A Method and System for Solving Intelligent Mission Planning of Agile Earth Observation Satellites

Abstract

The present invention discloses a method and system for solving the intelligent mission planning of an agile earth observation satellite. The method includes designing a continuous coding representation for describing the solution of the intelligent mission planning of the agile earth observation satellite, constructing a minimum subgraph model for solving the intelligent mission planning of the agile earth observation satellite, giving the properties and theorems for constructing the minimum subgraph and searching for the optimal path, designing a parallel decoding and solving algorithm for constructing the minimum subgraph by bidirectional traversal and searching for the optimal path by dynamic programming, using the differential evolution algorithm framework, and applying the real number coding and decoding algorithm to the solution of the mission planning of the fully agile earth observation satellite. Based on the properties and laws existing in the background problem graph model, the present invention optimizes and parallelizes multiple links such as constructing the minimum graph model, searching the graph, and decoding into a feasible observation sequence, reduces the computational overhead, and the graph decoding algorithm itself is an exact optimization method, which has a good effect on improving the quality of the solution.

Description

Technical Field

[0001] The present invention relates to the technical field of satellite observation, and particularly to a method and system for solving the intelligent mission planning of agile earth observation satellites. Background Art

[0002] Earth observation satellites are equipped with different optical sensors to detect the earth's surface and obtain relevant information. As an important means of imaging the ground, according to the different attitude maneuvering capabilities and working mechanisms, earth observation satellites can be divided into conventional earth observation satellites and agile earth observation satellites. Compared with conventional earth observation satellites that only have the ability to slewing maneuver, agile earth observation satellites have the three-degree-of-freedom maneuvering capabilities of slewing, pitching, and yawing, and show stronger observation mission planning capabilities. With the rapid progress of space technology and satellite technology, agile earth observation satellites, as a new type of earth observation satellites, have become the objects that major space powers compete to develop due to their low cost, short development cycle, strong attitude maneuvering capabilities, and high adjustment accuracy.

[0003] In the optimization process of the intelligent mission planning of agile earth observation satellites, not only a feasible observation subsequence needs to be determined in the set of observations to be scheduled, but also the start time of each observation in the subsequence needs to be solved (with an integer second as the unit interval). Therefore, this problem is both a combinatorial optimization problem and an integer programming problem.

[0004] In recent years, for the mission planning of semi-agile earth observation satellites with limited conditions (satisfying the FIFO property), heuristic algorithms based on permutation coding have been adopted and achieved good results. However, the existing coding and heuristic iterative algorithms are only applicable to semi-agile observation satellites, rely on insertion operations based on limited conditions, and cannot be extended to fully agile satellites. Moreover, for agile observation satellites, the reason why continuous evolutionary algorithms are rarely used currently is that there is a non-linear switching time between adjacent observations, that is, the constraints are difficult to handle, resulting in difficult decoding and high computational complexity. In addition, evolutionary algorithms require a series of operations such as selection, mutation, boundary checking, decoding, and calculation of the objective function for each individual in the population, with a large computational overhead, resulting in poor solution. Therefore, the present invention proposes a method and system for solving the intelligent mission planning of agile earth observation satellites to solve the problems existing in the prior art. Summary of the Invention

[0005] Aiming at the above problems, the purpose of the present invention is to propose a method and system for solving the intelligent mission planning of agile earth observation satellites, to solve the problems that in the current mission planning of semi-agile earth observation satellites, the existing coding and heuristic iterative algorithms are only applicable to semi-agile observation satellites and cannot be extended to fully agile satellites, as well as difficult decoding, high computational complexity, large computational overhead, and poor solution.

[0006] To achieve the objectives of the present invention, the present invention is implemented through the following technical solutions: A method for solving the intelligent mission planning of agile earth observation satellites, comprising the following steps:

[0007] Step 1: On the continuous real number space, design a continuous coding representation for describing the solution of the intelligent mission planning of agile earth observation satellites;

[0008] Step 2: Based on the coding representation in Step 1, construct a minimum subgraph model for solving the intelligent mission planning of agile earth observation satellites to reduce the complexity of problem solving;

[0009] Step 3: Give the properties and theorems for constructing the minimum subgraph and searching for the optimal path;

[0010] Step 4: According to the minimum subgraph model in Step 2 and the properties in Step 3, design a parallel decoding and solving algorithm for bidirectionally traversing to construct the minimum subgraph and dynamically programming to search for the optimal path;

[0011] Step 5: Based on the continuous coding and decoding technology in Step 4, use the differential evolution algorithm framework, apply the real number coding and decoding algorithm to the solution of the full agile earth observation satellite mission planning, and design mutation and crossover operators related to the problem characteristics.

