Multi-satellite on-orbit cooperative earth observation task planning method and system

Abstract

The application discloses a kind of multi-star on-orbit cooperation observation task planning method and system, by designing heuristic initial solution generation strategy, roulette selection algorithm based on Boltzmann selection probability, adaptive crossover operator and mutation operator based on population dispersion degree, stop condition based on population convergence coefficient, from initial solution, selection operator, crossover operator, mutation operator, stop condition etc. Various angles improve the genetic algorithm for multi-star cooperative observation task planning, compared with the genetic algorithm based on circle next cross, variation, significantly improve the execution benefit of observation task and the observation number of target point, construct the scheduling system of constellation cloud computing system, reach the purpose of on-orbit cooperation multi-star efficient execution observation task.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of satellite observation, and particularly relates to a multi-satellite on-orbit cooperative earth observation task planning method and system. BACKGROUND

[0002] Satellites fly in orbit, and can observe a two-dimensional scanning band with a certain width with the center line being a subsatellite point by using attitude maneuver. However, due to the limitations of the on-orbit movement of the satellite, the field of view angle of the satellite-borne remote sensing device, the side swing range of the satellite-borne remote sensing device and other factors, the satellite can only observe the ground target point in a limited time window, and sometimes the ground target point cannot be observed in the planning time. In the multi-satellite scenario, a target point can have multiple time windows with a satellite, and can also be invisible to multiple satellite nodes. Multi-satellite cooperative earth observation task refers to that a user specifies a batch of target points to be observed, a time interval for observation and a priority weight of each target point, and through a scheduling strategy, multiple satellite nodes are cooperated to observe the target points as much as possible and to obtain the highest observation task benefit under the condition of satisfying the satellite imaging constraints.

[0003] Multi-satellite cooperative earth observation has been proved to be an NP-hard problem, and the exponential explosion feature is very obvious, so it is difficult to obtain the optimal solution by direct calculation. The current solution method is to transform it into a combinatorial optimization problem to obtain an approximate solution. Many intelligent optimization algorithms including genetic algorithm, ant colony algorithm and particle swarm algorithm are the main means for solving the combinatorial optimization problem at present. However, due to the special scenario of multi-satellite cooperative earth observation task scheduling, the traditional intelligent optimization algorithm cannot be directly used to solve the problem, and the algorithm needs to be improved in terms of coding and decoding of scheduling scheme, formulation and improvement of operator operation, setting of iteration stopping condition and many other aspects according to the special environment of the constellation, so as to effectively improve the execution effect of multi-satellite earth observation task.

[0004] In addition, the traditional multi-satellite scheduling algorithm is simulated in the modeling system, and the applicable scheduling algorithm needs to be further applied to the distributed cluster system to cooperatively observe multiple satellites on orbit, so as to produce actual engineering and economic benefits. SUMMARY

[0005] The technical problem to be solved by the application is to provide a multi-satellite on-orbit cooperative earth observation task planning method and system to solve the problems in the prior art, improve the overall benefit of the observation task, increase the number of successfully observed target points, and realize on-orbit planning of the multi-satellite cooperative earth observation task.

[0006] The application adopts the following technical scheme:

[0007] A multi-satellite on-orbit cooperative earth observation task planning method comprises the following steps:

[0008] S1, obtain the geographic position, task priority, and task completion time interval information of each target point to be observed, access etcd through the master node to obtain real-time synchronized satellite node state data, and calculate the visible time window between each satellite node and the target point to be observed according to real-time orbit information of N satellites;

[0009] S2, generate a coding result based on the satellite node state data obtained in step S1 and the geographic position, task priority, and task completion time interval information of each target point to be observed;

[0010] S3, based on the coding result of step S2 and the visible time window of step S1, an initial solution generation strategy with task execution benefit priority is adopted, the task priority is sequenced to plan a feasible observation window for each target point under the premise of meeting the constraints, and an observation task execution time dynamic adjustment strategy is adopted to adjust the task execution time, to obtain an initial population;

[0011] S4, calculate the fitness value of the initial population obtained in step S3, and use a roulette selection method based on Boltzmann selection probability to select individuals from the initial population for genetic, and use an adaptive crossover and mutation method based on orbit circle order to obtain a new generation population;

[0012] S5, iterate the new generation population obtained in step S4 until the end, and output a scheduling scheme of task planning; the master node schedules the observation tasks of the target points to the corresponding satellite nodes through the inter-satellite link according to the scheduling scheme, and performs earth observation.

[0013] Specifically, step S2 is specifically:

[0014] First, chromosome coding is performed, each chromosome represents a feasible scheduling scheme, and each gene in the chromosome represents the scheduling of an observation target point. For N satellites in orbit, N+1 task scheduling queues are created for each chromosome, each scheduling queue corresponds to a satellite, and each target point to be observed in the scheduling queue is represented by a structure. The N+1 scheduling queue is used to store the target points that are not successfully scheduled in this task scheduling process, and the coding result is obtained.

