Movatterモバイル変換


[0]ホーム

URL:


CN119759455B - Task rapid scheduling method and device under satellite network fault - Google Patents

Task rapid scheduling method and device under satellite network fault
Download PDF

Info

Publication number
CN119759455B
CN119759455BCN202510272535.2ACN202510272535ACN119759455BCN 119759455 BCN119759455 BCN 119759455BCN 202510272535 ACN202510272535 ACN 202510272535ACN 119759455 BCN119759455 BCN 119759455B
Authority
CN
China
Prior art keywords
satellite
population
task
initial
computing
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active
Application number
CN202510272535.2A
Other languages
Chinese (zh)
Other versions
CN119759455A (en
Inventor
姚海鹏
王雅琪
王富
张琦
许书彬
赵君
张尼
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Beijing University of Posts and Telecommunications
Original Assignee
Beijing University of Posts and Telecommunications
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Beijing University of Posts and TelecommunicationsfiledCriticalBeijing University of Posts and Telecommunications
Priority to CN202510272535.2ApriorityCriticalpatent/CN119759455B/en
Publication of CN119759455ApublicationCriticalpatent/CN119759455A/en
Application grantedgrantedCritical
Publication of CN119759455BpublicationCriticalpatent/CN119759455B/en
Activelegal-statusCriticalCurrent
Anticipated expirationlegal-statusCritical

Links

Landscapes

Abstract

Translated fromChinese

本发明提供了一种卫星网络故障下的任务快速调度方法和装置,涉及卫星通信的技术领域,方法应用于卫星计算网络中的全局观测卫星,全局观测卫星获取网络中的全局信息,并在确定存在故障计算卫星时,获取目标故障计算卫星上所有未完成任务的任务信息;利用遗传算法生成初始种群,种群中的每个个体表示一种联合卸载策略;根据全局信息、任务信息、遗传算法和粒子群算法对初始种群进行迭代更新,直至达到预设迭代次数,以将全局最优个体对应的联合卸载策略作为目标故障计算卫星的任务调度策略。该方法的执行主体为应用遗传算法和粒子群算法的全局观测卫星,与强化学习算法相比,在保障卫星网络故障下的任务快速调度的前提下对卫星的内存要求更低。

The present invention provides a method and device for rapid scheduling of tasks under satellite network failure, which relates to the technical field of satellite communication. The method is applied to a global observation satellite in a satellite computing network. The global observation satellite obtains global information in the network, and when it is determined that there is a faulty computing satellite, obtains task information of all unfinished tasks on the target faulty computing satellite; uses a genetic algorithm to generate an initial population, and each individual in the population represents a joint unloading strategy; the initial population is iteratively updated according to global information, task information, genetic algorithm and particle swarm algorithm until a preset number of iterations is reached, so as to use the joint unloading strategy corresponding to the global optimal individual as the task scheduling strategy of the target faulty computing satellite. The execution subject of the method is a global observation satellite that applies a genetic algorithm and a particle swarm algorithm. Compared with a reinforcement learning algorithm, the method has lower memory requirements for the satellite under the premise of ensuring rapid scheduling of tasks under satellite network failure.

