R d (k) continuous scheduling tasks are deleted from the position S d (k) of the scheduling sequence of each circle k, and if the task exceeding the last position is deleted, the deletion is continued from the first task. Let S d(k)＝Sd(k)+Rd(k),Rd(k)＝Rd (k) +1, the next time the operator is invoked, the new parameter value is used. If S d (k) is greater than the number of tasks for the sequence of circle k, then the number of tasks is subtracted to return to the position forward of the sequence. If R d is greater than the number of tasks Let R d =1. This operator ensures that each scheduled task can be deleted at least once throughout the search. The perturbation operator is mainly used in this embodiment to prevent the algorithm from falling into a local optimum, so that the whole solution space is better searched. Each time the operator is invoked, the algorithm removes a removal subsequence consisting of a number of consecutive tasks from the sequence of scheduled tasks for each round, respectively. The operator defines two parameters: s d indicates the starting position of the removal sub-sequence in the currently scheduled task sequence; r d represents the size of the removed subsequence. The values of these two parameters are different on different rounds because the size of the scheduled sequence varies for each round. To guarantee diversity of searches, a larger-scale sequence of scheduled tasks corresponds to a larger R d.
Step 3.5: invoking an insertion operator, sequentially inserting the unscheduled tasks to be observed into a scheduled task sequence which is best solved on the circle k at present, and recording all feasible insertion and storage of each unscheduled task i on the circle kIn (1), ifOnly the least costly insert is reserved and stored in L k, which L k is used to store all possible inserts on round k, with at most one possible insert per unscheduled task. The feasible insertion in this embodiment is to record the feasible positions of inserting a target to be observed into the target to be observed at a certain turn. In this embodiment, the cost calculation method is as follows:
costi＝transji(esj,esi)+transi(j+1)(esi,es'j+1)-transj(j+1)(esj,esj+1)
where es j represents the earliest start time of task j, es i represents the earliest start time of object i to be observed, es' j+1 represents the earliest start time of task j+1 after updating, i is the object to be observed, and cost i represents the cost of inserting object i to be observed between positions j and j+1.
The specific insertion method of the insertion operator in this embodiment is as follows:
Step 3.5.1':
If an attempt is made to insert the observation target i between the scheduled tasks j and j +1,
If the observation start time t j of j is not earlier than the unreachable earliest time corresponding to i, i.e., t j＜uej (i), then the position is not pluggable and the position after j+1 is also not pluggable, without any further insertion attempt;
If the observation start time of j+1 is not later than the unreachable latest time corresponding to i, namely t j+1＜ulj+1 (i), the position is not insertable, and the position before j is also not insertable, and no insertion attempt is needed;
If the insertion position meets the requirement that the earliest and latest times are not reached, step 3.5.1 is shifted.
Step 3.5.1: for each circle k of the current solution, if the current circle k has no scheduled task, when only one insertion position exists, setting the earliest starting time corresponding to the insertion position as the visible time window starting time of the unscheduled target i to be inserted, and setting the latest starting time as the visible time window ending time;
step 3.5.2: when attempting to insert an unscheduled object i into each position of the current sequence,
If the insertion position is the sequence head, making the earliest start time es i of the insertion be the visible time window start time of the unscheduled target i to be inserted;
If the insertion position is the sequence tail, making the latest start time ls i of the insertion be the visible time window end time of the unscheduled target i to be inserted;
If the insertion position is between j and j+1, calculating the earliest starting time es i corresponding to the insertion of the object i to be observed into the position according to the earliest starting time es j of the task j before the position to be inserted in the current sequence, and calculating the latest starting time ls i corresponding to the insertion of the object i to be observed into the position according to the latest starting time ls j+1 of the task j+1 after the position to be inserted in the current sequence, if es i＜lsi, the insertion is feasible, otherwise, the insertion is not feasible.
