Multi-satellite cooperative task scheduling planning method of adaptive genetic algorithm based on fusion and merging observation

Abstract

The invention discloses an adaptive genetic algorithm multi-satellite cooperative task dynamic scheduling planning method based on fusion and merging observation, and the method comprises the steps: firstly, decomposing regional tasks through equal-ratio sampling, and constructing discrete observation points; secondly, tasks are converted into a weighted meta-task set, and selectable satellites and time windows are calculated in combination with satellite resources, tasks and attitude constraints; further, detecting and combining the meta-tasks meeting conditions, and constructing a combined observation model taking the maximum observation benefit as a target; and finally, solving is optimized by adopting a self-adaptive genetic algorithm, constraint is ensured to be met, and an executable multi-satellite collaborative observation scheduling scheme is generated. In the regional task planning and scheduling process, a regional discrete adaptive sampling method is adopted to quickly decompose regional tasks into point tasks, and the completion rate of regional observation tasks is increased; in addition, an effective meta-task merging observation method is designed, and the multi-satellite dynamic task planning benefit is effectively improved.

Description

Technical Field

[0001] The present invention relates to the technical field of multi-satellite task planning and scheduling, and in particular to a multi-satellite collaborative task scheduling and planning method based on an adaptive genetic algorithm with fused merged observations. By classifying all tasks, setting corresponding priorities, and in the process of regional task planning and scheduling, a regional discrete adaptive sampling method is used to quickly decompose regional tasks into point tasks, thereby improving the completion rate of regional observation tasks; further, when all tasks are converted into meta-tasks with weights for processing and planning, an effective meta-task merging observation method is designed, and an adaptive genetic algorithm is used to solve the constructed multi-satellite task planning model, thereby effectively improving the multi-satellite dynamic task planning and scheduling benefits. The algorithm can generate a task observation plan that simultaneously satisfies the maximum objective function and satisfies different constraint analyses, effectively utilizes satellite resources, and improves task planning efficiency, and belongs to the technical field of task planning and scheduling. Background Art

[0002] With the rapid progress of aerospace technology and the deepening of global space exploration, the satellite network, as an important bridge connecting the earth and the universe, is expanding at an alarming rate. This trend is not only reflected in the significant increase in the number of satellites, but also in the diversification of satellite types and the comprehensive improvement of their observation, communication, navigation and other capabilities. From high-resolution remote sensing satellites to deep space probes, from communication relay satellites to scientific experimental platforms, various types of satellites play an irreplaceable role in their respective fields.

[0003] However, with the increasing complexity of satellite networks, mission requirements are also showing an explosive growth trend. These tasks may involve earth observation, environmental monitoring, disaster warning, resource exploration, military reconnaissance and other aspects, which put forward unprecedented high requirements on the real-time, accuracy and flexibility of mission planning. Traditional mission planning methods, such as rule-based scheduling and simple optimization algorithms, are often difficult to cope with such complex and changing mission environments, especially when dealing with multi-satellite collaborative mission planning, their limitations are more obvious.

[0004] Multi-satellite collaborative mission planning is a highly complex system engineering problem, which requires the algorithm to optimize task allocation, resource scheduling and path planning while considering multiple constraints such as satellite orbit, attitude, payload capacity, task priority, time window, etc., to ensure that all tasks can be completed efficiently and accurately. In this process, it is necessary to consider not only the capabilities and limitations of a single satellite, but also the collaborative and competitive relationship between satellites to maximize the overall mission benefits.

[0005] Traditional mission planning methods are often inefficient when dealing with multi-satellite systems and are difficult to meet the needs of complex tasks. The adaptive genetic algorithm provides an effective solution for multi-satellite mission planning and scheduling. Genetic algorithm is an optimization algorithm based on natural selection and genetic variation. It simulates the biological evolution process to find the optimal solution. In multi-satellite mission planning and scheduling, the adaptive genetic algorithm can automatically adjust the algorithm parameters according to the characteristics of the mission and the status of satellite resources to improve the search efficiency and solution quality. It can effectively handle complex constraints, while taking into account the priority and time window of the task, to achieve the reasonable allocation of multi-satellite resources and the efficient execution of tasks. Summary of the invention

[0006] In order to overcome the defects in the above-mentioned prior art, the present invention adopts a regional discrete adaptive sampling method to quickly decompose regional tasks into point tasks in the process of regional task planning and scheduling, thereby improving the completion rate of regional observation tasks; when processing and planning all point tasks, an effective meta-task merging observation method is designed; in the process of solving the task planning model with an adaptive genetic algorithm, adaptive mutation and crossover probabilities are designed according to the number of population iterations and individual fitness, and the constructed multi-satellite task planning optimization objective function is solved, effectively improving the multi-satellite dynamic task planning benefits and algorithm solving efficiency; the present invention can improve the utilization rate of satellite resources and solve the timeliness problem of multi-satellite task execution. According to different task situations, different processing methods are selected to achieve reasonable allocation of satellite resources, increase the number of single-track observation tasks, and improve response timeliness.

