CN108268710A - A kind of IMA system dynamic restructuring policy optimization methods based on genetic algorithm - Google Patents
A kind of IMA system dynamic restructuring policy optimization methods based on genetic algorithm Download PDFInfo
- Publication number
- CN108268710A CN108268710A CN201810003355.4A CN201810003355A CN108268710A CN 108268710 A CN108268710 A CN 108268710A CN 201810003355 A CN201810003355 A CN 201810003355A CN 108268710 A CN108268710 A CN 108268710A
- Authority
- CN
- China
- Prior art keywords
- ima
- time
- load
- module
- task
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/10—Geometric CAD
- G06F30/15—Vehicle, aircraft or watercraft design
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/12—Computing arrangements based on biological models using genetic models
- G06N3/126—Evolutionary algorithms, e.g. genetic algorithms or genetic programming
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Geometry (AREA)
- General Physics & Mathematics (AREA)
- Biophysics (AREA)
- Evolutionary Computation (AREA)
- Life Sciences & Earth Sciences (AREA)
- General Engineering & Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Computer Hardware Design (AREA)
- Molecular Biology (AREA)
- Automation & Control Theory (AREA)
- Data Mining & Analysis (AREA)
- Computing Systems (AREA)
- Computational Linguistics (AREA)
- Biomedical Technology (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Artificial Intelligence (AREA)
- General Health & Medical Sciences (AREA)
- Aviation & Aerospace Engineering (AREA)
- Genetics & Genomics (AREA)
- Computational Mathematics (AREA)
- Mathematical Analysis (AREA)
- Mathematical Optimization (AREA)
- Pure & Applied Mathematics (AREA)
- Physiology (AREA)
- Stored Programmes (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a kind of IMA system dynamic restructuring policy optimization methods based on genetic algorithm, belong to Integrated Electronic System design field.Allocation plan State-explosion problem when the present invention solves IMA systems progress dynamic restructuring using Genetic Algorithm Technology, by adding in IMA system reliabilities, real-time and static load balancing analysis method, a kind of effective IMA systems dynamic restructuring Policy model has been obtained.By the way that configuration codes are reconstructed to IMA system hardware and softwares and time configuration, IMA system deployment scenario AADL models can be converted into, reliability, real-time constraint and the multi-dimensional load balancing that IMA systems are carried out by AADL models are analyzed, and can obtain the lethal gene and fitness function in Analysis of Genetic Algorithms optimization.By operatings of genetic algorithm, preferably IMA systems dynamic restructuring allocation plan can be effectively sought.
Description
Technical field
The invention belongs to Integrated Electronic System design field, the synthesization aviation being related in a kind of avionics system design
Electronic system configurations method, and in particular to one kind passes through genetic algorithm in synthetic aviation electronic system (IMA) dynamic restructuring
The method optimized to its software and hardware configuration strategy.
Background technology
The functions such as avionics system is related to communicating, navigates, identifies, detecting, flight management, display control are pilot
Comprehensive and accurate battlefield situation information is provided, potential threat in battlefield surroundings is identified with enhancing pilot, monitor, is analyzed
It is the nervous centralis of entire opportunity of combat with the ability of reaction.Avionic development experience discrete, association type, composite type with
And advanced synthesis type several stages, it is intended to enhance avionics system function, performance, reduce the same of aircraft life cycle cost
When, solve the upgrading Transplanting Problem of system hardware and software.
Since modern aircraft structure, function are complicated, abort situation, the variation formed, operational mode, running environment and knot
The limitation of the problems such as structure, engineer can not possibly define all configurations to cope with the crucial feelings taken off and be likely to occur afterwards
Condition, thus system needs to be directed in the works or unplanned events generate every configuration, i.e., the system reconfiguration that we often say in real time.System
The dynamic restructuring of system is the main means for improving system reliability, in Embedded System Design, is failed to system dynamic restructuring
And its mistake carries out the thing that modeling is part difficulty.