The calculation method of the earliest start time es j comprises the following steps:
Given a sequence of scheduled tasks (j-1) -j- (j+1) - (j+2) over a certain turn, ignoring the superscript k, calculating the earliest start time of each task j sequentially from front to back by es j＝EarliestStartTime(j-1)j(esj-1), wherein the earliest start time of the forefront scheduled task j-1 is set to its window start time st j-1, the earliest start time of an observed task over the next visible time window has been calculated as the observation start time every second over the previous visible time window already at step 3.1, thus upon attempted insertion, the earliest start time of a task over the next visible time window is obtained from the start time es j-1 of the given previous visible window task, es j. Also, the latest start time ls j: on the basis of the pre-calculation, the latest start time of each task is calculated sequentially from back to front by ls j＝LastestStartTimej(j+1)(lsj+1), the latest start time of the last task (j+2) is set to (et j+2-dj+2),etj+2 is the window end time of task j+2, and d j+2 is the observation duration of task j+2), so that the latest start time of the last visible time window is given, and the latest start time of the previous visible time window can be obtained.
In this embodiment, the insertion operator selects a part of non-scheduling targets, inserts the non-scheduling targets into the designated positions of the current solution according to the schedulable window of the non-scheduling targets, and ensures the feasibility of the current solution after the insertion. The solution defined in this embodiment is composed of several sequences of scheduled tasks, each corresponding to a round-robin scheduling scheme. When inserting a certain observation task into the current solution, the subsequent task after the insertion position may have to reschedule or postpone the observation time due to the influence of the constraint of the transition time, thereby ensuring the feasibility of the solution. The transition time between these subsequent tasks needs to be updated one by one to detect if it violates the visible time window constraint. Because of the existence of the visible time window constraint, a large number of insertion attempts are not feasible, and the larger the size of the scheduled sequence, the more times the conversion time needs to be updated and checked per insertion, which occupies a large amount of computing resources, and therefore, the "full" feasibility detection cannot meet the requirement of efficient solution. To reduce this large amount of computing resources, the insertion operator designed in this embodiment is a fast insertion method for time window constraints and transition time constraints. By calculating the earliest and latest start times of each task, any time point in [ es j,lsj ] can be used as the observation start time for any task j in the sequence according to the first-in first-out rule, and the feasibility of the whole sequence is not affected. When the observation start time of any one task in the sequence is determined, the observation start time of the rest tasks in the sequence can be determined through a backtracking method. In fig. 6, when a new task i is attempted to be inserted into a position between j and (j+1) of the current sequence, the earliest start time es i and the latest start time ls i corresponding to the insertion of the task i into the position are calculated from es i＝EarliestStartTimeji(esj) and ls i＝LastestStartTimei(j+1)(lsj+1), respectively. If es i≤lsi, the insertion is possible, otherwise, the insertion of the position would cause the remaining scheduled tasks to violate the window constraint. By adopting the rapid insertion method, the conversion time of the rest scheduled tasks of the current sequence does not need to be recalculated, so that the time for detecting the insertion feasibility is greatly reduced. It should be noted, however, that after each successful insertion of a task, the earliest start times of all subsequent tasks after the insertion location, and the latest start times of all preceding tasks before the insertion location, need to be updated.
By the operator rapid insertion method, the feasibility of any insertion position of all unscheduled tasks in the current solution can be efficiently evaluated, and all feasible insertion can be screened out. By performing such an insertion operation for all turns, there may be some goal in the current solution that is scheduled over multiple turns, i.e., that violates the multi-turn uniqueness constraint. To ensure the feasibility of the solution, an allocation operation (marked Assignment ()) is defined, which is allocated to only one of the feasible turns for the repeatedly scheduled observation target according to a preset rule. According to preliminary experiments, a flexible distribution strategy is adopted, so that a good distribution effect can be obtained. The flexible allocation strategy refers to preferentially allocating the object to be observed to the circle with less total time window. This is because tasks are preferentially allocated to the number of turns with a small number of time windows, and then there is enough space to accommodate more observation tasks with a larger number of turns. After the allocation operation is performed, an insertion of a new task is retried and the next cycle is entered. The insertion operator terminates when there are no viable inserts available.