[0015] Specifically, in step S3, the initial solution generation strategy with task execution benefit priority is specifically:

[0016] For the observation task of one-time user demand, firstly, multiple target points in the observation task are sorted according to the priority specified by the user, and then an initial solution of the algorithm is generated according to the arrangement order; for each target point, the time window visible between each satellite and the corresponding target point is obtained, and a time window is randomly selected for observation; when selecting the time window, it is first judged whether the constraint condition is met after the task scheduling to the time window corresponding satellite, if any of the constraint conditions is not met, the next time window is selected for scheduling; if all the selectable time windows do not meet the constraint condition, the corresponding target point scheduling fails.

[0017] Further, the constraint conditions include the side swing time of the satellite-borne remote sensing equipment required during observation, the stabilization time of the remote sensing equipment, and whether the on-board storage is sufficient.

[0018] Specifically, in step S3, the observation task execution time dynamic adjustment strategy is specifically:

[0019] For the observation task of the front and rear two target points scheduled to the same satellite node, the observation execution time of the first target point is set as the start time of the selected time window, when the time window of the second target point comes, if the satellite-borne remote sensing equipment side swing is not completed or the camera is not stabilized, the observation is executed after waiting for the conversion time to be completed; if the conversion activity is not completed when the selected time window of the second target point ends, the target point observation fails.

[0020] Specifically, in step S4, the roulette selection method of Boltzmann selection probability is used to select individuals from the initial population for genetic, and the Boltzmann selection probability P(i) is as follows:

[0021]

[0022] Wherein, f i is the fitness value of individual i in the population, and b is used to control the selection intensity.

[0023] Specifically, in step S4, the cross and mutation mode based on the orbit circle is specifically:

[0024] The cross rate and mutation rate are dynamically changed by measuring the dispersion degree of the current population, and the dispersion degree of the population is evaluated by the standard deviation of the fitness values of all individuals in the population; when the standard deviation σ of the parent population is less than the preset threshold, the cross and mutation probability is increased, and when the standard deviation σ of the parent population is greater than the preset threshold, the cross and mutation probability is reduced.

[0025] Further, the standard deviation σ of the parent population is as follows:

[0026]

[0027] Wherein, N is the total number of individuals in the population, and fi Let i be the fitness value of individual i in the population. The average fitness value of all individuals in the first generation of the population;

[0028] Crossover rate P c And the probability of mutation P m The adaptive update is as follows:

[0029]

[0030]

[0031] Among them, P c and P m P represents the crossover rate and mutation rate of the previous generation population. c and P m The numbers represent the crossover rate and mutation rate of the updated population, α is the update rate control factor, σ is the standard deviation of the population fitness, and d is the preset threshold.

[0032] Specifically, in step S5, the stopping condition is determined by the convergence coefficient of the population. Iteration stops when the convergence coefficient of the population is less than or equal to a preset threshold for 5 consecutive generations. The population convergence coefficient is evaluated by the average fitness of the population as follows:

[0033]

[0034] in, This represents the average fitness of the parent population. This represents the average fitness of the offspring population.

[0035] Another technical solution of the present invention is a multi-satellite on-orbit collaborative Earth observation mission planning system, comprising:

[0036] The data module acquires the geographical location, task priority, and task completion time interval information of each target point to be observed. It accesses etcd through the master control node to obtain real-time synchronized satellite node status data and calculates the visible time window between each satellite node and the target point to be observed based on the real-time orbit information of N satellites.

[0037] The encoding module generates encoding results based on the satellite node status data obtained from the data module and the geographical location, task priority, and task completion time interval information of each target point to be observed.

[0038] The population module, based on the encoding results and the visible time window, adopts an initial solution generation strategy that prioritizes task execution benefits. It plans feasible observation windows for each target point in order of task priority, while satisfying the constraints, and uses a dynamic adjustment strategy for observation task execution time to adjust the task execution time, thus obtaining the initial population.

[0039] The calculation module calculates the fitness value of the initial population obtained from the population module, and selects individuals from the initial population for inheritance using the roulette wheel selection method based on Boltzmann selection probability. The genetic results are processed using an adaptive crossover and mutation method based on orbital cycles to obtain a new generation of population.

[0040] The planning module iterates through the new generation population obtained by the calculation module until the end, and outputs the scheduling scheme of the task planning. According to the scheduling scheme, the master control node schedules the observation tasks of the target point to the corresponding satellite nodes through the inter-satellite link to carry out Earth observation.

[0041] Compared with the prior art, the present invention has at least the following beneficial effects:

[0042] This invention presents a multi-satellite on-orbit collaborative Earth observation mission planning method. Designed for multi-satellite collaborative Earth observation missions, it improves upon the genetic algorithm by employing a heuristic initial solution generation strategy prioritizing mission gains, allowing high-weight missions to be scheduled first. Benefiting from Boltzmann selection probability, dominant individuals have a higher probability of being inherited. Adaptive crossover and mutation rates enable faster convergence when population diversity is high and avoid getting trapped in local optima when population diversity is low. The convergence stopping condition ensures that individuals in the population tend to be consistent, improving average gains and enhancing the algorithm's stability. Finally, the above mission planning algorithm is used to improve the scheduling strategy of traditional cloud computing systems, achieving on-orbit scheduling of multi-satellite collaborative Earth observation missions and providing key scheduling support for building constellation cloud computing systems.