[0007] To achieve the above object, the present invention includes the following technical solutions:

[0008] The method for dynamic scheduling planning of multi-satellite cooperative tasks based on adaptive genetic algorithm of fusion combined observations includes the following steps

[0009] Step 1: Task classification stage: construct a set of tasks to be planned based on task attributes;

[0010] Step 2: Regional task decomposition stage, analyze the task characteristics and transform the complex large-area observation task into a simple task;

[0011] Step 3: In the data preprocessing stage, all tasks are converted into meta-tasks. Based on the schedulable satellite resources, the optional satellite resources and corresponding time windows of each meta-task are analyzed and calculated;

[0012] Step 4: Task merging observation processing stage, merge observation constraint detection is performed on all meta-tasks in turn, and meta-tasks that meet the merging conditions are merged and observed;

[0013] Step 5: Optimize model construction, take the maximum observation benefit as the optimization goal, establish the conflict resolution principle in the scheduling planning process, and establish a multi-satellite multi-task combined observation scheduling planning model;

[0014] Step 6: Adaptive genetic algorithm is used to solve the established scheduling planning model and optimize the parameters to obtain the combined observation scheduling result that meets the task planning constraints;

[0015] Preferably, constructing the set of tasks to be planned in step 1 includes classifying and prioritizing the tasks, specifically including the following steps:

[0016] S11, task classification;

[0017] The input of the scheduling planning task mainly includes task information and satellite information. According to the task attributes, all task sets T = {T1, T2, T3, …, T M}, classified into point task set T poi ={T 1_poi ,T 2_poi ,T 3_poi ,…,T m1_poi}、Regional task set T are ={T 1_are ,T 2_are ,T 3_are ,…,T m2_are} and the dynamic task set T dyn ={T 1_dyn ,T 2_dyn ,T 3_dyn ,…,T m3_dyn}. Among them, the point task is composed of the longitude (lon i ) and latitude (lat i ) is determined, then the subtask T in the task set is i_poi The specific position is represented by T i_poi =(lat i ,lon i ); The position and shape of the regional observation task are determined by the longitude and latitude of each vertex, so the subtask T in the regional task set i_are Represented as T i_are ={(lat1,lon1),…,(lat i ,lon i )|i=1,2,…,N}.

[0018] S12, priority setting;

[0019] According to the rule that the priority of dynamic tasks is higher than that of point tasks, which is higher than that of regional tasks, corresponding priorities are set for all tasks in all task sets to be planned, where subtasks in the same set have the same priority.

[0020] Preferably, the regional task decomposition in step 2 specifically includes the following steps:

[0021] S21, read the longitude and latitude range of the regional task [lat1, lat i ], latitude range [lon1,lon i ], respectively set the discrete granularity △λ of the latitudinal decomposition of the region lat_i , latitude discrete granularity △λ lon_i , the longitude and latitude discrete granularity conforms to the following geometric relationship According to the above-mentioned regional adaptive discrete sampling method, the task area is divided into grids, and each sampling point is a vertex of the grid and corresponds to a position on the ground.

[0022] S22, filter the sampling points within the task area. After dividing the task area by geometric sampling, some sampling points will still be outside the area. In this case, these points need to be removed. Take one of the sampling points as an example to illustrate: Assume that the coordinates of one of the sampling points are (a, b), if lat1≤a≤lat i and lon1≤b≤lon i , then set the flag of this sampling point to true, otherwise to false, so as to determine all sampling points in the task area. The specific process is as follows Figure 2 shown.

[0023] S23, for the multi-satellite scheduling planning problem of regional missions, regional coverage is an important indicator to measure the quality of the planning scheme. For each sampling point in a given satellite imaging strip and regional mission, the imaging strip covers the sampling points of the discrete mission area. According to the positional relationship between the sampling point and the imaging strip, determine whether the discrete grid is covered. If all vertices of the grid are covered, the grid is judged to be covered by the imaging strip; otherwise, the grid is judged to be uncovered, and all covered grid points are counted to calculate the coverage rate of the mission area.

[0024] S24, performing steps S21 to S23 on the regional observation task in the dynamic task to decompose it.

[0025] Preferably, in step 3, all tasks are converted into meta-tasks, and optional satellite resources and corresponding time windows of each meta-task are analyzed and calculated based on schedulable satellite resources, which specifically includes the following steps:

[0026] S31, after processing through steps S1-2, all tasks are decomposed into three types of point tasks with different priorities. Now all point tasks are converted into meta-tasks with three different weights. The converted meta-task set is T init =0.5·T dyn ∪0.3·T are ∪0.2·T poi .

[0027] S32, after the conversion is completed, the constraints of the meta-task are consistent with the constraints of the original task. Before this, some constraint variables and parameters involved are defined:

[0028] Table 1 Satellite related parameter definition table

[0029]

[0030] Table 2 Task related parameter definition table

[0031]

[0032]

[0033] S33, converts the original task constraints and satellite constraints into meta-task constraints, mainly including:

[0034] 1) Synthetic aperture radar can penetrate clouds and observe all-weather without any requirements, while optical imaging satellites must meet the minimum solar altitude angle requirement when imaging: sun_angle>sunlight;

[0035] 2) The imaging resolution of the selected observation satellite sensor cannot be less than the resolution required by the meta-task: T init _resolution ≤ S_resolution;