Step 3.6: sorting all the inserts in the L k from large to small according to the income of the corresponding observation target, only reserving the first (1-Greed) |L k | inserts, randomly selecting one insert by adopting a roulette manner, executing the insert into the current solution, updating the earliest starting time of each task after the insert position and the latest starting time of each task before the insert position, and returning to the step 3.5 until any feasible insert cannot be found in all circles;
The larger the outer loop's random factor parameter Greed (0 < green < 1), the more random the insertion operator will be, assuming that there are 10 possible insertions (selected tasks to be inserted and insertion positions) for the current solution, ordered in a profit-big-small order, then consider only the previous (1-Greed) 10 insertions, and then randomly select one from these insertions to perform the insertion in the manner of a roulette wheel. After each successful insertion of a task, the earliest start time of all subsequent tasks after the insertion location, and the latest start time of all preceding tasks before the insertion location, need to be updated.
Step 3.7: calculating the total gain of the newly generated solution, if the total gain of the newly generated solution is larger than the current best solution, taking the newly generated solution as the new current best solution, setting the number of internal loop iteration to be zero, if the newly generated solution is worse than the current best solution, adding 1 to the number of internal loop iteration, if the number of the internal loop iteration exceeds the maximum continuous iteration step number, jumping to the step 3.8, otherwise returning to the step 3.4, and enabling the disturbance factor parameter vector S d＝Sd+Rd,Rd＝Rd +1;
Step 3.8: let the random factor Greed = Greed-GREEDDECREASE, GREEDDECREASE be the random factor step size, if Greed is still greater than or equal to the random factor minimum STARTGREED-GREEDRANGE, GREEDRANGE, which is the random factor fluctuation range, the current best solution is set as the current solution, and step 3.2 is returned, otherwise, the current best solution is output.
Step 4: and (3) outputting the solving result in the step (3) to obtain a scheduling scheme of the agile satellite.
The invention also provides an agile satellite scheduling system considering time-dependent switching time, which comprises a processor and a memory connected with the processor, wherein the memory stores a program of an agile satellite scheduling method considering time-dependent switching time, and the program realizes the steps of the method when being executed.
The effectiveness of the method of the invention is verified experimentally as follows:
(1) Experimental example and algorithm parameters
Experiments mainly compare the single star Algorithm (ALNS) and the multi-star algorithm (A-ALNS) in the latest research literature, and the two algorithms mainly focus on reconnaissance activities in two areas: chinese area (3°n-53°n) and global area (74°e-133°e). The example assumes that the observation targets are uniformly randomly generated in these areas. As preliminary experiments show that for global experiments, both a comparison algorithm and the algorithm can schedule almost all observation targets, the global area is too simple to reflect the algorithm difference and is not in the consideration range of the comparison experiment. In the example, the benefits and observation durations of all the observation targets are randomly generated (in seconds) following the uniform distribution of [1,10] and [15,30], and the scheduling period is set to 24 hours.
The algorithm involves 4 input parameters, STARTGREED, GREEDRANGE, GREEDDECREASE each, and the maximum number of consecutive non-increasing iteration steps. Preliminary experiments have shown that algorithms achieve better results when STARTGREED =1, greenrange=0.2, greenDeclean=0.02. For the inner layer ILS loop, the inner layer ILS has been able to converge smoothly to a certain stable value when the maximum number of consecutive non-increasing iteration steps is set to 300. The experiment was run on a machine with an intel core22.5ghz processor and 8GB memory. For each example, the results of the algorithm were taken as the average of 5 independent replicates
(2) Single star example comparison
Table 1 comparison of single star calculation experiments in China
Table 1 shows the comparison between the present algorithm and the best algorithm ALNS in the literature on the single star scheduled Chinese regional example. Wherein, P t represents the total income of all objects to be observed in the current computing example, P s represents the income value of the scheduling scheme of the algorithm, and Outper represents the percentage of the solving quality of the algorithm superior to ALNS in the current computing example. The results in table 1 show that our algorithm GRILS has a scheduling scheme that is 52% higher than the average of the ALNS algorithm. In addition, the difference in the solution quality increases as the scale of the calculation increases. For example, the figure of merit for GRILS is nearly twice ALNS for the example with a target number of 400, 500, 600. This is a good demonstration that the present algorithm is superior to ALNS algorithms, especially on large scale. In addition, the algorithm takes several times less solution time than ALNS.