[0012] The further improvement lies in: In Step 1, the specific steps for designing the continuous coding representation are: using a real value to represent a specific task observation, the integer part represents the selection of the visible time window when performing the task observation, the decimal part represents the specific observation time within the selected time window, and the calculation method for converting the real number solution vector into an integer solution vector.

[0013] The further improvement lies in: In Step 2, map the problem solving to the problem solving of graph theory, and construct a minimum subgraph model for solving the problem, specifically including the definition, properties and characteristics of the minimum subgraph.

[0014] The further improvement lies in: In Step 4, in the minimum subgraph construction algorithm, traverse each vertex in topological order as a whole, which is represented as forward traversal.

[0015] The further improvement lies in: In Step 4, in the minimum subgraph construction algorithm, based on the forward traversal, search and draw the edges pointing to the current vertex in the reverse order of topological sorting of the current vertex, which is represented as backward traversal.

[0016] The further improvement lies in: In Step 4, when searching for the optimal path, utilize the optimal substructure property of the directed acyclic graph, and use the dynamic programming method to solve the optimal objective and search for the optimal path.

[0017] A further improvement lies in that: in step five, when designing mutation and crossover operators related to problem characteristics, a specific crossover operation that preserves the component values on the longest path is retained.

[0018] A system for solving the intelligent mission planning of agile earth observation satellites includes:

[0019] A continuous coding design module, which is used to design a continuous coding representation for describing the solution of the intelligent mission planning of agile earth observation satellites, enabling the continuous evolutionary algorithm to be applied to the solution of the problem;

[0020] A minimum graph model construction module, which uses the bidirectional traversal method to construct a minimum subgraph model for solving the intelligent mission planning of agile earth observation satellites;

[0021] A parallel solution algorithm design module, which is used to design a dynamic programming algorithm for parallel solution of constructing the minimum subgraph and searching for the optimal path;

[0022] A mutation and crossover operator design module, which is used to design mutation and crossover operators related to problem characteristics;

[0023] A property and theorem design module, which is used to give the necessary properties and theorems for constructing the minimum subgraph and searching for the optimal path and prove the correctness of the real number encoding and decoding algorithm;

[0024] An agile earth observation satellite mission planning solution module, which is applicable to the solution of full-agile and semi-agile earth observation satellite mission planning.

[0025] The beneficial effects of the present invention are as follows: Based on the properties and laws existing in the background problem graph model, the present invention optimizes and parallelizes in multiple links such as constructing the minimum graph model, searching the graph, and decoding into a feasible observation sequence, reducing the computational overhead. Moreover, the graph decoding algorithm itself is an exact optimization method, which has a good effect on improving the quality of the solution and can be used for the solution of the full-agile earth observation satellite mission planning. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1 is a schematic flow chart of the method for solving the intelligent mission planning of agile earth observation satellites of the present invention;

[0027] Figure 2 is a schematic diagram of an example of the minimum subgraph for solving the problem in Embodiment 1 of the present invention;

[0028] Figure 3 is a schematic diagram of an example of a non-minimum subgraph for solving the mission in Embodiment 1 of the present invention;

[0029] Figure 4 is a schematic diagram of an example of the decoding calculation process of the real number encoding of the AS-01 satellite 3 mission in Embodiment 2 of the present invention;

[0030] Figure 5 This is a schematic diagram of the decoding calculation result of real - number coding for the AS - 01 satellite 3 mission in the second embodiment of the present invention. Detailed implementation manners

[0031] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without making creative efforts belong to the scope of protection of the present invention.