[0036] 3) The time interval between two adjacent time windows of the satellite needs to satisfy the constraint of being greater than the satellite attitude adjustment time: et j-1 +trantime_i≤st j , among which, et j-1 It indicates the time when the last mission of the satellite ended, and trantime_i indicates the time when the attitude of satellite i was adjusted;

[0037] 4) After the mission is planned, the specific time window for the target mission to be executed by the satellite needs to meet the constraints within the mission's effective observation period, that is: i ,et i ∈d i ;

[0038] 5) The time interval between two adjacent observations of the same observation task needs to meet the minimum time interval constraint, that is: i+1 -et i ≥ time_interval;

[0039] 6) The same satellite available time period cannot be assigned to two or more tasks at the same time, that is, it cannot violate the mutual exclusion constraint: Among them, C i is a set of task decision variables, which is a Boolean variable. The value 0 indicates that the task is not planned, and the value 1 indicates that the task is planned;

[0040] 7) The satellite continuous working time cannot exceed the maximum continuous power-on time constraint:

[0041]

[0042] Where sat_type is the satellite type variable, the continuous working time of opt type satellite is 300s, and the continuous working time of sar type satellite is 600s.

[0043] S34, calculate the optional time window of the task. First, according to the position of each meta-task, calculate the satellite that can execute the task; further, according to the constraints of the task and the satellite in S32, calculate the optional satellite working time window of the task to form a set of optional time windows for the task.

[0044] Preferably, in step 4, all meta-tasks are sequentially subjected to merge observation constraint detection, and meta-tasks meeting the merge conditions are subjected to merge observation processing, which specifically includes the following steps:

[0045] S41, the conditions for meta-tasks to conduct merged observations mainly include the following three conditions: 1) multiple meta-tasks must be within the width of the satellite scan on the ground; 2) when conducting merged observations, the resolution of the same satellite must meet the resolution requirements of multiple meta-tasks; 3) the time window after merging multiple meta-tasks must be less than the single longest working time of the satellite.

[0046] S42, satellite attitude angle processing after merged observation: the observation angle difference of the observation satellite for any two meta-tasks should be within the satellite field of view angle. In order to cover all meta-tasks for synthetic observation, the average of the maximum observation angle and the minimum observation angle is selected as the attitude angle of the satellite in the merged observation time window.

[0047] S43, perform pairwise detection on all tasks with visible time windows of satellite i in the first orbital cycle. If both the deflection angle and observation time constraints are satisfied, the two meta-tasks are merged; if not, they are not merged.

[0048] S43, the combined observation task obtained in S42 is combined and checked again with the remaining meta-tasks of the round. If the combination constraint condition is met, it is further merged with it; if not, jump to S45.

[0049] S44, sequentially merge and detect the merged task with other tasks in the orbital cycle until the merge detection of all meta-tasks on the first orbital cycle of the satellite is completed, and then go to S45.

[0050] S45, traverse the other orbits of the satellite in sequence.

[0051] S46, traverse all satellites in sequence.

[0052] S47, the merging is completed, and the merged time window and satellite attitude are calculated.

[0053] Preferably, in step 5, the maximum observation benefit is taken as the optimization goal, the conflict resolution principle in the scheduling planning process is established, and a multi-satellite multi-task combined observation scheduling planning model is established, which specifically includes the following steps:

[0054] S51, design and optimize the objective function based on the benefits of the meta-task:

[0055] maxF=f(T dyn )+f(T are )+f(T poi )+f P

[0056] Among them, f(T dyn ) is the dynamic task benefit; f P Negative payoff for violating constraints; f(T are ) is the regional mission income, and the specific calculation method is:

[0057] Where score i_area is the revenue that regional task i deserves, p_a i is the number of sampling points actually executed after the regional task is discretized, spt i is the total number of sampling points after discretization of region i;

[0058] f(T poi ) is the point task income, and the specific calculation method is:

[0059] Where score i_point is the revenue that task i deserves, frequency is the frequency at which the task is executed, T init _frequency is the frequency at which the task needs to be executed.

[0060] S52 adopts the conflict resolution principle combining priority and benefit, that is, the meta-task with higher weight is executed first. When the weights of two meta-tasks are the same, the task with higher benefit is executed first.

[0061] Preferably, in step 6, the established scheduling planning model is optimized and solved by using an adaptive genetic algorithm to obtain a combined observation scheduling result that meets the task planning constraints, which specifically includes the following steps:

[0062] S61, encoding method, uses a matrix M to represent the multi-satellite multi-task visible window, that is:

[0063]

[0064] Among them, the element W ij (i=1,2,...,N s ; j = 1, 2, ..., N t ) is the set of satellite visible windows for mission j, N t is the number of tasks, N S is the number of satellites. Based on the matrix M, an ordered set of visible time windows corresponding to each task is created: in The visible window of each satellite is sorted according to the visible time, and the order remains unchanged during the entire calculation process. Let the dimension of the chromosome be two-dimensional, the first dimension represents the satellite serial number of the observation task, and the second dimension represents the start time of the satellite to perform the task. According to the above statement, the visible time window of the satellite for the task is encoded as:

[0065]

[0066] Among them, x i =0,1,2,...,N s represents the satellite number of mission j, y i =0,1,2,...,N t Indicates the visible time window number during which the satellite performs its mission.