Avionics system of new generation needs have fault-tolerant, recombination function, i.e., when hardware fault occurs, needs to system
It is reconstructed, i.e., repartitions function to the module of safety.Due to the versatility of IMA system modules, the mode of system configuration
Be it is varied optional, and mission requirements variation (such as cruise, search, the switching of attack isotype) or failure not really
It is qualitative (module failure, subregion failure, application software failure when) system is caused to reconstruct as needed when system configuration mode
It is that various, different configuration variation forms different allocation plan and configuration path, the synthesis of a variety of different allocation plans
The height of performance is also different.The existing analysis method to IMA reconstruct is only limitted to be based on existing IMA system deployment scenarios
The calculating analysis of model, the quality of allocation plan is evaluated by the comparison of characteristics more between different allocation plans, and lacks needle
To IMA restructuring procedures, the strategy-generating method of reliability, real-time and load balance characteristic synthesis based on model.For
Dynamic restructuring strategy design problem under multi job mode and multi-constraint condition, there is an urgent need to study a kind of IMA systems dynamic weight
The comprehensive designing method of structure strategy builds a kind of design supporting method, and the synthesization design for IMA systems lays the foundation, and also will
The further development of support next generation's IMA dynamic restructuring technologies.
IMA system configurations are related to application software and are deployed to after IMA platforms, and whether key characteristic meets system design
Requirement, domestic and international related research institutes have carried out special item to each individual characteristic.It is special currently based on the IMA systems of model
Property analysis had many perfect evaluation methods, the reliability consideration for IMA is concentrated mainly on Solution Architecture
In reliability design, some are proposed based on ARINC653 and ASSAC Specification Designs system to improve the method for reliability, for
The requirement of real-time of system proposes the whole real time analysis method of the system dynamic restructuring configuration based on model.Herein it
Outside, also extremely lack for the load balance analysis method of system reconfiguration configuration, it can not be in indexs such as system reliability real-times
The quality of more different allocation plans in the case of satisfaction.Meanwhile these methods are only applicable to evaluate existing IMA system configurations
Scheme.Although it by exhaustive all possible allocation plan and can be evaluated to screen optimal configuration side in theory
Case, but in practical IMA system dynamic restructuring analysis of strategies, since possible allocation plan existence Space Explosion is asked
Topic is difficult to realize by the configuration of exhaustive optimization dynamic restructuring.It, can not for the software and hardware binding relationship of IMA system complex
Optimal IMA system dynamic restructuring configuration strategies are found out by exhaustive possible allocation plan, need one kind can be in high efficiency
Assessment screening carries out IMA system dynamic restructuring strategies to obtain optimal IMA systems by comprehensive multifrequency nature under searching algorithm
The optimization method of system allocation plan.
Invention content
Allocation plan State space explosion when the present invention solves IMA systems progress dynamic restructuring using Genetic Algorithm Technology
Problem by adding in IMA system reliabilities, real-time and static load balancing analysis method, has obtained a kind of effective IMA systems
System dynamic restructuring Policy model.By the way that configuration codes are reconstructed to IMA system hardware and softwares and time configuration, can be converted into
IMA system deployment scenario AADL models, reliability, real-time constraint and the multidimensional that IMA systems are carried out by AADL models load
Equilibrium analysis can obtain lethal gene and fitness function in Analysis of Genetic Algorithms optimization.It, can by operatings of genetic algorithm
Effectively to seek preferably IMA systems dynamic restructuring allocation plan.
Specifically, a kind of IMA system dynamic restructuring policy optimization methods based on genetic algorithm of the invention, are divided into following
Step：
Step 1: establish the IMA system dynamic restructuring analysis of strategies models based on AADL；
IMA system structures and dynamic restructuring process are modeled using AADL language, establish IMA system dynamic restructuring plans
Slightly analysis model, and add in the optimization of IMA system strategies in IMA system dynamic restructuring analysis of strategies models and analyze required system
System reliability, real-time and load balance analysis attribute.It, mainly can including the basic of each software and hardware of system for fail-safe analysis
By degree and task transition probability.For real time analysis, main response and run time including system software, task transfer is generally
Rate and origin identification etc..It is analyzed for load balance, needs to provide the time scheduling mechanism of IMA modules, specify each IMA modules and permit
Perhaps main time frame length and timeslice unit, additionally need the top layer of communication interface and channel between clear and definite IMA system partitionings
Configuration, the i.e. configuration under different software and hardware matching relationships.