[0032] The symbols and variables involved in the present invention are as follows:

[0033] J = {j 1 ,…,j n}: The set of observation tasks given by the user, where n represents the total number of observation tasks to be scheduled;

[0034] During the scheduling time range, the set of orbital revolutions of the observation satellite orbiting the earth, where m o represents the total number of orbits;

[0035] For any observation j i ∈J, the following variables are defined:

[0036] p i : The revenue obtained by completing the observation task J i ;

[0037] d i : The working duration required to execute the observation task J i ;

[0038] During the scheduling time range, the set of optional observation time periods (i.e., visible time windows) for the observation task J i , where m i is the total number of visible time windows for j i ;

[0039] vtw i,q =(tws i,q ,twe i,q ,o i,q ): The q - th visible time window of the observation task J i , where tws i,q is the start time of this time window, tws i,q is the end time of this time window, and o i,q is the orbital index number where this time window is located;

[0040] yi : The decision variable for solving the problem. If j i is selected and the observation is executed, then y i = 1; otherwise, y i = 0;

[0041] t i : The decision variable for solving the problem. The start observation time for executing observation j i , measured in integer seconds;

[0042] In the observation period within the time window vtw i,q , the observation angle vector with t i as the start observation time. This observation angle is composed of three angle components: pitch α, roll β, and yaw γ. The specific angles are known data and are determined according to the actual observed satellite and mission;

[0043] The observation scheduling scheme for the intelligent mission planning of an agile earth observation satellite, which is also the solution to this problem. Among them, sch k is the feasible observation sequence scheduled on orbit k;

[0044] The constraint conditions involved in the problem model to be solved in the present invention are as follows:

[0045] For any j i ∈J, at most one start time is selected to execute the observation once. In addition, for any two consecutive adjacent observations j p and j s on the same orbit circle, the observation windows cannot overlap, and there must be enough gap time for attitude conversion of the observation angle. Among them, j p and j s represent the predecessor task and the successor task respectively. The constraint conditions that the feasible sequence S needs to satisfy are represented by the following formulas (1) and (2), where tr p,s is the switching time between adjacent tasks:

[0046]

[0047] t p + d p + tr p,s ≤ t s , if j p and j s are two consecutive observations and in sch k of S (2)

[0048] When the constraint (2) is satisfied, it is denoted as p → s, which is called j pReachable to j s , otherwise denoted as Taking the semi-agile earth observation satellite AS-01 as an example, the switching time between adjacent tasks is given by the following formulas (3) and (4), where v 1 = 1.5° / s, a 1 = 5, v 2 = 2° / s, a 2 = 10, v 3 = 2.5° / s, a 3 = 16, v 4 = 3° / s, a 4 = 22.

[0049]

[0050]

[0051] The objective function involved in the problem model to be solved by the present invention is as follows:

[0052] Any feasible observation sequence is a solution to the mission planning of the agile earth observation satellite. The quality of the solution is evaluated by the objective function, with the maximization of the observation benefit as the optimization objective, as shown in the following formula (5):

[0053]

[0054] Embodiment 1

[0055] Refer to Figure 1 , Figure 2 , Figure 3 , this embodiment provides a method for solving the intelligent mission planning of an agile earth observation satellite, including the following steps:

[0056] Step 1: On the continuous real number space, design a continuous coding representation for describing the solution of the intelligent mission planning of the agile earth observation satellite. The specific steps are as follows: Represent a specific observation task with a real value, where the integer part represents the selection of the visible time window when performing the task observation, and the decimal part represents the specific observation time within the selected time window. For the observation set J, use the real number vector X = (x 1 ,…x n ) to represent the solution of the problem. The real value x i corresponds to the observation j i ∈J, where 1 ≤ x i <m i +1. Obviously Then, the observation j i selects the visible time window as the time window for performing the observation task. The real number coding ensures that each task only selects a unique time window to participate in the observation. At the same time, the observation j iThe start time t i is calculated using the following formula (6):

[0057]

[0058] Step 2: Based on the encoded representation in Step 1, construct a minimum subgraph model for solving the intelligent mission planning of agile earth observation satellites to reduce the complexity of problem solving. The minimum subgraph model uses a graph model to solve problems, specifically including the characteristics and definitions of the minimum subgraph. For any real number solution, its integer vector solution is calculated through formula (6);