[0067] S62, determining the initial solution, the initial solution is determined in chronological order according to the start time of the point tasks, and the time windows of the satellites are sorted in chronological order. Specifically, the earliest started task is preferentially assigned to the satellite that can make observations the earliest.

[0068] S63, adaptive crossover and mutation operator, in order to avoid the convergence speed and accuracy of the algorithm caused by constant crossover and mutation probability, the adaptive crossover probability P is designed according to the number of population iterations and individual fitness. c and mutation probability P m , the specific calculation methods are as follows:

[0069]

[0070] Among them, P c1 ,P c2 is the maximum and minimum value of the initial crossover probability, f′ is the maximum fitness of both parties in the crossover, P m1 ,P m2 is the maximum and minimum value of the initial mutation probability, f is the fitness of the mutant individual, and f avg is the average fitness of the contemporary population, f max is the maximum fitness of the contemporary population, g is the number of iterations of the current population, and G is the maximum number of iterations of the population.

[0071] S64, select the operator, sort the population chromosomes in descending order according to their fitness, and select new populations of the same size as the initial population in turn. This can reduce the population size and reduce the algorithm calculation pressure, and can also further screen out chromosomes with higher fitness to speed up convergence.

[0072] S65, solve and obtain the optimal individual, and further output the optimal multi-satellite observation scheduling combination that meets the planning constraints.

[0073] In the process of regional task planning and scheduling, the present invention adopts the method of regional discrete adaptive sampling to quickly decompose regional tasks into point tasks, thereby improving the completion rate of regional observation tasks; in addition, an effective meta-task merging observation method is designed, and an adaptive genetic algorithm is used to solve the constructed multi-satellite merged observation task planning model, thereby effectively improving the benefits of multi-satellite dynamic task planning. Experimental results show that the multi-satellite collaborative task planning and scheduling algorithm designed based on the adaptive genetic algorithm of fusion merged observation can generate a task observation plan that satisfies both the maximum objective function and different constraint analysis, effectively utilizes satellite resources, and improves task planning efficiency.

[0074] An electronic device comprises a memory, a processor and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the method for dynamic scheduling and planning of multi-satellite collaborative tasks based on an adaptive genetic algorithm with fused and merged observations is implemented.

[0075] A computer-readable storage medium stores computer instructions, which, when executed by a processor, implement the multi-satellite collaborative task dynamic scheduling planning method based on an adaptive genetic algorithm with fused combined observations.

[0076] The advantages of the present invention are:

[0077] In processing multi-type observation tasks, the present invention adopts a regional discrete adaptive sampling method for regional tasks to decompose regional observation tasks and convert them into discrete observation points, and further converts all tasks into meta-tasks with weights; fully considers the constraints related to tasks and satellites, and calculates the optional time windows of each meta-task; in order to improve the efficiency of satellite execution tasks, a constraint-based multi-task merged observation method is proposed, and all time windows of satellites are merged and detected; an optimization model is constructed with the maximum observation benefit of multiple satellites as the goal, and on the basis of traditional genetic algorithms, adaptive crossover and mutation probabilities are designed according to the number of population iterations and individual fitness values, and an adaptive genetic algorithm is proposed to solve the optimization function, which improves the flexibility of the update strategy, makes the algorithm have better convergence, and can simultaneously process the planning and scheduling of multiple different types of tasks of multiple satellites. Simulation scheduling experiments show that with the increase in the number of different types of tasks, the algorithm has more obvious advantages in terms of operating efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0078] Figure 1 is a flow chart of a multi-satellite collaborative mission planning method based on an adaptive genetic algorithm of fusion and combined observations of the present invention,

[0079] Figure 2 This is a regional task decomposition flow chart of the present invention,

[0080] Figure 3 Satellite distribution effect map simulation diagram,

[0081] Figure 4 This is a comparison chart of the average benefits of the traditional genetic algorithm, the adaptive genetic algorithm, the genetic algorithm considering the combined observation, and the adaptive genetic algorithm integrating the combined observation.

[0082] Figure 5 This is a comparison chart of the average running time of the traditional genetic algorithm, the adaptive genetic algorithm, the genetic algorithm considering merged observations, and the adaptive genetic algorithm that integrates merged observations in this example. DETAILED DESCRIPTION

[0083] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0084] Depend on Figure 1 As shown, the multi-satellite collaborative mission planning method based on the adaptive genetic algorithm of fusion combined observation of the present invention includes the following specific steps:

[0085] S1, task classification stage, constructs a set of tasks to be planned based on task attributes;

[0086] S11, task classification, divided into point tasks, regional tasks and dynamic tasks;

[0087] S12, setting priorities for dynamic tasks, regional tasks and point tasks from high to low priorities;

[0088] S2, regional task decomposition stage, analyzing task characteristics;

[0089] S21, using the regional discrete adaptive sampling method to discretize the task area into grids;

[0090] S22, sampling points within the task range of the screening area;

[0091] S23, calculate the coverage rate of regional tasks;

[0092] S24, loop S21 to S23 to discretize all regional tasks;

[0093] S3, data preprocessing stage, converts all tasks into meta-tasks, and calculates the optional satellite resources and corresponding windows of all meta-tasks;

[0094] S31, assign different weight coefficients to all point task benefits according to their priorities and convert them into meta-tasks;

[0095] S32, constraint variables and parameter definitions;

[0096] S33, the mission constraints and satellite constraints are equivalently converted into meta-mission constraints;

[0097] S34, a set of optional time windows for computing tasks.