Step 2: setting IMA systematic genetic Algorithm Analysis algorithms；
IMA systems are analyzed using genetic algorithm, the matching of the software and hardware of IMA systems and time configuration are carried out first
Coding.It is determined in IMA system hardware and softwares parameters and top-level task chain, and sets in each subregion of IMA systems and only work
In the case of one software, the parameter of IMA system dynamic restructuring structure problems can be represented with time parameter and spatial parameter, institute
The time parameter stated includes the timeslice quantity and sequence that software occupies, and the spatial parameter refers to the hardware position where software
It puts.Binary coding may be used in the coding, software number, hardware number, timeslice quantity comprising system configuration and suitable
Sequence information.
Using the different characteristics of IMA systems as individual Rule of judgment come when generating and screening population, for reliability and
Real time analysis employs existing Stochastic Petri Net Model and analysis method is analyzed.In load balance analysis, for
The task load and traffic load of system, it is proposed that multiple evaluation indexes use run time on time and Spatial Dimension respectively
Than the task load that system is described with EMS memory occupation ratio；For traffic load, it is proposed that bus bandwidth occupancy describes system
Traffic load on Spatial Dimension.In summary specificity analysis obtains the basis for estimation of IMA system dynamic restructuring policy optimizations.
In the chess game optimization that IMA dynamic restructuring schemes are carried out using genetic algorithm, initial population has selected matching at random
Put scheme it is preferred on the basis of the iteration method that generates preferably initial population.In operatings of genetic algorithm select roulette wheel selection,
Single-point interior extrapolation method or binary system mutation algorithm.When terminating search, determine the algebraically upper limit of search as end condition.
Step 3: the reliability of computer system configurations is used as individual screening conditions with real-time；
By system hardware and software matching and time configuration codes, the AADL allocation models of IMA systems can be obtained, while
In the more specificity analysis basic models of IMA systems AADL, system software and the corresponding reliability of hardware and real-time have been contained
Attribute can calculate IMA systems by AADL model conversations for Petri net model and on the basis of Petri net model and work as
Reliability and real-time index under preceding configuration, and be compared with design permissible value, it is lethal to be used as in genetic algorithm
Gene judges whether individual survives.
It analyzes to obtain individual adaptation degree Step 4: carrying out load balance to allocation plan；
In the configuration of IMA system hardware and softwares and during dynamic restructuring, allocation plan is firstly the need of the performance requirement for meeting system, such as
Reliability and real-time demand.On the basis of these necessary conditions are met, we remain desirable to IMA system deployment scenarios can
Better adapt to current mission requirements so that the working efficiency of IMA systems maximizes, and the communication for reducing IMA intermodules needs
It asks, reduces the possibility to break down.By using for reference the analysis method of load balance problem in distributed system, for IMA systems
Carry out static load balancing analysis, by run time than with EMS memory occupation than describe IMA systems task processing capacity, pass through
The bus bandwidth that IMA intermodule communications occupy describes its communication capacity, using the evaluation index as IMA system deployment scenarios.It is comprehensive
Close fitness function of the three of the above evaluation index as individual.
Step 5: the optimal genetic decoding that Genetic algorithm searching obtains is obtained into IMA system dynamic restructuring allocation plans；
Genetic algorithm can obtain the optimal genes of individuals of algorithm, wherein wrapping after search convergence or completing regulation algebraically
The software and hardware matching of IMA system dynamic restructuring allocation plans and time scheduling information are included, corresponding software and hardware is obtained by decoding
The time scheduling information in matrix and each module is matched, these information are corresponded on the AADL models of system architecture, can be obtained
To corresponding IMA systems dynamic restructuring allocation plan.
The advantage of the invention is that：
1st, it employs genetic algorithm and dynamic restructuring policy optimization is carried out to IMA systems, solve IMA system dynamic restructurings and match
Put solution space explosion issues.
2nd, it when the configuration of IMA systems dynamic restructuring is described in genetic algorithm, specifically employs to system hardware and software
With the mode encoded with time configuration, coding information is added in IMA system AADL models in decoding and goes to obtain
Complete allocation plan AADL models.The reliability of IMA system configurations and real-time are added to hereditary behaviour as lethal gene
In work, the reliability of configuration and real-time sexual satisfaction IMA system requirements ensure that.