[0059] For any real number solution X = (x 1 ,…x n ), the integer vector solution Z = (t 1 ,…t n ) can be calculated by formula (6). Sort the observations in J according to the corresponding component values of Z to obtain an observation sequence where Obviously, the obtained observation sequence is mostly not a feasible observation sequence. Decoding the continuous encoding requires finding a subsequence of the above sequence that satisfies the above constraint condition (2). Therefore, the decoding process of finding the optimal subsequence with the maximum total benefit of observation tasks in the sequence is the optimal decoding and is also a combinatorial optimization problem;

[0060] In this embodiment, the above decoding process with the characteristics of combinatorial optimization is mapped to the problem of searching for the optimal path in a graph model, forming an exact algorithm for decoding the real number encoding. The graph model for solving the problem is defined as follows:

[0061] First, group and divide the observations in J by orbit, and then decode each orbit one by one. The observation tasks to be scheduled on the same orbit form the vertex set of the graph, denoted by V. For any two observations j p and j s , where t p <t s , p→s can be calculated by looking up the table through t p and t s using formula (3), or If p→s holds, draw a directed edge <v p ,v s >, pointing from vertex v p to vertex v s . The set of directed edges is denoted as E max , and G<V,E max > is a complete directed graph describing the problem;

[0062] In the complete graph G<V,E maxIn it, any simple path (a path with distinct vertices is called a simple path) constitutes a solution to the problem. The sum of the vertex benefits on the simple path is the objective function value. The optimal solution to the problem is to find an optimal path that maximizes the sum of the vertex benefits on the path.

[0063] In the complete graph G<V, E max >, if among all the simple paths (without loops) between any two vertices (v p and v s ), only the longest path (with the most vertices on the path) whose vertex sets are not contained in each other is retained, it forms the smallest subgraph G<V, E min > that can solve the problem. Figure 1 For an instance of G<V, E min >, {v 1 , v 2 , v 4, v 5 , v 8 , v 9}, {v 1 , v 2 , v 3 , v 5 , v 8 , v 9}, and {v 1 , v 2 , v 3 , v 6 , v 8 , v 9} are three non - inclusive longest paths between vertex 1 and vertex 9. Figure 1 In it, after observing j 1 and j 2 in sequence, because the observation path <v 1 , v 2 > bifurcates into {v 1 , v 2 , v 4} and {v 1 , v 2 , v 3}, and because the observation path <v 1 , v 2 , v 3 > bifurcates into {v 1 , v 2 , v 3 , v 5} and {v 1 , v 2 , v 3 , v 6}, finally forming the above - mentioned three non - inclusive longest paths.Figure 2 It is not the minimum subgraph because Figure 2 some of the paths existing in 1 ,v 2 ,v 3 ,v 5}, {v 1 ,v 5}, and {v 1 ,v 3 ,v 5};

[0064] Since the observation tasks have temporal order after being continuously encoded and decoded into integer encodings, the above graph model obviously has the following properties. The mathematical proof of the properties is shown in Step 3:

[0065] Property 1: The graph G<V, E max >> is a directed acyclic graph. If the integer solution vector Z = (t 1 , … t n ) of the start time is used as the key value pair to sort the alternative observation tasks from smallest to largest, and the sorting result on a certain orbit is represented as the observation sequence V = <v 1 , v 2 , …, v k >>, and is mapped to the vertices of the graph G<V, E max >>, where represents the start time of the observation task corresponding to the vertex v i , then the linear sorting of the vertex sequence V = <v 1 , v 2 , …, v k >> is the topological sorting of G<V, E max >>;

[0066] Property 2: According to the definition of G<V, E min >>, among any two vertices with directed edges, there is obviously no simple path passing through other vertices;

[0067] Property 3: G<V, E max >> has an optimal path optimal substructure. The optimal path between two vertices includes the optimal paths between other vertices;

[0068] Step 3: Give the properties and theorems for constructing the minimum subgraph and searching for the optimal path, and conduct a mathematical proof.