Table 2 comparison of multiple-star calculation experiments in China area
Table 2 shows a comparative experiment between the present invention and the literature best algorithm A-ALNS on the multi-star scheduling example in China. The calculation example is divided into two parts according to the size of the scale: 100-500 medium-scale modules and 600-1000 large-scale modules. The comparison shows that the method GRILS of the invention is much better than the A-ALNS algorithm in the multi-star example, the average scheduling benefit value is 30% higher, and the calculation time of the algorithm is much less than that of the A-ALNS.
The above is only a preferred embodiment of the present invention, and the protection scope of the present invention is not limited to the above examples, and all technical solutions belonging to the concept of the present invention belong to the protection scope of the present invention. It should be noted that modifications and adaptations to the invention without departing from the principles thereof are intended to be within the scope of the invention as set forth in the following claims.
Claims (9)
1. The agile satellite scheduling method considering time-dependent conversion time is characterized by comprising the following steps of:
step 1: acquiring a target sequence to be observed and preprocessing the target sequence;
Step 2: constructing an agile satellite scheduling model of time-dependent conversion time;
Step 3: inputting the preprocessed target sequence to be observed into an agile satellite scheduling model, and solving the agile scheduling model;
step 4: outputting the solving result in the step 3 to obtain a scheduling scheme of the agile satellite;
the construction method of the agile satellite scheduling model comprises the following steps:
The objective function is:
representing maximizing the overall benefit of scheduling tasks,
Wherein: for the decision variable, 1 represents that the ith target to be observed is scheduled on circle k, otherwise, 0, p i represents the profit of the ith target to be observed, T represents the set of the targets to be observed, and O represents the circle set in the scheduling period;
The constraint conditions are as follows:
Constraint (2) represents a multi-turn uniqueness constraint, namely each observation target can only be observed once at most in all turns;
constraint (3) represents a balance constraint, connecting decision variables And
For decision variables, 1 represents that i observation targets are observed and then j observation targets are observed, otherwise 0, i, j are observation target indexes, i, j epsilon T { s, e }, wherein { s, e } refers to virtual starting targets and virtual ending targets;
constraints (4) and (5) specify that the scheduling scheme starts at virtual target s and ends at virtual target e;
Constraint (6) specifies that the time interval between successive observations is not less than the transition time length, i k represents the observation task for observation target i at circle k, For an integer decision variable, representing the observation start time of the observation target i at the turn k, d i represents the observation duration of the observation target i,Representing the minimum pose transition time, i.e. the observation start time for a given observation target i at circle number kThe gesture conversion time corresponding to the earliest arrival time of the next observation target j, M represents the largest positive integer;
constraint (7) represents a visible time window constraint, i.e. the observation start time of each task must be within its visible time window; indicating the start time of object i in the visible time window of turn k, Indicating the end time of the visible time window of object i at turn k,Target i is in the visible time window time range of circle k;
Constraint (8) means that the observation target can only be scheduled on the circle with its visible time window; for binary parameters, when 1, the object i has a visible time window in circle k, otherwise, the object i is 0;
Constraint (9) represents the range of values of the decision variables in the model.
2. The method of claim 1, wherein the method of solving the agile satellite scheduling model in step 3 is a greedy random iterative local search heuristic.