3rd, in the step 4, to judge allocation plan fitness used by system load Equilibrium Analysis Method, it is comprehensive
The task load and traffic load of IMA system deployment scenarios are closed.Wherein task load includes describing by run time ratio
Time dimension task load and with EMS memory occupation than the task load on the Spatial Dimension of description.In the description of communication capacity
On, since the communication delay of IMA systems is much smaller than the task scheduling interval that temporally piece rotates, only includes and led to IMA intermodules
The traffic load on the Spatial Dimension of bus bandwidth description that letter occupies.
Description of the drawings
Fig. 1 is the IMA system dynamic restructuring policy optimization method flow charts provided by the invention based on genetic algorithm；
Fig. 2 is IMA system partitioning calling process schematic diagrames；
Fig. 3 is IMA system disparate modules EMS memory occupation schematic diagrames；
Fig. 4 is communication mode multilayered structure figure between IMA system partitionings；
Fig. 5 is example IMA system basic utility AADL model schematics；
Fig. 6 is example IMA system reliabilities, real time analysis software architecture Petri net model figure；
Fig. 7 is the AADL model schematics that IMA system deployment scenarios decode.
Specific embodiment
The present invention is described in detail with reference to the accompanying drawings and examples.
The existing analysis method to IMA system dynamic restructurings is only limitted to simply with the model generation comprising configuration information be matched
File is put, and system directly can only be judged by the schedulability and resource constraint under the configuration of configuration information simulation analysis difference
Whether dynamic restructuring configuration meets the single characteristic index such as functional requirement and reliability, real-time, can not be according to the entirety of system
Demand automatically generates suitable dynamic restructuring allocation plan.The height that the present invention should reach for IMA systems after dynamic restructuring can
By the demands such as property and performance be complete, in the base for establishing the IMA system dynamic restructuring analysis of strategies models comprising specificity analysis attribute
It on plinth, is analyzed in IMA system reliabilities, real-time and load balance characteristic respectively, establishes multiple IMA systems individual events point
Analyse model.And on this basis, the allocation plan of system is encoded by genetic algorithm, by system reliability, real-time
It is added in the screening search condition of genetic algorithm with load balancing requirements, is configured with obtaining satisfied IMA systems dynamic restructuring
Scheme completes a kind of IMA system dynamic restructuring policy optimization methods based on genetic algorithm.
Flow chart as shown in Figure 1, the IMA system dynamic restructurings policy optimization side provided by the invention based on genetic algorithm
Method includes the following steps：
Step 1: establish the IMA system dynamic restructuring analysis of strategies models based on AADL.
The use of AADL language is modeling language, IMA system architectures are modeled, in architecture Design model,
According to the High-level design document of IMA systems, connection relation models between the software, hardware and software and hardware of system first, including
The module number of system, zone configuration and bus connecting mode in module etc.；Then according to system detailed design document in mould
The functional attributes of each section, such as the configuration of bus bandwidth, port channel, the run time length and deposit that software needs are added in type
The main time frame length of maximum, timeslice granularity and memory space that storage space, system module allow etc..It is set in architecture
Dynamic Reconfiguration Model of Multi is established on the basis of meter model, IMA system reconfiguration trigger conditions is described with AADL mistake attachmentes etc., uses behavior
Attachment and mode describe the different conditions before and after restructuring procedure and reconstruct.It is excellent that IMA system strategies are added in Dynamic Reconfiguration Model of Multi
Change and analyze wanted system reliability, real-time and load balance analysis attribute, obtain the present invention includes specificity analysis category
The IMA system dynamic restructuring analysis of strategies models of property.
For fail-safe analysis, mainly include the basic reliability of each software and hardware of system and task transition probability.For reality
When property is analyzed, main response and run time including system software, task transition probability and origin identification etc..It is flat for load
Weighing apparatus analysis, needs to provide the time scheduling mechanism of IMA modules, specifies main time frame length and timeslice list that each module allows
Position, the top layer of communication interface and channel is configured between additionally needing clear and definite IMA system partitionings, i.e., in different software and hardware matching relationships
Under configuration.Need the specificity analysis attribute added in such as table 1, table 2.