[0098] S4, task merging observation processing stage, merge observation constraint detection is performed on all meta-tasks in turn, and meta-tasks that meet the merging conditions are merged and observed;

[0099] S41, meta-task merging observation constraints;

[0100] S42, satellite attitude angle processing after merging observations;

[0101] S43, performing a merge observation constraint check on all meta-tasks with visible time windows in the first orbital cycle of satellite i, and merging meta-tasks that meet the constraints;

[0102] S44, merge the combined observation task obtained in S43 with the remaining meta-tasks of the round again for detection, merge the meta-tasks that meet the constraints, if not, jump to S46

[0103] S45, sequentially performing combined detection on the meta-tasks in all orbital cycles of the satellite, and then proceeding to S46;

[0104] S46, traverse the other orbits of the satellite in sequence;

[0105] S47, traverse all satellites in turn;

[0106] S48, the merging is completed, and the merged time window and satellite attitude are calculated.

[0107] S5, establish the conflict resolution principle and build a multi-satellite multi-task combined observation scheduling planning model;

[0108] S51, design and optimize the objective function based on the task benefits;

[0109] S52, adopts the conflict resolution principle combining priority and benefit;

[0110] S6, adaptive genetic algorithm solution, parameter optimization solution of the established scheduling planning model;

[0111] S61, two-dimensional encoding method;

[0112] S62, determination of the initial solution;

[0113] S63, design adaptive crossover and mutation operators based on the number of population iterations and individual fitness values;

[0114] S64, select operator;

[0115] S65, solve the optimal individual and output a planning and scheduling solution that meets the task and satellite constraints.

[0116] In this embodiment, a data set is made using two publicly available satellite TLE ephemeris data. Each TLE contains two lines of text, with a total of 69 characters, which can provide sufficient information to determine the orbital position and velocity of the earth-orbiting satellite. These data include the longitude, latitude and altitude of the satellite at each time point and the solar altitude angle of the satellite relative to the subsatellite point. The calculation is completed using SGP4, which is a general and widely accepted standard algorithm suitable for most earth-orbiting satellites. It is mainly used to predict the position and velocity of low-Earth orbit satellites, calculate the orbit of the satellite based on TLE data, and can approximate the orbital perturbations caused by factors such as the earth's gravity, the perturbations of the sun and the moon, and atmospheric drag. The algorithm test scenario is described in JSON format. The scenario file contains the satellites used, key strategic observation points and areas, and possible dynamic adjustment information. The length and structure of this information may vary greatly. The area of ​​the regional mission is in square kilometers. The area is calculated under the EPSG:6933 coordinate reference system. This coordinate system is an equal-area projection suitable for calculating the actual area on the earth. The scene settings are shown in Table 3:

[0117] Table 3: Detailed scene settings

[0118] Scene complexity Low middle high Number of strategic observation points 45 180 450 Number of strategic observation areas 5 20 50 Additional points 5 20 50 Add Region Quantity 1 5 10 Number of optical satellites 20 52 132 Number of SAR satellites 10 28 68

[0119] The simulation experiment was set up with three different scenarios of high, medium and low complexity. Considering the timeliness requirement of multi-satellite mission planning, the simulation experiment set the observation time requirement of one mission to be 24 hours, that is, the simulation time was from 00:00:00 to 23:59:59, and a single-day mission was planned. Some scheduling planning results of the low-complexity scenario are shown in Table 4:

[0120] Table 4 Results of some planning schemes for low-difficulty scenarios

[0121] Satellite ID Task Number Start time End Time Roll Angle Opt01 [28 / 46] 22:21:17 22:22:07 13.37° Opt02 [18 / 11 / 12 / 27] 18:31:47 18:33:10 28.08° Opt03 [44 / 47 / 40] 14:39:54 14:40:53 13.77° Opt04 [55 / 1 / 16 / 6 / 5 / 29] 17:56:58 17:58:57 28.94° Opt05 [4] 16:33:56 16:34:06 21.22°20 Opt06 [19 / 31 / 43 / 37 / 50] 14:36:21 14:36:41 -13.21° Sar01 [5] 2:42:20 2:42:40 2.15° Sar02 [33 / 21 / 2 / 24 / 17] 18:43:08 18:46:03 12.74° Sar03 [7 / 32 / 56 / 13] 5:15:41 5:15:56 13.34° Sar04 [45 / 22 / 38 / 3 / 39] 5:47:11 5:49:03 3.91° Sar05 [14 / 37 / 9 / 40 / 42] 10:20:14 10:21:44 4.18° Sar06

[54] 18:30:40 18:31:00 3.35°

[0122] In order to evaluate the effectiveness and solution effect of the algorithm proposed in this invention, a genetic algorithm (GA) and an adaptive genetic algorithm (AGA) without region segmentation are selected for comparison. Each algorithm is performed five times in scenarios of the same difficulty. The average running time and average benefit of the algorithm of the task planning scheme are selected as evaluation criteria.