[0043] Furthermore, chromosome encoding is the primary problem to be solved using genetic algorithms. Based on real-time satellite status data, chromosomes can establish scheduling queues for satellite nodes that are flying normally and have good health checks. Each scheduling queue stores the target point tasks scheduled to be performed on that satellite node. Each task stores detailed scheduling information through a structure to facilitate the detailed output of the final scheduling scheme.

[0044] Furthermore, a heuristic initial solution generation strategy prioritizing task benefits is adopted. The target points to be scheduled are sorted according to their task weights, and high-weight tasks are selected for observation first, so that as many high-weight tasks as possible can be completed, which to a certain extent guarantees the overall benefits of Earth observation task planning.

[0045] Furthermore, whether each target point observation task can be successfully scheduled to a certain satellite depends on the following constraints: the target point observation task must be carried out within the visible time window; the capacity required for satellite imaging cannot exceed the maximum storage capacity supported by the satellite node; and when two target point observation tasks are executed before and after the same satellite, the time interval between the tasks must be greater than the attitude stabilization time of the remote sensing equipment between the two tasks.

[0046] Furthermore, since executing two tasks sequentially on the same satellite requires a certain amount of lateral switching time, the compact strategy means that for each observation task within a selected time window, the start time of the time window should be used as the start time of the observation, so as to avoid the earlier observation task on the same satellite affecting the later observation task.

[0047] Furthermore, due to the heuristic solution generation strategy and fitness calculation method, the fitness values ​​of individuals in the initial population are not significantly different. In this case, the traditional roulette wheel algorithm tends to be a general random selection algorithm, making it difficult to highlight the superior individuals. To alleviate this problem, Boltzmann selection probability is used instead of the original selection probability in the roulette wheel method, so that the superior individuals can be inherited more effectively.

[0048] Furthermore, traditional genetic algorithms use fixed crossover and mutation probabilities, set empirically. Both excessively high and low probabilities are detrimental to population evolution. Adaptive crossover and genetic operators can dynamically adjust these probabilities during algorithm iteration, enabling genetic algorithms to better balance global and local search, avoiding a reduction in the search space while accelerating convergence.

[0049] Furthermore, the population standard deviation reflects the degree of deviation between the fitness values ​​of all individuals in the population and the population mean fitness value, embodying the dispersion of the population, and is an important parameter for determining the dynamic adjustment of crossover and mutation probabilities. When updating the crossover and mutation probabilities, α is used to represent the update speed, and L represents the size of each update. When the population standard deviation is greater than the threshold, the crossover and mutation probabilities will decrease, and when the population standard deviation is less than the threshold, the crossover and mutation probabilities will increase.

[0050] Furthermore, the population convergence coefficient represents the average fitness of a generation of the population. If the average fitness remains within a certain threshold for several consecutive iterations, it means that the algorithm is converging, thus ensuring the stability of the algorithm.

[0051] In summary, this invention provides important support for the on-orbit automatic collaborative multi-node execution of observation task scheduling in constellation cloud computing systems.

[0052] The technical solution of the present invention will be further described in detail below through embodiments. Detailed Implementation

[0053] The technical solutions in the embodiments of the present invention will be clearly and completely described below. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0054] In the description of this invention, it should be understood that the terms "comprising" and "including" indicate the presence of the described features, integrals, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or collections thereof.

[0055] It should also be understood that the terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to limit the invention. As used in this specification and the appended claims, the singular forms “a,” “an,” and “the” are intended to include the plural forms unless the context clearly indicates otherwise.

[0056] It should also be further understood that the term "and / or" as used in this specification and the appended claims refers to any combination of one or more of the associated listed items and all possible combinations, and includes such combinations.

[0057] This invention provides a multi-satellite on-orbit collaborative Earth observation mission planning method. By designing a heuristic initial solution generation strategy, a roulette wheel selection algorithm based on Boltzmann selection probability, adaptive crossover and mutation operators based on population dispersion, and stopping conditions based on population convergence coefficients, this invention improves the genetic algorithm for multi-satellite collaborative Earth observation mission planning from various perspectives, including initial solution, selection operators, crossover operators, mutation operators, and stopping conditions. Compared with genetic algorithms based on cycle-based crossover and mutation, this invention can significantly improve the execution benefits of observation missions and the number of observations at target points. It also constructs a scheduler for a constellation cloud computing system to achieve the goal of efficiently executing multi-site observation missions in on-orbit collaborative multi-satellite operation.

[0058] This invention discloses a multi-satellite on-orbit collaborative Earth observation mission planning method, comprising the following steps:

[0059] S1. The constellation cloud computing system obtains the geographical location, task priority, and task completion time interval information of each target point to be observed. It accesses etcd through the master control node to obtain real-time synchronized satellite node status data, and calculates the visible time window between each satellite node and the target point to be observed based on the real-time orbit information of N satellites.