Description

Task rapid scheduling method and device under satellite network fault
Technical Field
The invention relates to the technical field of satellite communication, in particular to a method and a device for rapidly scheduling tasks under satellite network faults.
Background
Satellite computing networks are typically composed of satellites distributed in different orbits, equipped with computing devices and communication modules, capable of data processing and information interaction in space. However, satellite networks are subject to environmental conditions, operating systems, limited resources, etc., and risk node failure. The satellite network fault has extremely serious influence, so that the satellite network fault is agilely dealt with, and the network robustness is ensured.
In the prior art, in order to cope with satellite node faults, an intelligent algorithm such as reinforcement learning is selected to be deployed on a satellite node, but the deployment method needs the satellite node to store a large amount of experience data of interaction between an intelligent body and the environment, and as the learning process advances, the data volume is continuously increased, which is a heavy burden on satellites with intense memory resources.
Disclosure of Invention
The invention aims to provide a method and a device for rapidly dispatching tasks under satellite network faults, which are used for reducing the memory requirements of satellites on the premise of ensuring rapid dispatching of tasks under satellite network faults.
The method comprises the steps of obtaining global information in a satellite computing network, wherein the satellite computing network comprises global observation satellites, a cloud platform, a plurality of computing satellites and a plurality of ground terminals, the global information comprises state information of the cloud platform, state information of the computing satellites, state information of the ground terminals, inter-satellite link information and satellite-ground link information, task information of all incomplete tasks on target fault computing satellites is obtained when the fact that the fault computing satellites exist in the satellite computing network is determined, the target fault computing satellites represent any one satellite of all the fault computing satellites, a genetic algorithm is utilized to initialize a population to obtain an initial population based on the task information of all the incomplete tasks, each individual in the initial population represents a combined unloading strategy, the combined unloading strategy represents a set of the task unloading strategy of all the incomplete tasks, the initial population is iteratively updated based on the global information, the task information, the genetic algorithm and the particle swarm algorithm until the number of times is reached, the combined unloading strategy corresponding to the optimal individual is used as the target fault computing strategy, the combined unloading strategy is generated, the algorithm is optimized based on the fact that the individual sub-population is updated based on the genetic algorithm, and the algorithm is optimized according to the relative order of the initial population, and the offspring of the algorithm is optimized to the initial population is generated based on the relative order of the algorithm.
The method comprises the steps of S201, selecting two parent individuals from a current population by using a roulette algorithm, S202, carrying out cross operation on the two parent individuals to obtain two crossed parent individuals, S203, carrying out mutation on the two crossed parent individuals according to self-adaptive mutation probability to obtain two initial child individuals corresponding to the two parent individuals, and repeatedly executing the steps S201 to S203 until the number of the initial child individuals is greater than or equal to that of the individuals in the initial population, wherein the set of all the initial child individuals is used as the initial child population.
Optionally, performing cross operation on two parent individuals, wherein the cross operation comprises randomly generating a start cross point and a stop cross point of a gene segment to be crossed, and exchanging the gene segments of the two parent individuals between the start cross point and the stop cross point to obtain two crossed parent individuals, wherein the gene segments represent a plurality of continuous gene points, and each gene point represents a task unloading strategy of incomplete tasks in the combined unloading strategy.
Alternatively, the two crossed parent individuals are mutated according to the adaptive mutation probability to obtain two initial offspring individuals corresponding to the two parent individuals, which comprises the following steps of using the formulaCalculating the mutation probability of the parent individuals in the current population to obtain the self-adaptive mutation probability,Indicating a preset initial probability of variation of the sample,The predetermined final probability of variation is indicated,Representing the current number of iterations of the population,The method comprises the steps of representing preset iteration times, randomly generating random numbers between 0 and 1 for target parent individuals, wherein the target parent individuals represent any one of two crossed parent individuals, mutating at least one gene point of the target parent individuals if the random numbers are smaller than adaptive mutation probability to obtain initial child individuals corresponding to the target parent individuals, and taking the target parent individuals as the initial child individuals corresponding to the target parent individuals if the random numbers are larger than or equal to the adaptive mutation probability.
Optionally, the initial offspring population generated based on the genetic algorithm is optimized by using a particle swarm algorithm, and the method comprises the steps of mapping each initial offspring individual in the initial offspring population to particles in the particle swarm algorithm to obtain a first particle swarm, updating positions and speeds of all particles in the first particle swarm to obtain a second particle swarm, and mapping each particle in the second particle swarm to an individual in the genetic algorithm to obtain an optimized offspring population.
Optionally, the task unloading strategy of each incomplete task comprises one of unloading to a cloud platform, unloading to a healthy computing satellite, unloading to a ground terminal, unloading to the cloud platform through the healthy computing satellite, mapping each initial child individual in the initial child population to particles in a particle swarm algorithm, wherein the method comprises the steps of matching a corresponding probability distribution interval for each joint unloading strategy based on the total number of incomplete tasks on the target fault computing satellite and the selectable number of the task unloading strategies, determining a probability distribution interval corresponding to the joint unloading strategy corresponding to the target initial child individual to obtain a target probability distribution interval, wherein the target initial child individual represents any one body in the initial child population, taking the interval median of the target probability distribution interval as the position of the particles corresponding to the target initial child individual, and obtaining the speed of the particles in a random initialization mode.
Alternatively, the formula is usedCalculating the fitness function value of the individual, wherein,Representing the transmission delay of the ith outstanding task in the combined offloading policy,Representing the computational latency of the ith incomplete task in the joint offload policy,Representing the total number of outstanding tasks on the target failure computing satellite.
The invention provides a task rapid scheduling device under a satellite network fault, which is applied to a global observation satellite in a satellite computing network, and comprises a first acquisition module, an initialization module, an updating and determining module, and a second acquisition module, wherein the first acquisition module is used for acquiring global information in the satellite computing network, the satellite computing network comprises the global observation satellite, a cloud platform, a plurality of computing satellites and a plurality of ground terminals, the global information comprises state information of the cloud platform, state information of the computing satellites, state information of the ground terminals, inter-satellite link information and satellite-ground link information, the second acquisition module is used for acquiring task information of all incomplete tasks on a target fault computing satellite when the fault computing satellite exists in the satellite computing network, the target fault computing satellite represents any one satellite in the fault computing satellite, the initialization module is used for initializing a population by utilizing a genetic algorithm based on the task information of all incomplete tasks, each individual in the initial population represents a joint unloading strategy, the joint unloading strategy represents a set of the task unloading strategy of all tasks, the updating and the initial population is updated by the updating and determining module based on the global information, the task information, the genetic algorithm and the particle swarm algorithm until the optimal algorithm is used as an optimal iteration algorithm in the initial population, the optimal population is generated by using the iterative algorithm based on the optimal algorithm, and constructing a next generation population by utilizing the optimized offspring population and individuals with the front fitness function values in the initial offspring population in descending order, wherein the fitness function values of the individuals are inversely related to the task processing time delay of the corresponding combined unloading strategy.