[0069] Proof of Property 1: If there exist two vertices v p and v s , v p can reach v s through a simple path, and can return from v s to v pForm a circuit. According to the definition of G<{,E max >, and will hold simultaneously. Since a contradiction occurs, then G<V,E max > has no circuit and is a directed acyclic graph. For any edge <v max ,v p ,v s > in G<V,E p >, if v is a pre-order vertex, then in V = {v 1 ,v 2 ,…,v k}, v p must also precede v s in order. According to the definition of topological sorting, V = {v 1 ,v 2 ,…,v k} is the topological sorting of G<V,E max >;

[0070] Proof of Property 3: Let V = {v 1 ,v 2 ,…,v k} be the topological sorting of graph G<V k ,E max > obtained using the theorem, where the longest path between v 1 and v k is denoted as path 1,k , and path 1,s is a sub-path included in path 1,k . If path 1,k is not the longest path, then there exists a longest path path' 1,s . Thus, replacing path 1,s with path' 1,s can construct a longer path path' 1,k , which contradicts the fact that path 1,k is the longest path;

[0071] Step 4: According to the minimum subgraph model in Step 2 and the properties in Step 3, design a parallel decoding and solution algorithm for constructing the minimum subgraph and dynamically programming to search for the optimal path. To effectively search for the optimal path, this embodiment uses Property 1 and Property 2 to construct G<V,E min >, and uses Property 3 to search for the optimal path based on G<V,E min . When constructing the minimum subgraph, the vertex set is traversed globally in the order of topological sorting (forward traversal), and the predecessor vertices of the current vertex are traversed in the direction order of topological sorting and the edges pointing to the current vertex are drawn. The following information needs to be maintained during the traversal:

[0072] PVS j : v j The set of predecessor vertices to be traversed, PVS j The initial value of min is all the predecessor vertices of v in the topological sorting of G<V, E j ;

[0073] PRVS j : The set of predecessor vertices that can reach v through a simple path, PRVS j The initial value of j is

[0074] PPVS j : In G<V, E min >, the set of predecessor vertices that point to v through a directed edge, PRVS j The initial value of j is

[0075] MPP j : The optimal path with v min as the end point in G<V, E j >, MPP j The initial value of

[0076] MP j : The optimal profit obtained from the optimal path with v min as the end point in G<V, E j >, MP j The initial value of is the observed profit corresponding to v j ;

[0077] The minimum subgraph construction and optimal path search algorithm traverses each vertex in topological sorting order, which is expressed as forward traversal. During the forward traversal, assume the currently visited vertex is v j , and perform the following operations:

[0078] For the predecessor vertices in PVS j , always traverse in reverse order from near to far in topological order, which is called backward traversal. For example, at the beginning of the backward traversal, select to visit v j-1 . If v j-1 →v j holds, then an edge <v min , v j-1 , v j > can be drawn in G<V, E

[0079] PPVSj = PPVS j ∪{v j-1} (7)

[0080] PRVS j = PRVS j ∪{v j-1}∪PRVS j-1 (8)

[0081]

[0082]

[0083] If holds, then execute formula (10). When PVS j is equal to , the backward traversal terminates. Next, according to Property 3, the optimal benefit MP j with v j as the end point is equal to the maximum value of the optimal benefits with all the precursor vertices in PPVS j as the end points plus p j . MPP j is the optimal path corresponding to the vertex that can obtain the maximum optimal benefit among all the precursor vertices in PPVS j plus v j to obtain the optimal path corresponding to the optimal benefit MP j . The calculations of MP j and MPP j are shown in the following formulas (11) and (12) respectively:

[0084] MP j = max{p j + MP e | v e in PPVS j} (11)

[0085] MPP j = path e ∪{v j}, when p j + MP e = MP j (12)

[0086] Step Five: Based on the continuous encoding and decoding technology, and according to the solution algorithm in Step Four, use the differential evolution algorithm framework. When designing the mutation and crossover operators related to the problem characteristics, retain the specific crossover operation of the component values on the longest path;

[0087] Map a class of scheduling problems with time-dependent characteristics constraints into a directed acyclic graph, and use the relevant characteristics and properties of the directed acyclic graph to simplify the problem and correctly solve the problem;

[0088] Step 6: Utilize Steps 1 to 5 to map a class of scheduling problems with time-dependent characteristics constraints into a directed acyclic graph, and use the relevant characteristics and properties of the directed acyclic graph to simplify the problem and correctly solve the problem; This real number encoding and decoding algorithm is applicable to the solution of full-agility and semi-agility earth observation satellite mission planning.