[0060] The constellation cloud computing system, based on the constellation system, utilizes lightweight virtualization technology to break down the physical barriers between individual satellite nodes. It organically integrates various resource-constrained satellite nodes in a space resource pooling manner. By establishing a task scheduling and resource awareness platform for the constellation system, it achieves fully automated processing of the entire process, including receiving observation commands, on-orbit collaborative multi-satellite Earth observation, on-orbit collaborative multi-satellite target detection, and returning detection results to the ground control station. This saves satellite-to-ground link transmission bandwidth, enhances the ability of the constellation to collaboratively complete tasks, and improves the constellation system's perception capabilities and decision-making timeliness. It serves as the "brain" for the constellation system's unified on-orbit planning and collaborative control, laying a solid foundation for my country's space information security.

[0061] S2. Generate coding results based on the satellite node status data obtained in step S1 and the geographical location, task priority, and task completion time interval information of each target point to be observed;

[0062] In a genetic algorithm, each chromosome represents a feasible solution. Similarly, in multi-satellite Earth observation, each chromosome represents a feasible scheduling scheme, and each gene in the chromosome represents an observation of a specific ground target point.

[0063] If the current satellite cloud computing system is running N satellites, then N+1 task scheduling queues are created for each satellite. Each scheduling queue corresponds to one satellite and consists of observation targets scheduled for execution on that satellite, arranged in the order of task execution time. Each observation target in the scheduling queue is represented by a structure, which includes information such as the target point's location, priority, specific execution time, and which satellite the target point was scheduled to be observed on. The N+1th scheduling queue is used to store target points that were not successfully scheduled during the current task planning process.

[0064] S3. Based on the encoding results of step S2 and the visible time window of step S1, an initial solution generation strategy prioritizing task execution benefits is adopted. Feasible observation windows are planned for each target point in order of task priority, under the premise of satisfying constraints. The task execution time is adjusted by a dynamic adjustment strategy of observation task execution time to obtain the initial population.

[0065] To improve the performance benefits of the observation task, an initial solution generation strategy oriented towards performance benefits was designed, and a time-compact strategy was adopted to observe more target points.

[0066] For a user-defined observation task, the multiple target points are first sorted according to the user-specified priority, and then the initial solution for the genetic algorithm is generated based on this sorting order. For each target point, the visibility time between each satellite in the constellation cloud computing system and that target point is calculated. A time window is randomly selected from these visibility times to perform the observation. It is important to note that when selecting a time window, it is necessary to determine whether the constraints for scheduling a particular satellite meet certain requirements, including the onboard remote sensing equipment's side-swing time, the equipment's stabilization time, and the adequacy of onboard storage. If any constraint is not met, an alternative time window must be selected for scheduling.

[0067] Furthermore, a time-compact strategy was employed to maximize the number of target point observations. Specifically, for observations of two consecutive target points within a specific scheduling queue on a chromosome, the observation time for the target point was set to the start time of the selected time window whenever possible. If the remote sensing equipment's tilting was not yet complete or stable when the time window for the later target point observation arrived, the observation was delayed until the transition time was complete. If the transition time was not completed by the end of the time window for the later target point observation, the observation of that target point failed.

[0068] S4. Calculate the fitness value of the initial population obtained in step S3, and select individuals from the initial population for inheritance using the roulette wheel selection method based on Boltzmann selection probability. Adaptive crossover and mutation based on orbital cycles are used on the inheritance results to obtain a new generation population.

[0069] The roulette wheel selection operator is commonly used in genetic algorithms. Its main idea is that the probability of a chromosome being passed on to the next generation is proportional to its fitness. Let f be the fitness value of a chromosome. i The probability that this chromosome will be inherited by the next generation is:

[0070]

[0071] Where N represents the total number of chromosomes in the first generation population.

[0072] However, due to the profit-oriented heuristic initial solution generation strategy and fitness calculation method for constellation scenarios, the fitness values ​​of chromosomes in the initial population are not significantly different. In this case, the traditional roulette wheel algorithm tends to be a general random selection algorithm, making it difficult to highlight high-quality chromosomes.

[0073] To alleviate this problem, we consider combining the Boltzmann selection algorithm with the roulette wheel selection algorithm, that is, replacing the roulette wheel selection probability with the Boltzmann selection probability. The Boltzmann selection probability is defined as follows:

[0074]

[0075] Here, b is used to control the selection intensity. When b>0, the selection probability of individuals with high fitness is increased, thereby solving the above problem.

[0076] Adaptive crossover and mutation probabilities based on population dispersion

[0077] In genetic algorithms, the crossover operator determines the algorithm's global search capability, while the mutation operator determines its local search capability. Therefore, the crossover and mutation operators profoundly influence the algorithm's trajectory. An excessively high crossover probability can disrupt the algorithm's mechanism, while an excessively low crossover probability will slow down its search speed. An excessively high mutation probability causes the algorithm to tend towards random search, while an excessively low mutation probability can easily lead to getting trapped in local optima.