[0124] The algorithm parameters are set as follows:

[0125] Genetic algorithm (GA): population size 20, iteration number 100;

[0126] Genetic algorithm combined with regional adaptive discreteness (S-GA): population size 20, iteration number 100;

[0127] Adaptive genetic algorithm (AGA): population size 20, iteration number 100;

[0128] Adaptive genetic algorithm combined with regional adaptive discretization (S-AGA): population size is 20 and number of iterations is 100.

[0129] Figure 3 Satellite distribution renderings show that point observation missions are mainly distributed in my country's northwestern border areas, the South China Sea and Russia's Far East; regional observation missions are mainly distributed in my country's northeastern border areas and the Pacific Islands.

[0130] Figure 4 The average benefits of four algorithms in five experiments in three scenarios of different complexity are given. It can be seen from the figure that in scenarios of the same complexity, the adaptive genetic algorithm obtains higher benefits than the traditional genetic algorithm; in addition, for the same algorithm, the use of the combined observation method will further improve the score of the algorithm.

[0131] Figure 5 The average running time results of four algorithms in five experiments in three scenarios of different complexity are given. It can be seen from the figure that in scenarios of the same complexity, the efficiency of the adaptive genetic algorithm is higher than that of the traditional genetic algorithm; after the two algorithms introduce the design of combined observation, they can greatly shorten the running time of the algorithm and effectively improve the running efficiency of the algorithm.

[0132] In summary, the multi-satellite collaborative task planning method based on the adaptive genetic algorithm of fusion and merged observations of the present invention has the characteristics of observation benefits and fast convergence speed. While ensuring that the constraints are not violated as much as possible, it can effectively solve the problem of simultaneous planning of multiple satellites and multiple tasks.

[0133] The above are only preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A multi-satellite collaborative task dynamic scheduling planning method based on an adaptive genetic algorithm with fusion and merged observations, characterized in that: The method comprises the following steps S1: Task classification stage, based on task attributes, a set of tasks to be planned is constructed; S2: Regional task decomposition stage, analyzing the task characteristics and transforming the complex large-area observation task into a simple task; S3: Data preprocessing stage, convert all tasks into meta-tasks, and analyze and calculate the optional satellite resources and corresponding time windows of each meta-task based on the schedulable satellite resources; S4: Task merging observation processing stage, merge observation constraint detection is performed on all meta-tasks in turn, and meta-tasks that meet the merging conditions are merged and observed; S5: Optimize model construction, take the maximum observation benefit as the optimization goal, and establish the conflict resolution principle in the scheduling planning process. Establish a multi-satellite multi-task combined observation scheduling planning model; S6: Adaptive genetic algorithm is used to solve the established scheduling planning model and optimize the parameters to obtain the combined observation scheduling result that meets the task planning constraints.

2. The method for dynamic scheduling and planning of multi-satellite collaborative tasks based on an adaptive genetic algorithm with fusion and combined observations according to claim 1 is characterized in that: S1, task classification stage, constructs a set of tasks to be planned based on task attributes. The specific steps are as follows; S11, task classification, the input of the scheduling planning task mainly includes task information and satellite information. According to the task attributes, all task sets T = {T1, T2, T3, ..., T M }, classified into point task set T poi ={T 1_poi ,T 2_poi ,T 3_poi ,…,T m1_poi }、Regional task set T are ={T 1_are ,T 2_are ,T 3_are ,…,T m2_are } and the dynamic task set T dyn ={T 1_dyn ,T 2_dyn ,T 3_dyn ,…,T m3_dyn }, where the point task is the longitude (lon i ) and latitude (lat i ) is determined, then the subtask T in the task set is i_poi The specific position is represented by T i_poi =(lat i ,lon i ); The position and shape of the regional observation task are determined by the longitude and latitude of each vertex, so the subtask T in the regional task set i_are Represented as T i_are ={(lat1,lon1),…,(lat i ,lon i )|i=1,2,…,N}, S12, according to the rule that the priority of dynamic tasks is higher than the priority of point tasks, and the priority of point tasks is higher than the priority of regional tasks, corresponding priorities are set for all tasks in all task sets to be planned, wherein subtasks in the same set have the same priority.

3. The method for dynamic scheduling and planning of multi-satellite collaborative tasks based on an adaptive genetic algorithm with fusion and combined observations according to claim 1 is characterized in that ,S2, regional task decomposition stage, analyze the task characteristics, transform the complex large-area observation task into a simple task, which specifically includes the following steps; S21, read the longitude and latitude range of the regional task [lat1, lat i ], latitude range [lon1,lon i ], respectively set the discrete granularity △λ of the latitudinal decomposition of the region lat_i , Latitude discrete granularity △λ lon_i , the longitude and latitude discrete granularity conforms to the following geometric relationship According to the above geometric sampling method, the task area is divided into grids. Each sampling point is a vertex of the grid and corresponds to a position on the ground. S22, filter the sampling points within the task area. After dividing the task area by geometric sampling, there will still be some sampling points outside the area. At this time, these points need to be eliminated. Suppose the coordinates of one of the sampling points are (a, b). If lat1≤a≤lat i and lon1≤b≤lon i , then set the flag of this sampling point to true, otherwise to false, so as to determine all the sampling points in the task area. S23, for the multi-satellite scheduling planning problem of regional tasks, the regional coverage rate is an important indicator to measure the quality of the planning scheme. For each sampling point in a given satellite imaging strip and regional task, the imaging strip covers the sampling points of the discrete task area. According to the positional relationship between the sampling point and the imaging strip, it is determined whether the discrete grid is covered. If all vertices of the grid are covered, the grid is judged to be covered by the imaging strip; otherwise, the grid is judged to be uncovered. All covered grid points are counted, and then the coverage rate of the task area is calculated. S24, performing steps S21 to S23 on the regional observation task in the dynamic task to decompose it.