[0089] This embodiment also provides a system for solving the intelligent mission planning of agile earth observation satellites, including a continuous coding representation design module, a minimum graph model construction module, a dynamic programming parallel solution algorithm design module, a mutation and crossover operator design module, a property and theorem design module, and an agile earth observation satellite mission planning solution module. The continuous coding representation design module is used to design the continuous coding representation for describing the solution of the intelligent mission planning of agile earth observation satellites. The minimum graph model construction module is used to construct the minimum subgraph model for solving the intelligent mission planning of agile earth observation satellites through bidirectional traversal. The parallel solution algorithm design module is used for the parallel solution algorithm of dynamic programming to search for the optimal path. The mutation and crossover operator design module designs mutation and crossover operators related to the problem characteristics based on the continuous differential evolution algorithm. The property and theorem design module is used to give the necessary properties and theorems for constructing the minimum subgraph and searching for the optimal path and prove the correctness of the real number encoding and decoding algorithm. The agile earth observation satellite mission planning solution module is applicable to the solution of full-agility and semi-agility earth observation satellite mission planning.

[0090] Embodiment 2

[0091] S1. Real number encoding method and example

[0092] In this embodiment, the information of 3 observation tasks is extracted from the satellite AS-01 public dataset as an example for real number encoding. The known data information of the observation tasks is shown in Table 1 below:

[0093] Table 1 Example of 3-task data information of satellite AS-01

[0094] <![CDATA[j i > <![CDATA[vtw i，q > <![CDATA[d i > <![CDATA[p i > <![CDATA[tws i，q > <![CDATA[twe i，q > orbit 1 1 25 4 24283 24589 21 1 2 25 4 59157 59429 26 2 1 24 8 18539 18845 20 2 2 24 8 58881 59182 26 3 1 16 6 18459 18709 20 3 2 16 6 58984 59254 26

[0095] Randomly take the real number encoding values of three tasks: X = (2.8, 2.0, 2.3). From the meaning of the encoding and the above formula (6), the selected time window and the start time of the observation task can be calculated. The calculation result is Z = (59354, 58881, 59060). The specific calculation process is as follows Figure 4 as shown.

[0096] According to the start time of the observation task obtained by decoding, from the public dataset, through look-up table, the observation angles of the specific observation task are as shown in Table 2 below:

[0097] Table 2 Example of Observation Angle Data Information for AS-01 Satellite 3 Task

[0098] <![CDATA[j i > <![CDATA[vtw i，q > <![CDATA[t i > roll pitch yaw 1 2 59354 20.742 -24.406 2.915289 2 2 58881 31.75 45 4 3 2 59060 -19.021 24.114 4.3924

[0099] According to the start time of the observation task, the observation order is determined to be j 2 、j 3 、j 1 , according to Table 3 below, the change value of the observation angle vector can be calculated as follows:

[0100] Δθ 2，3 =(31.75 + 19.021)+(45 - 24.114)+(4.3924 - 4)=72.0494

[0101] Δθ 3，1 =(20.742 + 19.021)+(24.406 + 24.114)+(4.3924 - 2.9153)=89.76

[0102] According to the above formula (1), the switching time and the satisfaction of the constraint conditions are calculated as follows:

[0103] tr 2，3 =16 + 72.0494 / 2.5 = 44.81976

[0104] t 2 +tr 2，3+ d 2 =58881 + 44.81976 + 24 = 58949 < t 3 =59060

[0105] tr 3，1 =16 + 89.76 / 2.5 = 51.74708

[0106] t 3 +tr 3，1 +d 3 =59060 + 51.74708 + 16 = 59128 < t 1 =59354

[0107] In summary, the randomly generated real number solutions are directly feasible solutions, and the formed feasible sequence and scheduling scheme are as Figure 5 shown.

[0108] S2. Graph Theory Decoding Optimization Method and Example

[0109] Suppose Figure 2The real number encoding of the 9 observation tasks in the embodiment 1 is mapped to the observation sequence obtained by the method steps in the embodiment 1 as <v 1 , v 2 , ..., v 9 >, the vertex sequence is Figure 2 The topological sorting of , and the reachability of any two ordered vertices can be calculated by the decoded integer solution according to the method steps in the first embodiment. The result of the reachability calculation in the graph optimization process is shown in Table 3, where the symbol √ is reachable, the symbol × is unreachable, and the symbol * indicates that the adjacent observations of the reachability calculation can be omitted in the optimization process of constructing the minimum subgraph, which reduces the computational complexity. In addition, in the process of traversing the vertices, construct G <V,E min >The method of finding the optimal path at the same time also reduces the computational complexity.