[0078] Based on chromosome encoding, this invention employs a crossover and mutation method based on orbital cycles. For two chromosomes in a given generation of the population, a target point on a certain orbit is randomly selected from the scheduling queue corresponding to the same satellite for crossover. For a single chromosome in a given generation of the population, a target point on a certain orbit is randomly selected from the scheduling queue corresponding to the same satellite for mutation. However, the crossover and mutation probabilities in this operator are fixed values, relying on empirical settings, which is not conducive to population evolution. To address this, an adaptive crossover and mutation operator based on the population dispersion is designed. This operator dynamically changes the crossover and mutation rates by measuring the dispersion of the current population, allowing the genetic algorithm to better balance global and local search. The population dispersion is assessed by the standard deviation of the fitness values ​​of all chromosomes in the population, as follows:

[0079]

[0080] When the standard deviation of the parent population is less than a preset threshold, the crossover and mutation probabilities are increased; when the standard deviation of the parent population is greater than the preset threshold, the crossover and mutation probabilities are decreased.

[0081] Crossover rate P c And the probability of mutation P m The update calculation formula is as follows:

[0082] P c `= c +L

[0083] P m `= m +L

[0084]

[0085] Among them, P c and P mP represents the crossover rate and mutation rate of the previous generation population. c `and P m ` represents the crossover rate and mutation rate of the updated population, α is the update rate control factor, σ is the standard deviation of the population fitness, and d is the preset threshold.

[0086] Furthermore, the present invention limits the range of crossover rate variation to [0.8, 0.95] and the range of mutation rate variation to [0.05, 0.15].

[0087] S5. Iterate the new generation population obtained in step S4 until the end, and output the scheduling scheme of the task planning; the master control node of the constellation cloud computing system schedules the observation tasks of the target point to the corresponding satellite nodes through the inter-satellite link according to the scheduling scheme to conduct Earth observation.

[0088] This invention determines the stopping point of the algorithm iteration by using the convergence coefficient of the population. Specifically, the algorithm stops execution when the convergence coefficient of the population is less than or equal to a preset threshold for five consecutive generations.

[0089] The convergence coefficient of a population is defined as:

[0090]

[0091] in, This represents the average fitness of the parent population. This represents the average fitness of the offspring population.

[0092] In another embodiment of the present invention, a multi-satellite on-orbit collaborative Earth observation mission planning system is provided. This system can be used to implement the above-mentioned multi-satellite on-orbit collaborative Earth observation mission planning method. Specifically, the multi-satellite on-orbit collaborative Earth observation mission planning system includes a data module, an encoding module, a population module, a calculation module, and a planning module.

[0093] The data module acquires the geographical location, task priority, and task completion time interval information of each target point to be observed. It accesses etcd through the master control node to obtain real-time synchronized satellite node status data and calculates the visible time window between each satellite node and the target point to be observed based on the real-time orbit information of N satellites.

[0094] The encoding module generates encoding results based on the satellite node status data obtained from the data module and the geographical location, task priority, and task completion time interval information of each target point to be observed.

[0095] The population module, based on the encoding results and the visible time window, adopts an initial solution generation strategy that prioritizes task execution benefits. It plans feasible observation windows for each target point in order of task priority, while satisfying the constraints, and uses a dynamic adjustment strategy for observation task execution time to adjust the task execution time, thus obtaining the initial population.

[0096] The calculation module calculates the fitness value of the initial population obtained from the population module, and selects individuals from the initial population for inheritance using the roulette wheel selection method based on Boltzmann selection probability. The genetic results are processed using an adaptive crossover and mutation method based on orbital cycles to obtain a new generation of population.

[0097] The planning module iterates through the new generation population obtained by the calculation module until the end, and outputs the scheduling scheme of the task planning. According to the scheduling scheme, the master control node schedules the observation tasks of the target point to the corresponding satellite nodes through the inter-satellite link to carry out Earth observation.

[0098] In another embodiment of the present invention, a terminal device is provided, comprising a processor and a memory. The memory stores a computer program, which includes program instructions. The processor executes the program instructions stored in the computer storage medium. The processor may be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. It is the computing and control core of the terminal, suitable for implementing one or more instructions, specifically suitable for loading and executing one or more instructions to achieve corresponding method flows or corresponding functions. The processor described in this embodiment can be used for the operation of a multi-satellite on-orbit collaborative Earth observation mission planning method, including:

[0099] The system acquires the geographical location, task priority, and task completion time interval information of each target point to be observed. It obtains real-time synchronized satellite node status data via the master node accessing etcd. Based on the real-time orbit information of N satellites, it calculates the visible time window between each satellite node and the target point. Encoding results are generated based on the obtained satellite node status data and the geographical location, task priority, and task completion time interval information of each target point. Based on the encoding results and the visible time window, an initial solution generation strategy prioritizing task execution benefits is adopted. Feasible observation windows are planned for each target point according to task priority, while satisfying constraints. A dynamic adjustment strategy for observation task execution time is used to adjust the task execution time, resulting in an initial population. The fitness value of the initial population is calculated, and individuals are selected from the initial population using a roulette wheel selection method based on Boltzmann selection probability for genetics. Adaptive crossover and mutation based on orbital cycles are used on the genetic results to obtain a new generation population. The new generation population is iterated until completion, and a task planning scheduling scheme is output. According to the scheduling scheme, the master node schedules the observation tasks of the target points to the corresponding satellite nodes via inter-satellite links for Earth observation.