4. The method for dynamic scheduling and planning of multi-satellite collaborative tasks based on an adaptive genetic algorithm with fusion and combined observations according to claim 1 is characterized in that ,S3,In the data preprocessing stage, all tasks are converted into meta-tasks. According to the schedulable satellite resources, the optional satellite resources and the corresponding time windows of each meta-task are analyzed and calculated. The specific steps include: S31, after processing through steps S1-2, all tasks are decomposed into three types of point tasks with different priorities. Now all point tasks are converted into meta-tasks with three different weights. The converted meta-task set is T init =0.5·T dyn ∪0.3·T are ∪0.2·T poi , S32, after the conversion is completed, the constraints of the meta-task are consistent with the constraints of the original task. Before this, some constraint variables and parameters involved are defined: S: Satellite resource set S = {sat1, sat2, ..., sat n }, SD: Available satellite set; S_geometry: The latitude and longitude (longitude and latitude) of the subsatellite point, that is, the geographical location of the point on the Earth's surface directly below the satellite at a certain moment in orbit; sun_angle: the solar altitude angle of the subsatellite point; sunlight: the minimum solar altitude angle required for effective imaging by optical satellites; S_resolution: The spatial resolution of the satellite sensor, which indicates the size of the smallest ground feature that the sensor can identify (unit: m); startup_min: the shortest time required for the satellite to start executing the mission and end the current mission (unit: s); roll: maximum deflection angle, which indicates the angle by which satellite i can rotate left or right relative to its forward direction; orbit_max: the longest continuous working time of satellite i in one orbit around the earth; Visualfield: The width of the ground covered by the satellite sensor (unit: km) transtime: the time required for satellite attitude adjustment (unit: s); T init _resolution: the sensor spatial resolution required by the task (unit: m); T init _frequency: represents the number of times the task should be executed within the specified time; time_interval: the time interval between two effective imaging tasks (unit: h); min_time: the earliest time at which the task can be executed; max_time: the latest time at which the task can be executed; score: the maximum score of the task; TSM i : The set of mission execution time of satellite i; The start time of satellite i performing mission j; The end time of satellite i performing mission j; The time window in which satellite i performs mission j; The duration set of each task i in the meta-task set; The deflection angle of satellite i when performing mission j; S33, converts the original task constraints and satellite constraints into meta-task constraints, mainly including: 1) Synthetic aperture radar can penetrate clouds and observe all-weather without any requirements, while optical imaging satellites must meet the minimum solar altitude angle requirement when imaging: sun_angle>sunlight; 2) The imaging resolution of the selected observation satellite sensor cannot be less than the resolution required by the meta-task: T init _resolution ≤ S_resolution; 3) The time interval between two adjacent time windows of the satellite needs to satisfy the constraint of being greater than the satellite attitude adjustment time: et j-1 +trantime_i≤st j , among which, et j-1 It indicates the time when the last mission of the satellite ended, and trantime_i indicates the time when the attitude of satellite i was adjusted; 4) After the mission is planned, the specific time window for the target mission to be executed by the satellite needs to meet the constraints within the mission's effective observation period, that is: i ,et i ∈d i ; 5) The time interval between two adjacent observations of the same observation task needs to meet the minimum time interval constraint, that is: i+1 -et i ≥ time_interval; 6) The same satellite available time period cannot be assigned to two or more tasks at the same time, that is, it cannot violate the mutual exclusion constraint: Among them, C i is a set of task decision variables, which is a Boolean variable. The value 0 indicates that the task is not planned, and the value 1 indicates that the task is planned; 7) The satellite continuous working time cannot exceed the maximum continuous power-on time constraint: In the formula, sat_type is the satellite type variable, the continuous working time of opt type satellite is 300s, and the continuous working time of sar type satellite is 600s. S34, calculate the optional time window of the task. First, according to the position of each meta-task, calculate the satellite that can execute the task; further, according to the constraints of the task and the satellite in S32, calculate the optional satellite working time window of the task to form a set of optional time windows for the task.