[0110] Table 3. Examples of inter-vertex reachability

[0111]

[0112] The integer digits of the real number code corresponding to the observation task determine which time window of the observation task is selected, and the decimal digits of the real number code corresponding to the task determine the integer value of the task start time. The selected tasks are grouped on the corresponding orbits to obtain the observation order of the tasks on the same orbit. When the number of tasks is large and the time windows are dense, adjacent tasks may not meet the constraints. Therefore, the decoding and optimization of the real number solution is the decoding process of finding the optimal sub-observation sequence under the constraints, which is a combinatorial optimization problem.

[0113] According to properties 1 to 3 and formulas (7) to (12) in the first embodiment, this embodiment designs a parallel computing method for bidirectional traversal to construct a minimum subgraph and dynamically search for an optimal path. The specific implementation process is as follows:

[0114] 1. For vertex sequence <v 1 ,v 2 ,…,v 9 >For every vertex v j Initialize the maintained collection: v j The preceding vertex set For example: PVS 3 = {v 1 , v 2}; v can be reached through a simple path j The preceding vertex set The set of predecessor vertices pointing to v{ With v j The best path to the end point With v j The maximum benefit that can be obtained by a simple path to the end point j MPj ←p i ;

[0115] 2. Forward traverse the topological sort <v 1 , v 2 ,..., v 9 >. For each vertex in the forward traversal, mark the vertex being visited in the forward traversal as the current vertex;

[0116] 3. Perform a backward traversal of the current vertex. Assume the current vertex is v 3 . First, backward visit the nearest vertex v 2 . After calculation, v 2 →v 3 . Draw the directed edge <v 2 , v 3 > and perform the following operations:

[0117] PPVS 3 =PPVS 3 ∪{v 2}={v 2}

[0118] PRVS 3 =PRVS 3 ∪{v 2}∪PRVS 2 ={v 2}∪{v 3}={v 2 , v 3}

[0119]

[0120] MP 3 =max{p 3 +MP e |v e in PPVS 2}=13 + 9 = 22

[0121] MPP 3 =(MPP e ∪{v 2}, when MP e +P i =MP i )={v 1 , v 2 , v 3}

[0122] 4. Jump to step 3 and repeatedly execute step 3 until ;

[0123] 5. Jump to step 2 until the topological sort <v1 , v 2 , ..., v 9 >The vertices in are traversed;

[0124] 6. Output the maximum MP j and the corresponding path MPP j .

[0125] The pseudo code of the above calculation process is shown in Algorithm 1 below, and Table 3 and Figure 2 The corresponding calculation example is shown in Table 4 below, where the first column represents forward traversal of the vertices in the graph, the second column represents backward traversal of the predecessor vertex set of the current vertex, the value of the set element in the table represents the sequence number of the vertex, the symbol "+" represents the set union operation, and the symbol "-" represents the set difference operation.

[0126] Algorithm 1: Optimal path solving algorithm based on graph theory

[0127]

[0128]

[0129] Table 4 Figure decoding optimization algorithm calculation process example

[0130]

[0131] S3. Continuous Differential Evolution Algorithm for Intelligent Mission Planning of Agile Earth Observation Satellites

[0132] The algorithm first randomly generates an initial population of size np, with an individual vector length equal to the number of tasks n. The population is represented by pop. i,j Represents the jth component of the i-th individual in the population, which corresponds to the real value of the j-th observation task encoded by a real number. It is randomly generated within the legal real number range based on the number of time windows of the observation task.

[0133] Each iteration of the evolution process includes mutation and crossover operators, and records the best and worst solutions in the population. During the iteration, each individual in the population is pop i In the mutation operation, the mutation operator adopts the idea of ​​approaching the optimal solution and moving away from the worst solution. For each component of the individual vector, the operation represented by the following formula (13) is performed to generate a new individual v i , where parameter c 1 and c 2 is a randomly generated random number in the interval [0, 1], pop best，j is the jth component of the optimal individual obtained in the evolution process, pop worst，j is the jth component of the worst individual obtained in the evolution process. If the individual vector component pop corresponding to a certain observation task j isi，j If the boundary constraint is not satisfied, a real value that satisfies the boundary constraint is randomly generated.