1 fail-safe analysis of table inputs attribute
2 real time analysis of table inputs attribute
Step 2: IMA system dynamic restructuring configuration codes, set IMA systematic genetic Algorithm Analysis algorithms；
The task that IMA systems are run in practical aircraft uses is more, distributes task to the IMA moulds in IMA systems
Two stage arrangement mode on block and for the corresponding timeslice of IMA module assignments is again very flexible and changeable, thus in IMA system configurations and
During reconstruct, there are problems that allocation plan State space explosion.When AADL language is used to model IMA systems, need
The partitioned organization of IMA inside modules is modeled using ARINC653 attachmentes, by AADL structural models component processor,
Process and thread correspond to the module in ARINC653 attachmentes, process in subregion and subregion respectively.It is right in ARINC653 attachmentes
255 partitioned process should can be bound in practical general purpose I MA system standards, the subsystem of a corresponding IMA module.It will appoint
Business is assigned to after subregion, it is also necessary to the run time of configuration task, when being that task divides in uncertain main time frame
Between piece, this further increases the state space for leading to allocation plan.
An if IMA system configuration spatial model：
F=(d, s, n, T, t0)
Wherein d is the module number of IMA systems, and s is the amount of software of IMA systems；N is that the module of IMA systems can accommodate
Maximum number of partitions；The maximum main time frame period that T allows for IMA systems, t0Module for IMA systems divides timeslice
Granularity, T is t0Multiple.
For a medium sized IMA system, it is assumed that：
F=(d, s, n, T, t0)=(4,20,225,120,10)
Since each module is equivalent in IMA systems, the dynamic of IMA systems can be calculated by the second class Stirling number method
State reconstructs allocation plan number：
It can be seen that before the time dimension configuration of the IMA systems second level is carried out, the possible reconstruct allocation plan of system
Number has had reached 10,000,000,000 ranks, if obtaining optimal reconstruct allocation plan using simply arrangement and screening technique, due to
The analyses such as the reliability real-time of system needs could carry out after system deployment scenario determines, can not generate whole reconstruct and match
The scheme of putting will cause follow-up work that can not carry out.
By calculating it is found that the optimal reconstruct allocation plan that the method enumerated and screened by violence obtains IMA systems is
It is unpractical.Accordingly, it is considered to the more excellent solution of reconfiguration scheme is found using suitable heuritic approach.Genetic algorithm is borrowed as one kind
Reflect the randomization searching method that the evolution laws of living nature develop, and directly structure objects can be operated, be not present
Derivation and the restriction of function continuity, and with inherent Implicit Parallelism and better global optimizing ability.Genetic algorithm is adopted
With the optimization method of randomization, the search space of optimization can be obtained and instructed automatically, the direction of search is adaptively adjusted, does not need to
Determining rule.And when carrying out optimal reconfiguration scheme to IMA systems and searching for, the problem of encountering in addition to search space it is excessive with
Outside, combined reliability also in need, real-time and load balance characteristic carry out scheme evaluation, and judgment criteria is sufficiently complex.For
Overcome the problems, such as these, selection carries out the analysis optimization of IMA system reconfiguration schemes using genetic algorithm.
IMA systems are analyzed using genetic algorithm, need to design genetic algorithm frame first.The search of genetic algorithm
Flow is to need to initially set up the initial population of IMA reconstruct allocation plans based on biology, and the individual in initial population includes
With binary-coded IMA systems dynamic restructuring allocation plan information.Then need to assess " nature individual in initial population
Environment " fitness, fitness are determined by the characteristic of needs assessment in reconstruct allocation plan.Initial population is by selecting and breeding
Generate next-generation, formation new population.At this moment, the individual of new population is represented is matched by gene intersection and the obtained new reconstruct that makes a variation
Put scheme.Genetic algorithm can be repeated continuously this process to find most adaptable individual, and setting end condition can have
One is obtained in the time of limit and preferably reconstructs allocation plan.As a result, genetic algorithm rudimentary algorithm it is practical including a) coding, b)
Fitness function, c) initial population design, d) genetic manipulation and end condition will be discussed in detail this some respectively below.
A) it encodes；
IMA system hardware and softwares parameters and top-level task chain determine and set only work in each subregion one it is soft
In the case of part, the parameter of IMA system reconfiguration structure problems can be represented with time parameter and spatial parameter, time ginseng
Number includes the timeslice quantity and sequence of software occupancy, and the spatial parameter refers to the hardware location where software.