5. The method for dynamic scheduling and planning of multi-satellite collaborative tasks based on an adaptive genetic algorithm with fusion and combined observations according to claim 1 is characterized in that ,S4, task merging observation processing stage, all meta-tasks are tested for merging observation constraints in turn, and meta-tasks that meet the merging conditions are processed for merging observations, which specifically includes the following steps; S41, the conditions for meta-tasks to conduct combined observations are mainly the following three: 1) multiple meta-tasks must be within the width of the satellite scan on the ground; 2) When performing combined observations, the resolution of the same satellite must meet the resolution requirements of multiple meta-tasks; 3) The time window after combining multiple meta-tasks must be less than the maximum single working time of the satellite. S42, processing of satellite attitude angle after combined observation: the difference in observation angle of any two meta-tasks of the observation satellite should be within the satellite field of view. In order to cover all meta-tasks for synthetic observation, the average of the maximum observation angle and the minimum observation angle is selected as the attitude angle of the satellite in the combined observation time window. S43, perform pairwise detection on all tasks with visible time windows of satellite i in the first orbital cycle. If both the deflection angle and observation time constraints are satisfied, the two meta-tasks are merged; if not, they are not merged. S43, the combined observation task obtained in S42 is combined again with the remaining meta-tasks of the round for detection. If the combined constraint condition is met, it is further combined with the meta-tasks; if not, jump to S45. S44, sequentially merge and detect the merged tasks with other tasks in the orbital cycle until the merge detection of all meta-tasks on the first orbital cycle of the satellite is completed, and then go to S45, S45, traverse the other orbits of the satellite in turn, S46, traverse all satellites in turn, S47, the merging is completed, and the merged time window and satellite attitude are calculated.

6. The method for dynamic scheduling and planning of multi-satellite collaborative tasks based on an adaptive genetic algorithm with fusion and combined observations according to claim 1 is characterized in that , S5, optimization model construction, taking the maximum observation benefit as the optimization goal, establishing the conflict resolution principle in the scheduling planning process, and establishing a multi-satellite multi-task combined observation scheduling planning model, as follows: S51, design and optimize the objective function based on the benefits of the meta-task: maxF=f(T dyn )+f(T are )+f(T poi )+f P Among them, f(T dyn ) is the dynamic task benefit; f P Negative payoff for violating constraints; f(T are ) is the regional mission income, and the specific calculation method is: Where score i_area is the revenue that regional task i deserves, p_a i is the number of sampling points actually executed after the regional task is discretized, spt i is the total number of sampling points after discretization of region i; f(T poi ) is the point task income, and the specific calculation method is: Where score i_point is the revenue that task i deserves, frequency is the frequency at which the task is executed, T init _frequency is the frequency at which the task needs to be executed. S52 adopts the conflict resolution principle combining priority and benefit, that is, the meta-task with higher weight is executed first. When the weights of two meta-tasks are the same, the task with higher benefit is executed first.

7. The method for dynamic scheduling and planning of multi-satellite collaborative tasks based on an adaptive genetic algorithm with fusion and combined observations according to claim 1 is characterized in that , S6, solving by adaptive genetic algorithm, performing parameter optimization on the established scheduling planning model, and obtaining a combined observation scheduling result that meets the task planning constraints, specifically including the following steps; S61, encoding method, uses a matrix M to represent the multi-satellite multi-task visible window, that is: Among them, the element W ij (i=1,2,...,N s ; j = 1, 2, ..., N t ) is the set of satellite visible windows for mission j, N t is the number of tasks, N S is the number of satellites, and an ordered set of visible time windows corresponding to each mission is created based on the matrix M: in The visible window of each satellite is sorted according to the visible time. The order remains unchanged during the entire calculation process. The dimension of the chromosome is two-dimensional. The first dimension represents the satellite serial number of the observation mission, and the second dimension represents the start time of the satellite to perform the mission. According to the above statement, the visible time window of the satellite for the mission is encoded as: Among them, x i =0,1,2,...,N s represents the satellite number of mission j, y i =0,1,2,...,N t Indicates the visible time window number of the satellite's mission. S62, determination of the initial solution. The initial solution is determined in chronological order based on the start time of the point tasks, and the time windows of the satellites are sorted in chronological order. Specifically, the earliest started task is preferentially assigned to the satellite that can make observations the earliest. S63, adaptive crossover and mutation operator, in order to avoid the convergence speed and accuracy of the algorithm caused by constant crossover and mutation probability, the adaptive crossover probability P is designed according to the number of population iterations and individual fitness. c and mutation probability P m , the specific calculation methods are as follows: Among them, P c1 ,P c2 is the maximum and minimum value of the initial crossover probability, f′ is the maximum fitness of both parties in the crossover, P m1 ,P m2 is the maximum and minimum value of the initial mutation probability, f is the fitness of the mutant individual, and f avg is the average fitness of the contemporary population, f max is the maximum fitness of the contemporary population, g is the number of iterations of the current population, G is the maximum number of iterations of the population, S64, selection operator, sorting the population chromosomes in descending order according to their fitness, and selecting new populations of the same size as the initial population in turn, which can reduce the population size and reduce the algorithm calculation pressure, and can also further select chromosomes with higher fitness to speed up the convergence speed. S65, solve and obtain the optimal individual, and further output the optimal multi-satellite observation scheduling combination that meets the planning constraints.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the program, the multi-satellite collaborative task dynamic scheduling planning method based on the adaptive genetic algorithm of fusion and combined observation as described in any one of claims 1 to 7 above is implemented.

9. A computer-readable storage medium having computer instructions stored thereon, characterized in that: When the computer instructions are executed by the processor, the method for dynamic scheduling and planning of multi-satellite collaborative tasks based on an adaptive genetic algorithm with fused and merged observations as described in any one of claims 1 to 7 is implemented.

Images