In a third aspect, the present invention provides an electronic device, including a memory, and a processor, where the memory stores a computer program that can be executed on the processor, and when the processor executes the computer program, the method for quickly scheduling tasks under satellite network failure in any of the foregoing embodiments is implemented.
In a fourth aspect, the present invention provides a computer readable storage medium storing computer instructions that, when executed by a processor, implement a method for task fast scheduling in the event of a satellite network failure according to any of the preceding embodiments.
The invention provides a task rapid scheduling method under satellite network faults, which provides a satellite computing network, and comprises a global observation satellite, a cloud platform, a plurality of computing satellites and a plurality of ground terminals, wherein the task rapid scheduling method is applied to the global observation satellite in the satellite computing network, the global observation satellite acquires global information in the satellite computing network, when determining that a fault computing satellite exists in the network, task information of all incomplete tasks on a target fault computing satellite is acquired, an initial population is generated by utilizing a genetic algorithm, each individual in the population represents a combined unloading strategy, and then the initial population is iteratively updated according to the global information, the task information, the genetic algorithm and a particle swarm algorithm until the preset iteration times are reached, and the combined unloading strategy corresponding to the global optimal individual is used as the task scheduling strategy of the target fault computing satellite. The execution main body of the method is a global observation satellite, and the genetic algorithm and the particle swarm algorithm are applied to solve the task scheduling problem. And the combination application of the genetic algorithm and the particle swarm algorithm can effectively improve the convergence stability of the algorithm.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings that are needed in the description of the embodiments or the prior art will be briefly described, and it is obvious that the drawings in the description below are some embodiments of the present invention, and other drawings can be obtained according to the drawings without inventive effort for a person skilled in the art.
Fig. 1 is a flowchart of a task fast scheduling method under satellite network failure according to an embodiment of the present invention;
Fig. 2 is a diagram of a task scheduling architecture of a satellite network in a satellite computing network according to an embodiment of the present invention;
FIG. 3 is a system architecture diagram of a satellite network failure detection and recovery mechanism implemented in a method according to an embodiment of the present invention;
FIG. 4 is a graph showing the performance of the improved adaptive probability genetic algorithm according to the embodiment of the present invention compared with that of the conventional genetic algorithm;
FIG. 5 is a functional block diagram of a task fast scheduling device under satellite network failure according to an embodiment of the present invention;
fig. 6 is a schematic diagram of an electronic device according to an embodiment of the present invention.
Detailed Description
For the purpose of making the objects, technical solutions and advantages of the embodiments of the present invention more apparent, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention, and it is apparent that the described embodiments are some embodiments of the present invention, but not all embodiments of the present invention. The components of the embodiments of the present invention generally described and illustrated in the figures herein may be arranged and designed in a wide variety of different configurations.
Thus, the following detailed description of the embodiments of the invention, as presented in the figures, is not intended to limit the scope of the invention, as claimed, but is merely representative of selected embodiments of the invention. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
Some embodiments of the present invention are described in detail below with reference to the accompanying drawings. The following embodiments and features of the embodiments may be combined with each other without conflict.
Example 1
Computing offloading is a technique to transfer computing tasks from a local device to other devices or servers with more computing power for processing. Computation offloading after satellite network failure is an effective strategy to cope with satellite network failure, aimed at ensuring continuity of service and quality of service by transferring computation tasks from the failed satellite network device to other available computation resources.
The embodiment of the invention provides a task rapid scheduling method under satellite network faults, which is applied to global observation satellites in a satellite computing network and is shown in fig. 1, and specifically comprises the following steps:
Step S102, global information in a satellite computing network is acquired.
Fig. 2 is a diagram of a task scheduling architecture of a satellite network in a satellite computing network, where the satellite computing network includes a global observation satellite, a cloud platform, a plurality of computation satellites, and a plurality of ground terminals, and the global observation satellite is used for observing global information in the network, where the global information includes state information of the cloud platform, state information of the computation satellites, state information of the ground terminals, inter-satellite link information, and inter-satellite link information, as shown in fig. 2.
In the embodiment of the invention, the state information of the cloud platform comprises computing resources of the cloud platform(I.e., computing frequency), computing satellite state information including idle computing resources and node operational states, computing satellite j state information may be expressed as: , wherein,Representing the free computing resources of satellite node j,Indicating the operational status of the satellite node j,=0 Indicates that satellite node j is faulty, i.e., satellite node j is a faulty computing satellite; Indicating that satellite node j is operating properly, i.e., satellite node j is a healthy computing satellite. The state information of the ground terminal comprises the computing resources of the ground terminal. The inter-satellite link information represents a data transmission rate of a link between a satellite node in the satellite computing network, and the earth-satellite link information represents a data transmission rate of a link between the satellite node and the earth terminal in the satellite computing network.
Fig. 3 is a system architecture diagram of a satellite network fault detection and recovery mechanism implemented in the method according to the embodiment of the present invention, where, as shown in fig. 3, a global observation satellite includes a global observation module, a scheduling decision module, a calculation satellite includes a fault prediction module and a task calculation module, and the global observation satellite periodically and continuously detects a satellite node state, and collects and updates global network information. Meanwhile, the calculation satellite can conduct fault prediction and actively send fault information to the global observation satellite. The embodiment of the invention essentially applies a hybrid mechanism of global observation and active report, and provides guarantee for satellite fault quick judgment and task quick scheduling.
When determining that a fault computing satellite exists, the global observation satellite can conduct task scheduling decision on tasks which are not completed, and a scheduling instruction is sent to the computing satellite. The computing satellite can unload the tasks to different positions for processing according to the scheduling instruction, so that reliable execution of the tasks is ensured. Compared with the mode of carrying out scheduling decision through the ground control center, the method in the embodiment of the invention is executed by the global observation satellite, so that the instruction transmission delay can be reduced, and the response speed can be improved. Specific embodiments of the scheduling decision module are described in detail below.
Step S104, when determining that a fault computing satellite exists in the satellite computing network, acquiring task information of all unfinished tasks on the target fault computing satellite.
Wherein the target failure computing satellite represents any one of all the failure computing satellites.
Specifically, if the global observation satellite determines that a fault calculation satellite exists in the network, task information of all incomplete tasks on each fault calculation satellite needs to be further acquired, wherein the i-th incomplete task on the target fault calculation satelliteCan be expressed as:, Representation ofIs used for the data size of the (c) data,Representing executionThe number of CPU cycles required.