3. The method of claim 2, wherein the greedy random iterative local search heuristic comprises the specific steps of:
Step 3.1: any pair of visible time windows VTW j-1、VTWj of the same circle are pre-calculated, and each observation starting time on the previous visible time window is opposite to each observation starting time The earliest start time es j on the latter time window, and with respect to each observation start time on the latter visible time windowThe latest start time ls j-1 on the previous time window;
Step 3.2: setting an initial value of a random factor Greed in a greedy random iterative local search algorithm to STARTGREED, and initializing a current solution to be null;
Step 3.3: initializing the iteration times of the inner loop, and initializing disturbance factor parameter vectors S d and R d;
Step 3.4: invoking a disturbance operator, deleting R d (k) continuous scheduled tasks from a position S d (k) in a task sequence of which the current solution is positioned in the circle k, wherein S d (k) represents the kth element of the parameter vector S d, namely, the starting position of a removed subsequence in the scheduled task sequence of the circle k, and R d (k) represents the kth element of the parameter vector R d, namely, the size of the removed subsequence of the circle k;
Step 3.5: invoking an insertion operator, sequentially inserting unscheduled objects to be observed into the current solution until the current solution has no feasible insertion, calculating all feasible insertion positions of all visible objects to be observed i on each circle k in the scheduled task sequence of the current circle k, and storing the positions In (1), ifThen only the insertion position with the minimum cost is reserved and stored in the L k, where the L k is used to store the feasible insertion of all the unscheduled objects on the round k, where each unscheduled object has at most one feasible insertion;
Step 3.6: sorting all the inserts in the L k from large to small according to the income of the corresponding observation target, only reserving the first (1-Greed) |L k | inserts, randomly selecting one insert by adopting a roulette manner, executing the insert into the current solution, updating the earliest starting time of each task after the insert position and the latest starting time of each task before the insert position, and returning to the step 3.5 until any feasible insert cannot be found in all circles;
step 3.7: calculating the total gain of the newly generated solution, if the total gain of the newly generated solution is larger than the current best solution, taking the newly generated solution as the new current best solution, setting the number of internal loop iteration to be zero, if the newly generated solution is worse than the current best solution, adding 1 to the number of internal loop iteration, if the number of the internal loop iteration exceeds the maximum continuous iteration step number, jumping to the step 3.8, otherwise returning to the step 3.4, and enabling the disturbance factor parameter vector S d＝Sd+Rd,Rd＝Rd +1;
Step 3.8: let the random factor Greed = Greed-GREEDDECREASE, GREEDDECREASE be the random factor step size, if Greed is still greater than or equal to the random factor minimum STARTGREED-GREEDRANGE, GREEDRANGE, which is the random factor fluctuation range, the current best solution is set as the current solution, and step 3.2 is returned, otherwise, the current best solution is output.
4. A method according to claim 3, characterized in that: the insertion operator is as follows:
step 3.5.1: for each circle k of the current solution, if the current circle k has no scheduled task, when only one insertion position exists, setting the earliest starting time corresponding to the insertion position as the visible time window starting time of the unscheduled target i to be inserted, and setting the latest starting time as the visible time window ending time;
step 3.5.2: when attempting to insert an unscheduled object i into each position of the current sequence,
If the insertion position is the sequence head, making the earliest start time es i of the insertion be the visible time window start time of the unscheduled target i to be inserted;
If the insertion position is the sequence tail, making the latest start time ls i of the insertion be the visible time window end time of the unscheduled target i to be inserted;
If the insertion position is between j and j+1, calculating the earliest starting time es i corresponding to the insertion of the object i to be observed into the position according to the earliest starting time es j of the task j before the position to be inserted in the current sequence, and calculating the latest starting time ls i corresponding to the insertion of the object i to be observed into the position according to the latest starting time ls j+1 of the task j+1 after the position to be inserted in the current sequence, if es i＜lsi, the insertion is feasible, otherwise, the insertion is not feasible.