[0134] v i，j = pop i，j + c 1 *(pop best，j - pop i，j ) + c 2 *(pop worst，j - pop i，j ) (13)

[0135] The new individual vector u generated through the crossover operation i , which is determined by whether the component real value corresponding to a certain observation task is copied from pop i，j or from v i，j . During crossover, first copy the individual components on the longest path obtained by graph optimization decoding from pop i，j to u i，j , that is, retain the information of the scheduled observation tasks that have been solved in the newly generated real - number solution u i . For the remaining components, the component values of vector u i，j are determined by the crossover probability to be copied from pop i，j or v i，j . This operator can ensure that the objective function value of the individual does not decrease with the number of generations during the evolution process, reduce the search space, and strengthen local search. The differential evolution algorithm based on graph decoding is shown in Algorithm 2 as follows:

[0136] Algorithm 2: Differential Evolution Solving Algorithm Based on Graph - Theoretic Decoding

[0137]

[0138]

[0139] The above - mentioned are only the preferred embodiments of the present invention, and are not intended to limit the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present invention shall be included within the protection scope of the present invention.

Claims

1. A method for solving intelligent mission planning of an agile earth observation satellite, characterized in that: The following steps are involved: Step 1: In the continuous real number space, a continuous coding representation is designed to describe the intelligent mission planning solution of the agile earth observation satellite. The specific steps of designing the continuous coding representation are: using a real value to represent a specific mission observation, the integer part represents the selection of the visible time window when performing the mission observation, the decimal part represents the specific observation time within the selected time window, and the calculation method of converting the real solution vector into the integer solution vector; Step 2: Based on the coding representation in step 1, a minimum subgraph model for solving the intelligent mission planning of agile earth observation satellites is constructed to reduce the complexity of problem solving. The problem solving is mapped to the problem solving of graph theory, and a minimum subgraph model for solving the problem is constructed, including the definition, properties and characteristics of the minimum subgraph. Step 3: Give the properties and theorems for constructing the minimum subgraph and searching for the optimal path; Step 4: According to the minimum subgraph model in step 2 and the properties in step 3, a parallel decoding and solving algorithm for bidirectional traversal to construct the minimum subgraph and dynamic programming to search for the optimal path is designed. In the algorithm for constructing the minimum subgraph, each vertex is traversed as a whole according to the topological sorting, which is represented as a forward traversal. On the basis of the forward traversal, the current vertex is searched and the edge pointing to the current vertex is drawn in the reverse order of the topological sorting, which is represented as a backward traversal. When searching for the optimal path, the optimal substructure property of the directed acyclic graph is used to adopt a dynamic programming method to solve the optimal target and solve the optimal path; Step 5: Based on the continuous encoding and decoding technology and according to the solution algorithm in step 4, use the differential evolution algorithm framework to apply the real number encoding and decoding algorithm to solve the fully agile earth observation satellite mission planning, and design mutation and crossover operators related to the problem characteristics.

2. A method for solving intelligent mission planning of agile earth observation satellites according to claim 1, characterized in that: In the step 5, when designing mutation and crossover operators related to the problem characteristics, a specific crossover operation of the component values ​​on the longest path is retained.

3. A system for solving the method for intelligent mission planning of agile earth observation satellites as claimed in claim 1, characterized in that: include: The continuous coding design module is used to design a continuous coding representation to describe the intelligent mission planning solution of agile earth observation satellites, so that the continuous evolutionary algorithm can be applied to solve the problem; The minimum graph model construction module uses the bidirectional traversal method to construct the minimum subgraph model for solving the intelligent mission planning of agile earth observation satellites; Parallel solution algorithm design module, used to design dynamic programming algorithms for constructing minimum subgraphs and searching for optimal paths for parallel decoding and solution; Mutation and crossover operator design module, used to design mutation and crossover operators related to problem characteristics; The property and theorem design module is used to provide the necessary properties and theorems for constructing the minimum subgraph and searching the optimal path; The agile earth observation satellite mission planning solution module is suitable for solving full-agile and semi-agile earth observation satellite mission planning.

Images