And step S106, initializing the population by using a genetic algorithm based on the task information of all the incomplete tasks to obtain an initial population.
In order to solve the problem of rapid scheduling of incomplete calculation tasks of a fault calculation satellite in satellite network faults, the embodiment of the invention provides a genetic algorithm based on an improved self-adaptive probability of particle swarm optimization so as to jointly optimize a multi-task unloading decision. First, the global observation satellite generates an initial population using a genetic algorithm, wherein each individual in the initial population represents a joint offloading policy that represents a set of task offloading policies for all outstanding tasks. In the embodiment of the invention, the task unloading strategy comprises four steps of unloading to a cloud platform, unloading to a healthy computing satellite, unloading to a ground terminal and unloading to the cloud platform through the healthy computing satellite.
For example, if there are M outstanding tasks on the target failure computing satellite, each outstanding task having 4 task offloading policies, then the target failure computing satellite is in commonA possible joint offloading strategy. Thus, Y individuals in the initial population are slavesY joint unloading strategies which are randomly extracted from the feasible joint unloading strategies. The object of the embodiment of the invention is thatAnd the optimal combined unloading strategy is explored from the feasible combined unloading strategies.
And S108, carrying out iterative updating on the initial population based on the global information, the task information, the genetic algorithm and the particle swarm algorithm until the preset iterative times are reached, and taking the joint unloading strategy corresponding to the global optimal individual as a task scheduling strategy of the target fault calculation satellite.
After the initial population is obtained, if only genetic algorithm is adopted, the globally optimal individual can be determined by iteratively updating the population according to the capacity of global searching in the solution space. However, the embodiment of the invention improves the genetic algorithm based on the particle swarm algorithm, and the particle swarm algorithm has high-efficiency local exploration capability and can improve the overall performance of the algorithm, so that the robustness and the adaptability of the method provided by the embodiment of the invention are further enhanced, and the convergence speed is increased.
Specifically, in the process of updating each generation of population, an initial offspring population generated based on a genetic algorithm is optimized by using a particle swarm algorithm, so that individuals with the optimized offspring population and the front fitness function value in the initial offspring population are sorted in descending order to construct a next generation population. That is, after the current population is updated through the genetic algorithm to obtain the initial offspring population, instead of directly using the current population as the next generation population, the offspring population needs to be further optimized through the particle swarm algorithm to obtain the optimized offspring population, the optimized offspring population is combined with the initial offspring population, the fitness function value of each individual in the combined population is calculated, and then relatively excellent individuals are screened from the combined population according to the fitness function value to construct the next generation population.
In the embodiment of the invention, the fitness function value of the individual is inversely related to the task processing time delay of the corresponding combined unloading strategy. The task processing time delay of the combined unloading strategy is larger, the fitness function value of the corresponding individual is smaller, and on the contrary, the task processing time delay of the combined unloading strategy is smaller, the fitness function value of the corresponding individual is larger. Thus, if there are X sub-individuals in the pooled population, the fitness function values of the X sub-individuals are sorted in descending order and the first Y sub-individuals are selected as individuals in the next generation population, where Y represents the number of individuals in the initial population.
The embodiment of the invention provides a task rapid scheduling method under a satellite network fault, which provides a satellite computing network, and comprises a global observation satellite, a cloud platform, a plurality of computing satellites and a plurality of ground terminals, wherein the task rapid scheduling method is applied to the global observation satellite in the satellite computing network, the global observation satellite acquires global information in the satellite computing network, when determining that a fault computing satellite exists in the network, task information of all incomplete tasks on a target fault computing satellite is acquired, an initial population is generated by utilizing a genetic algorithm, each individual in the population represents a combined unloading strategy, and then the initial population is iteratively updated according to the global information, the task information, the genetic algorithm and a particle swarm algorithm until the preset iteration times are reached, so that the combined unloading strategy corresponding to the global optimal individual is used as the task scheduling strategy of the target fault computing satellite. The execution main body of the method is a global observation satellite, and the genetic algorithm and the particle swarm algorithm are applied to solve the task scheduling problem. And the combination application of the genetic algorithm and the particle swarm algorithm can effectively improve the convergence stability of the algorithm.
In an alternative embodiment, the generation of the initial population of offspring is based on a genetic algorithm, comprising in particular the steps of:
step S201, selecting two parent individuals from the current population by using a roulette algorithm.
Specifically, the total fitness of the current population is calculated firstSetting the fitness function value of the y-th parent individual in the current population to be expressed asThen the total fitness of the current population is: where Y represents the size of the population, i.e., the total number of individuals in the population.
Next atGenerates a random number pick within the range of values,Representing the accumulated fitness of the first k parent individuals in the current population, if the random number pick satisfies the following formula: the kth parent individual is selected. With reference to the above manner, two random numbers are generated and the cumulative fitness range in which they reside is determined to select two parent individuals from the current population.
Step S202, performing cross operation on the two parent individuals to obtain two crossed parent individuals.
In order to increase diversity of population and promote global searching capability, after two parent individuals are selected, the two parent individuals are subjected to cross operation to obtain two crossed parent individuals. The embodiment of the invention does not limit the mode of the cross operation specifically, and a user can select according to actual requirements, for example, single-point cross, multi-point cross, uniform cross and the like.
And step S203, respectively mutating the two crossed parent individuals according to the adaptive mutation probability to obtain two initial child individuals corresponding to the two parent individuals.
In order to improve species diversity and further enhance global retrieval capability of an algorithm, after two crossed parent individuals are obtained, the embodiment of the invention respectively carries out mutation on the two crossed parent individuals based on self-adaptive mutation probability, so as to obtain two initial offspring individuals. The adaptive variation probability changes along with the iteration times of the population, and the more the iteration times of the population are, the smaller the adaptive variation probability is, so that the convergence speed of the algorithm is increased.
Step S201 to step S203 are repeatedly performed until the number of initial child individuals is greater than or equal to the number of individuals in the initial population, and the set of all initial child individuals is taken as the initial child population.
Two initial offspring individuals can be generated by executing a round of steps S201 to S203, in order to obtain all the initial offspring individuals, the steps need to be repeatedly executed, if the number Y of the individuals in the initial population is even, the number of times of the repeated execution is Y/2, if the number Y of the individuals in the initial population is odd, (y+1)/2, and correspondingly, the number of the individuals in all the offspring populations can be randomly deleted by one initial offspring individual or by other means to maintain the population size, and the population size can be changed into y+1, wherein the change of the population number does not affect the performance of the algorithm.
In an alternative embodiment, the step S202 performs a crossover operation on two parent individuals, and specifically includes the following steps:
Step S2021, randomly generating a start crossover point and a stop crossover point of the gene fragment to be crossed.
Step S2022, exchanging the gene segments of the two parent individuals between the initial cross point and the final cross point to obtain two crossed parent individuals, wherein the gene segments represent a plurality of continuous gene points, and each gene point represents a task unloading strategy of an unfinished task in the combined unloading strategy.
Based on the description, the embodiment of the invention adopts a multi-point crossing mode to carry out crossing operation on two parent individuals, and the gene segments of the parent individuals are exchanged in sections by randomly selecting two crossing points, so that the gene diversity can be better maintained, and the premature sinking into local optimum is avoided.
For ease of understanding, the following illustrates that if a total of M gene loci are included in a parent, the randomly generated starting and ending crossover points are 3 and 7, respectively, then the multi-point crossover of two parent individuals corresponds to the exchange of consecutive gene locus 3,4,5,6,7 corresponding gene segments.
In an alternative embodiment, in step S203, two crossed parent individuals are mutated according to the adaptive mutation probability, obtaining two initial offspring individuals corresponding to the two parent individuals, specifically comprising the following steps:
Step S2031, using the formulaCalculating the mutation probability of the parent individuals in the current population to obtain the self-adaptive mutation probability,Indicating a preset initial probability of variation of the sample,The predetermined final probability of variation is indicated,Representing the current number of iterations of the population,Representing a preset number of iterations.
Step S2032, randomly generating a random number between 0 and 1 for a target parent individual, wherein the target parent individual represents any one of the two crossed parent individuals.
Step S2033, if the random number is smaller than the adaptive mutation probability, mutating at least one gene point of the target parent individual to obtain an initial child individual corresponding to the target parent individual.
Step S2034, if the random number is greater than or equal to the adaptive mutation probability, using the target parent as the corresponding initial child.
Specifically, the embodiment of the invention adopts a self-adaptive variation probability function, the variation probability can be dynamically adjusted according to the current iteration times of the population as known by the expression of the self-adaptive variation probability, and the self-adaptive variation probability mechanism can be used for carrying out variation on individuals based on larger probability at the initial stage of an algorithm, so that species diversity is improved, individual variation probability is reduced at the later stage, and the convergence rate is accelerated.
In an alternative embodiment, the particle swarm algorithm is used to optimize an initial population of offspring generated based on the genetic algorithm, comprising the steps of:
Step S301, mapping each initial offspring individual in the initial offspring population to a particle in the particle swarm algorithm to obtain a first particle swarm.
Step S302, the positions and the speeds of all particles in the first particle swarm are updated to obtain a second particle swarm.
Step S303, mapping each particle in the second particle swarm to an individual in the genetic algorithm to obtain an optimized offspring population.
Specifically, the embodiment of the invention introduces a particle swarm algorithm to finely explore the initial offspring population, and can quickly converge to the vicinity of the local optimal solution through the movement and information sharing of particles in the solution space. Specifically, for each initial offspring population, a particle swarm mode is adopted to perform local exploration, each initial offspring individual in the population is firstly required to be mapped to particles in a particle swarm algorithm, the mapping method is not limited in detail, and a user can select according to actual requirements.
After the first particle swarm is obtained, the position and the speed of each particle are updated by adopting a speed update formula and a position update formula of the particles in a particle swarm algorithm of the prior art to obtain a second particle swarm, and finally, all the particles in the second particle swarm are mapped into a genetic algorithm by using a method of mapping the initial offspring swarm into the opposite of the first particle swarm, so that the optimized offspring swarm is obtained.
In an alternative implementation manner, the task offloading strategy of each incomplete task comprises one of offloading to a cloud platform, offloading to a healthy computing satellite, offloading to a ground terminal, offloading to the cloud platform through the healthy computing satellite, and mapping each initial child individual in the initial child population to a particle in a particle swarm algorithm in the step S301, wherein the method specifically comprises the following steps:
Step S3011, calculating a total number of outstanding tasks on the satellite and an optional number of task offloading policies based on the target failure, and matching a corresponding probability distribution interval for each joint offloading policy.
Step S3012, determining a probability distribution interval corresponding to a combined unloading strategy corresponding to a target initial child individual to obtain a target probability distribution interval, wherein the target initial child individual represents any one of initial child populations.
And step S3013, taking the interval median value of the target probability distribution interval as the position of the particle corresponding to the target initial offspring individual, and obtaining the speed of the particle by a random initialization mode.
For ease of understanding, referring to the example above, assuming there are M outstanding tasks on the target failure computing satellite, each outstanding task having 4 task offloading policies, then the target failure computing satellite is in commonA possible joint offloading strategy. Based on this, it can be determined thatThe probability distribution intervals are:,,... . And then carrying out matching binding on each joint unloading strategy and each probability distribution interval.
If the probability distribution interval corresponding to the combined unloading strategy corresponding to the target initial offspring individual isMedian value of intervalAnd the position of the particle corresponding to the target initial offspring individual is obtained. And the velocity of the particles may be represented as a vector whose components represent the direction and magnitude of adjustment, respectively, of the corresponding task offloading strategy.
Correspondingly, after the second particle swarm is obtained, the position value of the particles in the second particle swarm can be matched to a corresponding probability distribution interval, and then the matching relation between the determined combined unloading strategy and the probability distribution interval can be mapped to a genetic algorithm according to the matching relation, so that a corresponding optimized offspring individual is obtained.
In an alternative embodiment, the formula is utilizedCalculating the fitness function value of the individual, wherein,Representing the transmission delay of the ith outstanding task in the combined offloading policy,Representing the computational latency of the ith incomplete task in the joint offload policy,Representing the total number of outstanding tasks on the target failure computing satellite.
In the embodiment of the invention, the optimization target of the combined unloading strategy is to minimize the processing time delay (transmission time delay and calculation time delay) of the unfinished task, so that the fitness function value is inversely related to the task processing time delay of the combined unloading strategy. In view of complex satellite network states, the task processing time may have large difference, and in order to suppress the influence of large values on the fitness function value, the embodiment of the invention adopts logarithmic transformation processing on the processing time delay of all incomplete tasks, and the logarithmic transformation can compress the values, so that the fitness function value is more balanced, and various task allocation schemes are evaluated more fairly.
The transmission delay and the calculation delay of the ith unfinished task selection each task unloading strategy are specifically described below.
1. Offloading to health computing satellites:, , wherein,Representing satellite j and satelliteIs a data transmission rate of the inter-satellite link,Representing satellitesThe calculation frequency assigned to task i.
2. Unloading to a ground terminal:, , wherein,Representing the data transmission rate of the satellite-to-ground link between satellite j and ground terminal i,Representing the calculation frequency assigned by the ground terminal for task i.
3. Unloading to a cloud platform:, , wherein,Representing the data transfer rate of the link between satellite j and the cloud platform,Representing the computing frequency assigned by the cloud platform for task i.
4. Offloading to cloud platform through health computing satellites:, , wherein,Representing satellitesThe data transmission rate of the link with the cloud platform, i.e. the strategy takes into account the two transmission delays, satellite j to satellite respectivelyTransmission delay and satellite of (a)Transmission delay to the cloud platform.