[0100] In another embodiment of the present invention, a storage medium is provided, specifically a computer-readable storage medium (Memory). This computer-readable storage medium is a memory device in a terminal device used to store programs and data. It is understood that the computer-readable storage medium here can include both the built-in storage medium in the terminal device and extended storage media supported by the terminal device. The computer-readable storage medium provides storage space that stores the terminal's operating system. Furthermore, this storage space also stores one or more instructions suitable for loading and execution by a processor. These instructions can be one or more computer programs (including program code). It should be noted that the computer-readable storage medium here can be high-speed RAM or non-volatile memory, such as at least one disk storage device.

[0101] One or more instructions stored in a computer-readable storage medium can be loaded and executed by the processor to implement the corresponding steps of the multi-satellite on-orbit collaborative Earth observation mission planning method in the above embodiments; one or more instructions in the computer-readable storage medium are loaded and executed by the processor in the following steps:

[0102] The system acquires the geographical location, task priority, and task completion time interval information of each target point to be observed. It obtains real-time synchronized satellite node status data via the master node accessing etcd. Based on the real-time orbit information of N satellites, it calculates the visible time window between each satellite node and the target point. Encoding results are generated based on the obtained satellite node status data and the geographical location, task priority, and task completion time interval information of each target point. Based on the encoding results and the visible time window, an initial solution generation strategy prioritizing task execution benefits is adopted. Feasible observation windows are planned for each target point according to task priority, while satisfying constraints. A dynamic adjustment strategy for observation task execution time is used to adjust the task execution time, resulting in an initial population. The fitness value of the initial population is calculated, and individuals are selected from the initial population using a roulette wheel selection method based on Boltzmann selection probability for genetics. Adaptive crossover and mutation based on orbital cycles are used on the genetic results to obtain a new generation population. The new generation population is iterated until completion, and a task planning scheduling scheme is output. According to the scheduling scheme, the master node schedules the observation tasks of the target points to the corresponding satellite nodes via inter-satellite links for Earth observation.

[0103] This invention relies on a constellation cloud computing system to conduct simulation tests. Taking my country's already launched Yaogan-6, Yaogan-11, Yaogan-13, and Yaogan-16-A satellites as examples, the initial orbital parameters of each satellite are shown in Table 1, and the performance settings of the onboard observation equipment of each satellite are shown in Table 2.

[0104] Table 1 shows the orbital root numbers of each low-Earth orbit observation node in the satellite cloud computing system.

[0105]

[0106] Table 2. Observation performance of various low-Earth orbit observation satellites

[0107]

[0108]

[0109] To enhance the realism of the experimental example, this example sets the observation period for the low-Earth orbit satellite to be from 4:00:00 on March 18, 2021 to 4:00:00 on March 19, 2021. It selects the top 100 cities from the 2020 Kearney City Index report for observation. The priority weight for each target point is selected within the range of [1, 50], with higher priority indicating greater observation benefits for that target point. Furthermore, the visibility data between the low-Earth orbit observation satellite and the ground observation target point is obtained using STK software widely used in the aerospace field to avoid complex coordinate transformations and dynamic calculations.

[0110] By setting up observation tasks with three different target points of 50, 75, and 100, a total of four sets of test cases were generated. Each set of test cases was run 20 times to test the performance of the algorithm. The best return, the number of targets completed, the average return of the task, and the execution time of the algorithm were recorded. The average value of each record was taken as the final experimental result.

[0111] Based on the above test scenario and experimental parameter settings, the improved genetic algorithm IGA proposed in this invention is compared with the genetic algorithm GA based on crossover and mutation in cycles to verify the effectiveness of this invention. The parameter settings of the IGA algorithm are shown in Table 3, and the algorithm parameter settings of the GA algorithm are shown in Table 4. The quality comparison of the two methods is shown in Table 5.

[0112] Table 3 IGA Algorithm Parameter Settings

[0113]

[0114]

[0115] Table 4 GA Algorithm Parameter Settings

[0116] Parameter Value Population size 50 Crossover probability 0.9 Mutation probability 0.1 Maximum number of iterations 400

[0117] Table 5 Comparison of experimental results between genetic algorithm and improved genetic algorithm

[0118]

[0119] As shown in the table, the improved genetic algorithm of this invention outperforms the genetic algorithm based on crossover and mutation based on cycles in multiple test examples. Compared with the latter, the overall average reward for the observation task is improved by 2.58%, the overall optimal reward is improved by 1.92%, and the number of successfully observed target points is increased by 4.84%. Benefiting from Boltzmann selection probability, dominant individuals have a greater probability of being inherited, improving the observation reward. The convergence stopping condition can make individuals in the population tend to be consistent, improving the average reward and making the algorithm more stable. Dynamically adjusting the crossover rate and mutation rate allows for faster convergence when the population diversity is high and avoids getting trapped in local optima when the population diversity is low. At the same time, these strategies correspondingly increase the algorithm's running time, but it is still within an acceptable range.