Software number, hardware number, timeslice quantity and sequence are discrete variable, are had using binary coding direct
Semanteme.The free time leaf length that module has should equally be on time piece granularity t0, can equally use binary coding
It represents.
If function f (x1) and f (x2,x3,x4), wherein,
x1=1,2 ..., and s }, x2=1,2 ..., T/t0},x3={ 1,2 ..., n }, x4=1,2 ..., d }
Wherein s is the amount of software of IMA systems；The maximum main time frame period that T allows for IMA systems, t0For module
Divide the granularity of timeslice；N is the open ended maximum number of partitions of module；D is the module number of IMA systems.
Length is used to carry out block code for the string of binary characters of L=sl+dl', wherein,
L=l1+l2+l3=[log2(T/t0)]+[log2n]+[log2d]
L'=[log2(T/t0)]
l1,l2,l3The timeslice unit length shared by software, the scheduling cis-position of place module and place module are represented respectively,
The main time frame length of l' representation modules.Each numerical value is started counting up from 0.
All allocation plans form state space：
SL={ a1,a2,…,ak}
Wherein, being encoded to for any one reconstruct allocation plan k：
ak=(ak1,ak2,…,aks,ak'1,…,ak'd)
Encoding Factor (a for being wherein directed to software comprising sk1,ak2,…,aks) and the d encoding Factors for being directed to hardware
(ak'1,…,ak'd), respectively：
aks=(aks1,aks2,…,aksl),ak'd=(ak'd1,ak'd2,…,ak'dl')
Reconstruct allocation plan, that is, individual UVR exposure character string forms be：
sk=ak11ak12…ak1lak21…akslak'11ak'12…ak'1l'ak'21…ak'dl'It (turns left and represents successively from low level from the right side
To a high position).
For example, for F=(d, s, n, T, t0)=(4,20,225,120,10) an IMA system for, it is each to be configured
The code length of scheme be 296 binary codes, wherein l1=l'=4, l2=8, l3=2.
For some allocation plan, for the 1st software, it is encoded to aks1=00110000001000, code length is
14.Since every numerical value is all started counting up from 0 ,+1 is needed when being reduced to the decimal system from binary system.So for the software, wherein
Preceding l1Position is 0011, represents that software occupies the time leaf length of 4 units, later l2Digit 00000010 represents that the software is it
The 4th software in said module, last l3Position represents the software in the 1st module for 00；
For the 1st module, it is encoded to ak'd1=1000, the main time frame length for representing the module is 9, equal to this
The sum of all software holding time leaf length and all free time leaf length in module.
B) fitness function；
The Petri net model of IMA system reconfiguration allocation plans obtained by coding and decoding, and then calculate IMA systems
Reliability and real-time index.Reliability and real-time need to only meet system requirements, not as fitness function evaluation index, only
Consider when building subgroup as lethal gene.
In load balance specificity analysis, the time dimension of partition information computing system that can directly by containing in coding
Task load, and the task load of Spatial Dimension needs to read the Memory Allocation situation of top layer, and then the AADL moulds for passing through reduction
Type is calculated.Due to being free of systematic information interaction approach information in genetic algorithm encoding, system should be in the soft of top layer
In part task chain preset each task between subregion in and by stages information interaction approach.By the information exchange side for reading top layer
Formula can restore to obtain the AADL models for including port channel information, and then the corresponding traffic load balance index of computing system.
After index comprehensive is carried out, load balance index can be as the fitness of the part evaluation individual of fitness function.
When system load balance spends smaller, reconstruct allocation plan is more excellent.
Then fitness function f (x) works as costmax-cost(x)>0, f (x)=costmax-cost(x)；Otherwise, f (x)=
0。
Wherein, costmaxThe maximum load that expression system is likely to be breached is spent, i.e. selection load spends maximum configuration side
The load of system is spent during case；Cost (x) represents that the load of system during option and installment scheme x is spent.
C) initial population is chosen.
After the completion of written in code, actual verification determines the initial population choosing method of genetic algorithm, due to IMA systems
State space is very big and each task characteristic is influenced it is different, predict which kind of distribution method easily reach higher fitness compared with
For difficulty, so when selecting initial population, first the individual of random generation certain amount, then therefrom chooses best individual and adds
Into initial population.This continuous iteration of process, until number of individuals has reached predetermined scale in initial population.