In summary, the embodiment of the invention discloses a task rapid scheduling mechanism for satellite network faults, and provides an effective scheme for task reliable processing under a satellite node fault scene. According to the invention, the state information of each satellite node and the multidimensional information of the network are acquired by using the global observation satellite, and the calculation satellite can actively send the detected fault information, so that the global observation satellite can master the fault condition in time, and the scheduling delay caused by untimely information acquisition is avoided. The mixed mechanism of the global observation and the active report can accelerate the response speed of faults, reduce the influence of the faults on task execution, and provide guarantee for the rapid dispatching of network tasks. High-speed and reliable communication links are established between the global observation satellite and each satellite node, and network fault information and unloading instructions can be transmitted to nodes in the network. Based on abundant data experience, the global satellite can quickly make task scheduling decisions, reasonably allocate network resources and ensure that tasks are quickly and effectively transferred to other nodes for execution.
The embodiment of the invention adopts an improved self-adaptive genetic algorithm, and can effectively carry out joint optimization on the multitask unloading decision. The genetic algorithm can perform global search in the search space, and starts searching from one population, so that the genetic algorithm can explore a plurality of areas at the same time, the opportunity of finding a global optimal solution is increased, the search speed is also increased, and the genetic algorithm is very suitable for a satellite fault network task fast switching scene. The improved adaptive genetic algorithm introduces a multipoint crossover mechanism, which helps to maintain diversity of the population. Through a self-adaptive mutation probability mechanism, higher mutation probability is maintained at the initial stage of algorithm exploration, local optimum is prevented from being achieved too early, the mutation probability is reduced at the later stage of exploration, and the algorithm convergence speed is accelerated, so that the performance of the algorithm is effectively improved. Meanwhile, a particle swarm optimization improvement mechanism is introduced, so that the local exploration capacity is increased.
Fig. 4 is a graph of the performance comparison result of the improved adaptive probability genetic algorithm and the conventional genetic algorithm according to the embodiment of the present invention, and as can be seen from fig. 4, the algorithm adopted in the present invention has more stable convergence and higher adaptability than the conventional genetic algorithm, so that the present invention has better performance.
Example two
The embodiment of the invention also provides a task quick scheduling device under the satellite network fault, which is applied to a global observation satellite in a satellite computing network and is mainly used for executing the task quick scheduling method under the satellite network fault provided by the first embodiment.
Fig. 5 is a functional block diagram of a task fast scheduling device under satellite network failure according to an embodiment of the present invention, as shown in fig. 5, the device mainly includes a first obtaining module 10, a second obtaining module 20, an initializing module 30, and an updating and determining module 40, where:
The first obtaining module 10 is configured to obtain global information in a satellite computing network, where the satellite computing network includes a global observation satellite, a cloud platform, a plurality of computing satellites, and a plurality of ground terminals, and the global information includes state information of the cloud platform, state information of the computing satellites, state information of the ground terminals, inter-satellite link information, and inter-satellite link information.
And a second obtaining module 20, configured to obtain task information of all outstanding tasks on the target failure computing satellite when it is determined that the failure computing satellite exists in the satellite computing network, where the target failure computing satellite represents any one of all the failure computing satellites.
The initialization module 30 is configured to initialize a population by using a genetic algorithm based on task information of all incomplete tasks to obtain an initial population, where each individual in the initial population represents a combined unloading policy, and the combined unloading policy represents a set of task unloading policies of all incomplete tasks.
The updating and determining module 40 is configured to iteratively update the initial population based on global information, task information, genetic algorithm and particle swarm algorithm until a preset iteration number is reached, and calculate a task scheduling policy of the satellite by using a joint unloading policy corresponding to a globally optimal individual as a target fault, where in each generation of population updating process, the initial child population generated based on the genetic algorithm is optimized by using the particle swarm algorithm, so as to construct a next generation population by using the optimized child population and individuals with fitness function values in the initial child population ranked in descending order and before, and the fitness function values of the individuals are inversely related to task processing delays of the joint unloading policy corresponding to the individual.
The embodiment of the invention provides a task rapid scheduling device under satellite network faults, which provides a satellite computing network, and comprises a global observation satellite, a cloud platform, a plurality of computing satellites and a plurality of ground terminals, wherein the device is applied to the global observation satellite in the satellite computing network, the global observation satellite acquires global information in the satellite computing network, when determining that the fault computing satellite exists in the network, task information of all incomplete tasks on a target fault computing satellite is acquired, an initial population is generated by utilizing a genetic algorithm, each individual in the population represents a combined unloading strategy, and then the initial population is iteratively updated according to the global information, the task information, the genetic algorithm and a particle swarm algorithm until the preset iteration times are reached, and the combined unloading strategy corresponding to the global optimal individual is used as the task scheduling strategy of the target fault computing satellite. The execution main body of the device is a global observation satellite, and the task scheduling problem is solved by applying a genetic algorithm and a particle swarm algorithm. And the combination application of the genetic algorithm and the particle swarm algorithm can effectively improve the convergence stability of the algorithm.
Optionally, the device further comprises a generation module for generating an initial offspring population based on a genetic algorithm, the generation module comprising:
and the selection unit is used for selecting two parent individuals from the current population by using a roulette algorithm.
And the crossing unit is used for performing crossing operation on the two parent individuals to obtain two crossed parent individuals.
And the mutation unit is used for respectively mutating the two crossed parent individuals according to the adaptive mutation probability to obtain two initial child individuals corresponding to the two parent individuals.
The repeated calling unit is used for repeatedly calling the selection unit, the crossing unit and the mutation unit until the number of the initial offspring individuals is greater than or equal to the number of the individuals in the initial population, and the set of all the initial offspring individuals is used as the initial offspring population.
Optionally, the cross unit is specifically configured to:
The start and end crossover points of the gene segments to be crossed are randomly generated.
And exchanging the gene segments of the two parent individuals between the starting cross point and the ending cross point to obtain the two crossed parent individuals, wherein the gene segments represent a plurality of continuous gene points, and each gene point represents a task unloading strategy of an unfinished task in the combined unloading strategy.
Optionally, the mutation unit is specifically configured to:
By means of arithmeticCalculating the mutation probability of the parent individuals in the current population to obtain the self-adaptive mutation probability,Indicating a preset initial probability of variation of the sample,The predetermined final probability of variation is indicated,Representing the current number of iterations of the population,Representing a preset number of iterations.
And randomly generating a random number between 0 and 1 for a target parent individual, wherein the target parent individual represents any one of the two crossed parent individuals.
And if the random number is smaller than the self-adaptive mutation probability, mutating at least one gene point of the target parent individual to obtain an initial child individual corresponding to the target parent individual.