5. The method according to claim 4, wherein: before step 3.1, further comprising
Step 3.1' pre-calculating the "unreachable" earliest time vector for any one time window VTW i k over any number of turns kAnd "unreachable" latest time vectorWherein the method comprises the steps ofRepresents the earliest observation start time of the corresponding VTW i k when the VTW j k cannot be reached from the VTW i k,Indicating a latest observation start time at which j cannot be scheduled in VTW j k after observation of object i in VTW i k;
then step 3.5.1 'is also included before step 3.5.1':
If an attempt is made to insert the observation target i between the scheduled tasks j and j +1,
If the observation start time t j of j is not earlier than the unreachable earliest time corresponding to i, i.e., t j<uej (i), then the position is not pluggable and the position after j+1 is also not pluggable, without any further insertion attempt;
If the observation start time of j+1 is not later than the unreachable latest time corresponding to i, namely t j+1<ulj+1 (i), the position is not insertable, and the position before j is also not insertable, and no insertion attempt is needed;
If the insertion position meets the requirement that the earliest and latest times are not reached, step 3.5.1 is shifted.
6. The method of claim 4, wherein the cost in step 3.5 is:
costi＝transji(esj,esi)+transi(j+1)(esi,es'j+1)-transj(j+1)(esj,esj+1)
where es j represents the earliest start time of task j, es i represents the earliest start time of object i to be observed, es' j+1 represents the earliest start time of task j+1 after updating, i is the object to be observed, and cost i represents the cost of inserting object i to be observed between positions j and j+1.
7. A method according to claim 3, characterized in that: step 3.5 further includes, after performing an insertion operation on all the rounds, checking whether a certain target is scheduled to a task sequence of a plurality of rounds, if so, performing an allocation step, allocating the target to a task of one of the rounds according to a predetermined allocation policy, and deleting scheduled tasks corresponding to the remaining rounds.
8. The method according to claim 7, wherein: the preset allocation strategy refers to that the object to be observed is preferentially allocated to the circle with a smaller total time window number.
9. An agile satellite scheduling system that takes into account time-dependent transition times, characterized by: comprising a processor and a memory connected to the processor, the memory storing a program for an agile satellite scheduling method taking into account time dependent switching times, the program when executed implementing the steps of the method according to any of the preceding claims 1-8.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010654910.7A CN111651905B (en) | 2020-07-09 | 2020-07-09 | Agile satellite scheduling method considering time-dependent conversion time |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010654910.7A CN111651905B (en) | 2020-07-09 | 2020-07-09 | Agile satellite scheduling method considering time-dependent conversion time |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111651905A CN111651905A (en) | 2020-09-11 |
| CN111651905B true CN111651905B (en) | 2024-11-05 |
Family
ID=72345641
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010654910.7A Active CN111651905B (en) | 2020-07-09 | 2020-07-09 | Agile satellite scheduling method considering time-dependent conversion time |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111651905B (en) |
Families Citing this family (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112633643B (en) * | 2020-12-09 | 2022-09-23 | 合肥工业大学 | Multi-satellite multi-station task scheduling method and system |
| CN115186178B (en) * | 2022-06-29 | 2023-10-17 | 北京空间飞行器总体设计部 | Agile remote sensing satellite task solution space searching method based on step length self-adaption |
| CN115343735B (en) * | 2022-10-14 | 2022-12-20 | 北京航天驭星科技有限公司 | Satellite survey station tracking system, and calculation method, device and storage medium thereof |
| CN116599575B (en) * | 2023-07-17 | 2023-10-13 | 数字太空(北京)科技股份公司 | Simulation environment construction method and device for large-scale remote sensing task system |
| CN118075596B (en) * | 2024-04-19 | 2024-06-25 | 西安电子科技大学 | Agile imaging satellite instruction sequence autonomous generation method |