For example, for IMA system F=(d, s, n, T, t0)=(4,20,225,120,10), possible reconstruct configuration side
Case number rank more than 10,000,000,000.When obtaining initial population, random generation 100 every time reconstructs allocation plans, then therefrom
Choose best individual, continuous iteration is until obtaining 100 individuals as initial population.
D) Evolution of Population；
Genetic manipulation (Evolution of Population) includes following three basic genetic operators：Selection intersects and makes a variation.Individual inheritance is calculated
The operation of son is all carried out in random perturbation.Therefore, the rule that individual in population is migrated to optimal solution be it is random,
But this randomization operation is different with traditional stochastic search methods, is a kind of efficiently oriented stochastic search methods.It is using
When genetic algorithm carries out the chess game optimization of IMA reconstruct allocation plans, the present invention has selected relatively common genetic manipulation method.
In Selecting operation, using roulette wheel selection (roulette wheel selection), in the method, respectively
The select probability and its fitness value of individual are proportional.Individual adaptation degree is bigger, and selected probability is higher, otherwise also
So.Reliability and the ungratified individual of real-time are shown as not surviving, mating centering will not be chosen to.It calculates just
In beginning population after the select probability of each individual, for assortative mating individual, need to carry out more wheel selections.Each round generates one
Uniform random number between [0,1], by the random number, alternatively pointer is selected individual to determine.It, can be randomly after individual is selected
Composition mating pair, for crossover operation below.
Gene by above-mentioned mating pair in crossing operation swaps, and generates the new assortment of genes, beneficial it is expected
Gene can be combined.It employs most common single-point and intersects (one-point crossover) method.Concrete operations
It is：A crosspoint is set at random in individual is gone here and there, and when carrying out intersection, two individual part-structures before or after the point carry out
It exchanges, and generates two new individuals.
In mutation operator, the genic value on certain locus gone here and there to the individual in initial population makes variation, to keep away
Exempt from algorithm and enter local convergence.Employ binary system mutation algorithm, mutation probability 0.01.
E) end condition；
Set the algebraically upper limit of iteration as end condition to ensure that system can be completed to search within the suitable time
Journey is about F=(d, s, n, T, t for scale0)=(4,20,225,120,10) IMA systems, set the algebraically of genetic algorithm
For 200 generations.
Step 3: the reliability of computer system configurations is used as individual screening conditions with real-time.
By system hardware and software and time configuration codes, the AADL allocation models of IMA systems can be obtained, while in IMA
In the more specificity analysis basic models of system AADL, system software and the corresponding reliability of hardware and real-time category have been contained
Property, it can be by the way that AADL model conversations is Petri net models and computing system is being currently configured on the basis of Petri net model
Under reliability and real-time index, and with design permissible value be compared, to sentence in genetic algorithm as lethal gene
Whether disconnected individual survives.
It is according to transformation rule, the component in AADL models and connector is right respectively to carrying out IMA systems reliability analysis
The connection relation between component, connector should be corresponded into directed arc, by AADL to the library institute in Petri net model and transition
Model conversation is fail-safe analysis Petri net model, according to weighting SAPetri nets (Software Architecture Petri
Net, software architecture Petri network) migration probability generation test path PW, the migration probability of test path PW can calculate such as
Under：
Wherein, PtiRepresent the migration probability of arc i in software architecture Petri network.
Test path PW is C1→L1→C2→L2→……→Ln→Cn+1, including n+1 component CiCorresponding library institute and n
A connector LiCorresponding transition share 2n arc between library institute and transition, and the Calculation of Reliability of test path PW is as follows：
Wherein, RciIt is to build C in SAPN test paths PWiCorresponding library reliability, RLiIt is the SAPN test roads
Connector L in diameter PWiThe reliability of corresponding transition, RTiIt is directed arc T between library institute and transitioniReliability；
The reliability R of systemSA, i.e., the reliability and migration of the reliability of entire SAPN by whole m test path PW
Probability calculation obtains, wherein every test path consider n+1 library and n transition and its between transition process it is reliable
Property, such as following formula：
Wherein, PPWiRepresent the migration probability along i-th of test path PW, RPWiRepresent that i-th test path PW's is reliable
Property.
As the reliability R of reconstruct allocation planSAWhen being unsatisfactory for system requirements, which corresponds to the individual of coding
It should be excluded in corresponding population.