[0120] In summary, this invention provides a multi-satellite on-orbit collaborative Earth observation mission planning method and system. Based on the multi-satellite collaborative observation scenario, it improves the genetic model. Experiments have shown that this method improves the overall benefits of the observation mission, increases the number of missions to be completed, and improves the traditional distributed cloud computing system with this method as the core scheduling strategy. It realizes the on-orbit scheduling of multi-satellite Earth observation missions, provides important support for the construction of constellation cloud computing systems, and has positive engineering benefits and strategic significance.

[0121] The above content is only for illustrating the technical concept of the present invention and should not be construed as limiting the scope of protection of the present invention. Any modifications made to the technical solution based on the technical concept proposed in this invention shall fall within the scope of protection of the claims of this invention.

Claims

1. A method for planning multi-satellite on-orbit collaborative Earth observation missions, characterized in that, Includes the following steps: S1. Obtain the geographical location, task priority, and task completion time interval information of each target point to be observed. Access etcd through the master control node to obtain real-time synchronized satellite node status data. Calculate the visible time window between each satellite node and the target point to be observed based on the real-time orbit information of N satellites. S2. Generate coding results based on the satellite node status data obtained in step S1 and the geographical location, task priority, and task completion time interval information of each target point to be observed; S3. Based on the encoding results of step S2 and the visible time window of step S1, an initial solution generation strategy prioritizing task execution benefits is adopted. Feasible observation windows are planned for each target point according to task priority, while satisfying constraints. A dynamic adjustment strategy for observation task execution time is used to adjust the task execution time, resulting in an initial population. Specifically, the initial solution generation strategy prioritizing task execution benefits is as follows: For a user-defined observation task, the multiple target points in the observation task are first sorted according to the user-specified priority, and then the initial solution of the algorithm is generated according to the sorted order. For each target point, the visible time window between each satellite and the corresponding target point is obtained, and a time window is randomly selected to perform the observation. When selecting a time window, it is first determined whether the constraints are met after the current task is scheduled to the satellite corresponding to the time window. If any constraint is not met, the next time window is selected for scheduling. If all available time windows do not meet the constraints, the scheduling of the corresponding target point fails. The constraints include the side swing time of the onboard remote sensing equipment required by the satellite during observation, the stabilization time of the remote sensing equipment, and whether the onboard storage is sufficient. The specific strategy for dynamically adjusting the execution time of observation tasks is as follows: For two consecutive target point observation tasks scheduled to the same satellite node, the observation execution time of the first target point is set as the start time of the selected time window. When the time window for the second target point arrives, if the onboard remote sensing equipment has not completed its side swing or the camera is not stable, the observation will be performed after the conversion time is completed. If the conversion activity is not completed by the end of the selected time window for the second target point, the target point observation will fail. S4. Calculate the fitness values ​​of the initial population obtained in step S3, and select individuals from the initial population for inheritance using the roulette wheel selection method based on Boltzmann selection probability. Apply adaptive crossover and mutation based on orbital cycles to the inheritance results to obtain a new generation population. Again, select individuals from the initial population for inheritance using the roulette wheel selection method based on Boltzmann selection probability. as follows: in, Let i be the fitness value of individual i in the population. Used to control the intensity of selection This represents the total number of chromosomes in a generation of the population; The adaptive crossover and mutation method based on orbital cycles is specifically as follows: The crossover and mutation rates are dynamically adjusted to measure the dispersion of the current population. The dispersion of the population is assessed by the standard deviation of the fitness values ​​of all individuals in the population; when the standard deviation of the parent population... When the value is less than the preset threshold, the crossover and mutation probabilities are increased, and the standard deviation of the parent population is increased. When the value exceeds a preset threshold, the probability of crossover and mutation is reduced, and the standard deviation of the parent population is lowered. as follows: in, The total number of individuals in the population. Let i be the fitness value of individual i in the population. The average fitness value of all individuals in the first generation of the population; Cross rate and mutation probability The adaptive update is as follows: in, and This represents the crossover rate and mutation rate of the previous generation population. and This represents the crossover rate and mutation rate of the updated population. It is the update speed control factor. It is the standard deviation of population fitness. The preset threshold; S5. Iterate the new generation population obtained in step S4 until the end, and output the scheduling scheme of the task planning; according to the scheduling scheme, the master node schedules the observation tasks of the target point to the corresponding satellite nodes through the inter-satellite link to carry out Earth observation.

2. The multi-satellite on-orbit collaborative Earth observation mission planning method according to claim 1, characterized in that, Step S2 is as follows: First, chromosome encoding is performed. Each chromosome represents a feasible scheduling scheme, and each gene in the chromosome represents a scheduling of an observation target point. For N satellites in orbit, N+1 task scheduling queues are created for each chromosome. Each scheduling queue corresponds to one satellite. Each target point to be observed in the scheduling queue is represented by a structure. The N+1th scheduling queue is used to store target points that were not successfully scheduled during the current task scheduling process, thus obtaining the encoding result.