| CN118607895A (en) * | 2024-08-08 | 2024-09-06 | 中国人民解放军国防科技大学 | Method and device for scheduling earth observation satellite, electronic equipment and storage medium |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104063748A (en) * | 2014-06-28 | 2014-09-24 | 中国人民解放军国防科学技术大学 | Algorithm for imaging satellite-oriented time-dependent scheduling problem |
| CN107025363A (en) * | 2017-05-08 | 2017-08-08 | 中国人民解放军国防科学技术大学 | A kind of adaptive big neighborhood search method of Agile satellite scheduling |
Family Cites Families (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| IT201700056428A1 (en) * | 2017-05-24 | 2018-11-24 | Telespazio Spa | INNOVATIVE SATELLITE SCHEDULING METHOD BASED ON GENETIC ALGORITHMS AND SIMULATED ANNEALING AND RELATIVE MISSION PLANNER |
| CN107578178B (en) * | 2017-09-11 | 2018-08-28 | 合肥工业大学 | Based on the dispatching method and system for becoming neighborhood search and gravitation search hybrid algorithm |
| CN107608793B (en) * | 2017-09-13 | 2021-08-13 | 航天恒星科技有限公司 | Quick search agile satellite task planning method |
-
2020
- 2020-07-09 CN CN202010654910.7A patent/CN111651905B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104063748A (en) * | 2014-06-28 | 2014-09-24 | 中国人民解放军国防科学技术大学 | Algorithm for imaging satellite-oriented time-dependent scheduling problem |
| CN107025363A (en) * | 2017-05-08 | 2017-08-08 | 中国人民解放军国防科学技术大学 | A kind of adaptive big neighborhood search method of Agile satellite scheduling |
Also Published As
| Publication number | Publication date |
|---|---|
| CN111651905A (en) | 2020-09-11 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111651905B (en) | Agile satellite scheduling method considering time-dependent conversion time | |
| CN110889625B (en) | Task planning method for swarm unmanned aerial vehicle system | |
| CN113852405B (en) | Construction method and device of multi-beam relay satellite task scheduling model | |
| CN110412869B (en) | Multi-satellite cooperative space dynamic target real-time tracking method | |
| CN109709535B (en) | Beam resident scheduling method for cooperative distributed system | |
| JP5306051B2 (en) | Thermal power distribution device | |
| CN109886859A (en) | Data processing method, system, electronic equipment and computer readable storage medium | |
| CN107622699A (en) | All the period of time spatial domain conflict probe and solution desorption method based on sequential | |
| CN113269385A (en) | Emergency scheduling method and system for agile satellite resources | |
| CN113313349B (en) | Satellite task resource matching optimization method and device, storage medium and electronic equipment | |
| CN113313451A (en) | Multi-objective optimization logistics scheduling method based on improved cuckoo algorithm | |
| CN112862306A (en) | Multi-satellite online collaborative task planning method for time-sensitive moving target tracking | |
| CN116090742A (en) | Unmanned cluster distributed task allocation algorithm considering time window constraint | |
| CN115081936A (en) | Method and device for scheduling observation tasks of multiple remote sensing satellites under emergency condition | |
| CN113535387B (en) | Heterogeneous-aware GPU resource allocation and scheduling method and system | |
| CN113852406B (en) | Multi-beam relay satellite task scheduling method and device | |
| CN115423393B (en) | Order scheduling method and device of dynamic self-adaptive scheduling period based on LSTM | |
| CN113158549B (en) | Diversified task oriented ship formation grade repair plan compilation method | |
| CN114298294B (en) | Neural network memory optimization method and device based on hardware accelerator | |
| CN115766612A (en) | Scheduling method based on weight conversion probability and corresponding device | |
| CN114637331A (en) | Unmanned aerial vehicle multi-task path planning method and system based on ant colony algorithm | |
| CN115047906A (en) | Multi-unmanned aerial vehicle cooperative attack moving target task allocation method based on optimization particle swarm optimization algorithm | |
| Park et al. | Collaborative Computation Offloading Scheme Based on Deep Reinforcement Learning | |
| CN111260254B (en) | Online scheduling method for dynamically arriving tasks | |
| CN117009057B (en) | Concurrent transaction scheduling method based on dynamic value |
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 |