And if the random number is greater than or equal to the adaptive mutation probability, taking the target parent individual as the corresponding initial child individual.
Optionally, the device further comprises an optimization module, which is used for optimizing the initial offspring population generated based on the genetic algorithm by utilizing the particle swarm algorithm, and comprises:
the first mapping unit is used for mapping each initial child individual in the initial child population to particles in the particle swarm algorithm to obtain a first particle swarm.
And the updating unit is used for updating the positions and the speeds of all particles in the first particle swarm to obtain a second particle swarm.
And the second mapping unit is used for mapping each particle in the second particle swarm to an individual in the genetic algorithm to obtain an optimized offspring population.
Optionally, the task offloading policy of each incomplete task includes one of offloading to a cloud platform, offloading to a healthy computing satellite, offloading to a ground terminal, offloading to the cloud platform through the healthy computing satellite, and a first mapping unit, specifically configured to:
the total number of outstanding tasks on the satellite and the selectable number of task offloading policies are calculated based on the target failure, and a corresponding probability distribution interval is matched for each joint offloading policy.
And determining a probability distribution interval corresponding to a combined unloading strategy corresponding to the target initial child individual to obtain a target probability distribution interval, wherein the target initial child individual represents any one of the initial child populations.
Taking the interval median value of the target probability distribution interval as the position of the particle corresponding to the target initial offspring individual, and obtaining the speed of the particle by a random initialization mode.
Alternatively, the formula is utilizedCalculating the fitness function value of the individual, wherein,Representing the transmission delay of the ith outstanding task in the combined offloading policy,Representing the computational latency of the ith incomplete task in the joint offload policy,Representing the total number of outstanding tasks on the target failure computing satellite.
Example III
Referring to fig. 6, an embodiment of the present invention provides an electronic device comprising a processor 60, a memory 61, a bus 62 and a communication interface 63, the processor 60, the communication interface 63 and the memory 61 being connected by the bus 62, the processor 60 being arranged to execute executable modules, such as computer programs, stored in the memory 61.
The memory 61 may include a high-speed random access memory (RAM, random Access Memory), and may further include a non-volatile memory (non-volatile memory), such as at least one magnetic disk memory. The communication connection between the system network element and at least one other network element is achieved via at least one communication interface 63 (which may be wired or wireless), and may use the internet, a wide area network, a local network, a metropolitan area network, etc.
Bus 62 may be an ISA bus, a PCI bus, an EISA bus, or the like. The buses may be classified as address buses, data buses, control buses, etc. For ease of illustration, only one bi-directional arrow is shown in FIG. 6, but not only one bus or type of bus.
The memory 61 is configured to store a program, and the processor 60 executes the program after receiving an execution instruction, and the method executed by the apparatus for defining a process disclosed in any of the foregoing embodiments of the present invention may be applied to the processor 60 or implemented by the processor 60.
The processor 60 may be an integrated circuit chip having signal processing capabilities. In implementation, the steps of the above method may be performed by integrated logic circuitry in hardware or instructions in software in the processor 60. The processor 60 may be a general-purpose processor including a central Processing unit (Central Processing Unit, CPU), a network processor (Network Processor, NP), a digital signal processor (DIGITAL SIGNAL Processing, DSP), an Application Specific Integrated Circuit (ASIC), a Field-Programmable GATE ARRAY (FPGA), a discrete gate or transistor logic device, or a discrete hardware component. The disclosed methods, steps, and logic blocks in the embodiments of the present invention may be implemented or performed. A general purpose processor may be a microprocessor or the processor may be any conventional processor or the like. The steps of the method disclosed in connection with the embodiments of the present invention may be embodied directly in the execution of a hardware decoding processor, or in the execution of a combination of hardware and software modules in a decoding processor. The software modules may be located in a random access memory, flash memory, read only memory, programmable read only memory, or electrically erasable programmable memory, registers, etc. as well known in the art. The storage medium is located in a memory 61 and the processor 60 reads the information in the memory 61 and in combination with its hardware performs the steps of the method described above.
The computer program product of the method and apparatus for task fast scheduling under satellite network failure provided in the embodiments of the present invention includes a computer readable storage medium storing non-volatile program code executable by a processor, where the program code includes instructions for executing the method described in the foregoing method embodiments, and specific implementation may refer to the method embodiments and will not be repeated herein.
In addition, each functional unit in the embodiments of the present invention may be integrated in one processing unit, or each unit may exist alone physically, or two or more units may be integrated in one unit.
The functions, if implemented in the form of software functional units and sold or used as a stand-alone product, may be stored in a non-volatile computer readable storage medium executable by a processor. Based on this understanding, the technical solution of the present invention may be embodied essentially or in a part contributing to the prior art or in a part of the technical solution, in the form of a software product stored in a storage medium, comprising several instructions for causing a computer device (which may be a personal computer, a server, a network device, etc.) to perform all or part of the steps of the method according to the embodiments of the present invention. The storage medium includes a U disk, a removable hard disk, a Read-Only Memory (ROM), a random access Memory (RAM, random Access Memory), a magnetic disk, an optical disk, or other various media capable of storing program codes.
It should be noted that like reference numerals and letters refer to like items in the following figures, and thus once an item is defined in one figure, no further definition or explanation thereof is necessary in the following figures.
In the description of the present invention, it should be noted that, directions or positional relationships indicated by terms such as "center", "upper", "lower", "left", "right", "vertical", "horizontal", "inner", "outer", etc., are directions or positional relationships based on those shown in the drawings, or are directions or positional relationships conventionally put in use of the inventive product, are merely for convenience of describing the present invention and simplifying the description, and are not indicative or implying that the apparatus or element to be referred to must have a specific direction, be constructed and operated in a specific direction, and thus should not be construed as limiting the present invention. Furthermore, the terms "first," "second," "third," and the like are used merely to distinguish between descriptions and should not be construed as indicating or implying relative importance.
Furthermore, the terms "horizontal," "vertical," "overhang," and the like do not denote a requirement that the component be absolutely horizontal or overhang, but rather may be slightly inclined. As "horizontal" merely means that its direction is more horizontal than "vertical", and does not mean that the structure must be perfectly horizontal, but may be slightly inclined.
In the description of the present invention, it should also be noted that, unless explicitly specified and limited otherwise, the terms "disposed," "mounted," "connected," and "connected" are to be construed broadly, and may be, for example, fixedly connected, detachably connected, integrally connected, mechanically connected, electrically connected, directly connected, indirectly connected through an intermediary, or in communication between two elements. The specific meaning of the above terms in the present invention will be understood in specific cases by those of ordinary skill in the art.
It should be noted that the above embodiments are merely for illustrating the technical solution of the present invention and not for limiting the same, and although the present invention has been described in detail with reference to the above embodiments, it should be understood by those skilled in the art that the technical solution described in the above embodiments may be modified or some or all of the technical features may be equivalently replaced, and these modifications or substitutions do not make the essence of the corresponding technical solution deviate from the scope of the technical solution of the embodiments of the present invention.