3. The multi-satellite on-orbit collaborative Earth observation mission planning method according to claim 1, characterized in that, In step S5, the stopping condition is determined by the convergence coefficient of the population. Iteration stops when the convergence coefficient of the population is less than or equal to a preset threshold for 5 consecutive generations. The population convergence coefficient is evaluated by the average fitness of the population. as follows: in, This represents the average fitness of the parent population. This represents the average fitness of the offspring population.

4. A multi-satellite on-orbit collaborative Earth observation mission planning system, characterized in that, include: The data module acquires the geographical location, task priority, and task completion time interval information of each target point to be observed. It accesses etcd through the master control node to obtain real-time synchronized satellite node status data and calculates the visible time window between each satellite node and the target point to be observed based on the real-time orbit information of N satellites. The encoding module generates encoding results based on the satellite node status data obtained from the data module and the geographical location, task priority, and task completion time interval information of each target point to be observed. The population module, based on the encoding results and the visible time window, adopts an initial solution generation strategy that prioritizes task execution benefits. For each target point, a feasible observation window is planned according to task priority while satisfying constraints. A dynamic adjustment strategy for observation task execution time is used to adjust the task execution time, resulting in the initial population. Specifically, the initial solution generation strategy prioritizing task execution benefits is as follows: For a user-defined observation task, the multiple target points in the observation task are first sorted according to the user-specified priority, and then the initial solution of the algorithm is generated according to the sorted order. For each target point, the visible time window between each satellite and the corresponding target point is obtained, and a time window is randomly selected to perform the observation. When selecting a time window, it is first determined whether the constraints are met after the current task is scheduled to the satellite corresponding to the time window. If any constraint is not met, the next time window is selected for scheduling. If all available time windows do not meet the constraints, the scheduling of the corresponding target point fails. The constraints include the side swing time of the onboard remote sensing equipment required by the satellite during observation, the stabilization time of the remote sensing equipment, and whether the onboard storage is sufficient. The specific strategy for dynamically adjusting the execution time of observation tasks is as follows: For two consecutive target point observation tasks scheduled to the same satellite node, the observation execution time of the first target point is set as the start time of the selected time window. When the time window for the second target point arrives, if the onboard remote sensing equipment has not completed its side swing or the camera is not stable, the observation will be performed after the conversion time is completed. If the conversion activity is not completed by the end of the selected time window for the second target point, the target point observation will fail. The computation module calculates the fitness values ​​of the initial population obtained from the population module, and uses a roulette wheel selection method based on Boltzmann selection probability to select individuals from the initial population for inheritance. The inheritance results are then processed using adaptive crossover and mutation based on orbital cycles to obtain a new generation population. The Boltzmann selection probability is then used to select individuals from the initial population for inheritance. as follows: in, Let i be the fitness value of individual i in the population. Used to control the intensity of selection This represents the total number of chromosomes in a generation of the population; The adaptive crossover and mutation method based on orbital cycles is specifically as follows: The crossover and mutation rates are dynamically adjusted to measure the dispersion of the current population. The dispersion of the population is assessed by the standard deviation of the fitness values ​​of all individuals in the population; when the standard deviation of the parent population... When the value is less than the preset threshold, the crossover and mutation probabilities are increased, and the standard deviation of the parent population is increased. When the value exceeds a preset threshold, the probability of crossover and mutation is reduced, and the standard deviation of the parent population is lowered. as follows: in, The total number of individuals in the population. Let i be the fitness value of individual i in the population. The average fitness value of all individuals in the first generation of the population; Cross rate and mutation probability The adaptive update is as follows: in, and This represents the crossover rate and mutation rate of the previous generation population. and This represents the crossover rate and mutation rate of the updated population. It is the update speed control factor. It is the standard deviation of population fitness. The preset threshold; The planning module iterates through the new generation population obtained by the calculation module until the end, and outputs the scheduling scheme of the task planning. According to the scheduling scheme, the master control node schedules the observation tasks of the target point to the corresponding satellite nodes through the inter-satellite link to carry out Earth observation.

5. The multi-satellite on-orbit collaborative Earth observation mission planning system according to claim 4, characterized in that, In the encoding module, chromosome encoding is performed first. Each chromosome represents a feasible scheduling scheme, and each gene in the chromosome represents a scheduling of an observation target point. For N satellites in orbit, N+1 task scheduling queues are created for each chromosome. Each scheduling queue corresponds to one satellite. Each target point to be observed in the scheduling queue is represented by a structure. The N+1th scheduling queue is used to store target points that were not successfully scheduled during the current task scheduling process, thus obtaining the encoding result.

6. The multi-satellite on-orbit collaborative Earth observation mission planning system according to claim 4, characterized in that, In the planning module, the stopping condition is determined by the population's convergence coefficient. Iteration stops when the convergence coefficient of the population is less than or equal to a preset threshold for five consecutive generations. The population's average fitness is used to evaluate the convergence coefficient. as follows: in, This represents the average fitness of the parent population. This represents the average fitness of the offspring population.

Images