Claims (9)

The updating and determining module is used for carrying out iterative updating on the initial population based on the global information, the task information, the genetic algorithm and the particle swarm algorithm until the preset iterative times are reached, and taking a joint unloading strategy corresponding to a global optimal individual as a task scheduling strategy of the target fault calculation satellite, wherein in each generation of population updating process, the initial offspring population generated based on the genetic algorithm is optimized by utilizing the particle swarm algorithm, so that a next generation population is constructed by utilizing the optimized offspring population and individuals with the front fitness function value in the initial offspring population in descending order;
CN202510272535.2A2025-03-102025-03-10Task rapid scheduling method and device under satellite network faultActiveCN119759455B (en)

Priority Applications (1)

Application NumberPriority DateFiling DateTitle
CN202510272535.2ACN119759455B (en)2025-03-102025-03-10Task rapid scheduling method and device under satellite network fault

Applications Claiming Priority (1)

Application NumberPriority DateFiling DateTitle
CN202510272535.2ACN119759455B (en)2025-03-102025-03-10Task rapid scheduling method and device under satellite network fault

Publications (2)

Publication NumberPublication Date
CN119759455A CN119759455A (en)2025-04-04
CN119759455Btrue CN119759455B (en)2025-06-24

Family

ID=95179431

Family Applications (1)

Application NumberTitlePriority DateFiling Date
CN202510272535.2AActiveCN119759455B (en)2025-03-102025-03-10Task rapid scheduling method and device under satellite network fault

Country Status (1)

CountryLink
CN (1)CN119759455B (en)

Citations (2)

* Cited by examiner, † Cited by third party
Publication numberPriority datePublication dateAssigneeTitle
CN114884958A (en)*2022-07-122022-08-09北京邮电大学Method and device for unloading computing tasks in satellite-ground converged network and electronic equipment
CN116155728A (en)*2023-04-232023-05-23华东交通大学 Calculation offloading and resource optimization method in ultra-dense network

Family Cites Families (3)

* Cited by examiner, † Cited by third party
Publication numberPriority datePublication dateAssigneeTitle
CN116361009B (en)*2023-05-192023-11-10南京邮电大学MEC computing unloading, resource allocation and cache joint optimization method
CN119484644A (en)*2024-10-172025-02-18内蒙古大学 A resource scheduling optimization method for satellite distance education networks
CN119584159A (en)*2024-11-042025-03-07北京邮电大学 A cache-assisted satellite task offloading and resource allocation method

Patent Citations (2)

* Cited by examiner, † Cited by third party
Publication numberPriority datePublication dateAssigneeTitle
CN114884958A (en)*2022-07-122022-08-09北京邮电大学Method and device for unloading computing tasks in satellite-ground converged network and electronic equipment
CN116155728A (en)*2023-04-232023-05-23华东交通大学 Calculation offloading and resource optimization method in ultra-dense network

Also Published As

Publication numberPublication date
CN119759455A (en)2025-04-04

Similar Documents

PublicationPublication DateTitle
CN112398899B (en)Software micro-service combination optimization method for edge cloud system
CN109298930B (en) A cloud workflow scheduling method and device based on multi-objective optimization
CN118101500B (en)Service deployment method and system under edge environment based on improved genetic algorithm
CN113672295A (en)Collaborative computing unloading method based on genetic algorithm in mobile cloud environment
CN112817605A (en)Software-defined satellite network controller deployment method, device and related equipment
CN112016691A (en)Construction method and device of quantum line
CN118337640A (en) A microservice deployment method for cloud and multi-edge network node collaboration
CN113835899A (en)Data fusion method and device for distributed graph learning
CN115983392A (en)Method, device, medium and electronic device for determining quantum program mapping relation
CN116339932A (en) Resource scheduling method, device and server
CN113342504A (en)Intelligent manufacturing edge calculation task scheduling method and system based on cache
CN118095070A (en)Multi-scene multi-objective problem optimization method and device, electronic equipment and storage medium
CN119759455B (en)Task rapid scheduling method and device under satellite network fault
Kusetogullari et al.A reduced uncertainty-based hybrid evolutionary algorithm for solving dynamic shortest-path routing problem
CN115879562A (en)Quantum program initial mapping determination method and device and quantum computer
CN114297934A (en)Model parameter parallel simulation optimization method and device based on proxy model
Lee et al.A genetic algorithm based robust learning credit assignment cerebellar model articulation controller
CN111178529B (en)Data processing method and device, electronic equipment and readable storage medium
Wang et al.Multi-Agent Systems for Collaborative Inference Based on Deep Policy Q-Inference Network
Curry et al.A computational intelligence approach to system-of-systems architecting incorporating multi-objective optimization
CN111412795A (en) Test point setting scheme generation method and device
WO2017016417A1 (en)System control method and device, controller and control system
CN116614377A (en) Method and device for dynamic configuration of UAV swarm service function chain
Santiyuda et al.Adaptive Lévy Flower Pollination Algorithm for Multiple Objective Optimization.
Vivek et al.Parallel bi-objective evolutionary algorithms for scalable feature subset selection via migration strategy under Spark

Legal Events

DateCodeTitleDescription
PB01Publication
PB01Publication
SE01Entry into force of request for substantive examination
SE01Entry into force of request for substantive examination
GR01Patent grant
GR01Patent grant

[8]ページ先頭

©2009-2025 Movatter.jp