When carrying out real time analysis, first according to the transformation rule of IMA system AADL models to Petri net model by AADL
System architecture model conversation is into SAPetri pessimistic concurrency controls；In addition to the time attribute for describing tasks carrying and information exchange is converted to
Outside time response in position element and migration element, it is also necessary to the attribute for describing task response-time is converted, according to
Time SAPetri model and the task beginning and end position generation test path provided, can pass through breadth first search
(BFS) test path from initial point to terminating point is obtained in algorithm, and real time analysis, Ke Yiji are carried out to every test path
It calculates：
The time attribute of element on test path, which is added up, can obtain real time analysis result.2n+1 is shared on test path PW
A element O (including the transition corresponding with n connector of the corresponding library institute of n+1 component), whereinRepresent i-th component O's
Time attribute, TPWRepresent the total time that test path PW is spent.For all test paths, IMA reconstruct configuration strategy should expire
FootWhereinThe maximum time for representing the test path j of system permission limits, TPWjIt represents to be currently configured lower test
The practical cost time of path j.When reconstruction configuration method is unsatisfactory for above-mentioned condition, which should be arranged in the population at place
It removes.
It analyzes to obtain individual adaptation degree Step 4: carrying out load balance to reconstruct allocation plan；
In the configuration of IMA system hardware and softwares and dynamic restructuring, reconstruct allocation plan is firstly the need of the performance for meeting IMA systems
It is required that such as reliability and real-time demand.On the basis of these necessary conditions are met, we remain desirable to IMA system configurations
Scheme can better adapt to current mission requirements so that the working efficiency of system maximizes, and reduces the communication of intermodule
Demand reduces the possibility to break down.By using for reference the analysis method of load balance problem in distributed system, for IMA systems
System carry out static load balancing analysis, by run time than with EMS memory occupation than describe IMA systems task processing capacity, lead to
It crosses bus bandwidth that intermodule communication occupies and communication step-length describes its communication capacity, using as IMA system reconfiguration allocation plans
Evaluation index.
(1) run time ratio；
In IMA systems, dispatching method is seized different from the task on distributed system processing unit, task is divided first
Different subregions is fitted on, each subregion distributes different time windows in a main time frame.Subregion is based on timeslice wheel
The scheduling strategy turned can only obtain process resource and called execution in the time slot of distribution.As shown in Fig. 2, in a master
1 can be included in time frame to k subregion and part Free Partition, Free Partition can be located at any two work subregion it
Between.Sequence performs each subregion from left to right sequentially in time, and after current main time frame has been performed, the order period is followed
Inscription of loop.
For a general IMA system, system includes d module.There is k in wherein l-th of modulelA subregion.L
Task load of a module on time dimension can be expressed as：
Wherein, ζ1For the maximum operation load value of l-th of module, the calculation processing ability of the module is represented；ωf1For l
The mission payload of f subregions in a module.
Each subregion task in the same module of IMA systems is that timesharing carries out, so in l-th above-mentioned of module in the time
In task load formula in dimension, task load of l-th of module on time dimension is expressed as in l-th of module all points
Area mission payload and the ratio between with l-th module maximum operation load value.Structure and scheduling characteristic for IMA systems, I
Can to task load of l-th of the module on time dimension represent formula further be simplified.
It is based in Time Slice Circular Scheduling strategy in IMA systems, even if subregion breaks down in implementation procedure is called, such as
Subregion cannot terminate operation in the time slot of distribution, and system can force to terminate the subregion, when time window is moved to next
Between piece.Correspondingly, it is completed before the time trough end of distribution even if subregion is applied, when system will not be by this part free time
Between distribute to other tasks.The time dimension computing resource and the calculating demand of itself that i.e. task occupies are unrelated, and only with dividing
The time slot length matched is related.When the timeslice of distribution is very few so that when task can not be completed, the dispatching requirement of system cannot
Meet, reconstruct allocation plan can be vetoed directly, not need to further load balance analysis.So f points in l-th of module
The mission payload in areaFor：
Wherein, T1Main time frame length for l-th of module；The time being assigned to for f subregions in l-th of module
Leaf length.
So the representation formula of task load of l-th of module on time dimension can be converted into：
NoteThe working time total in a main time frame for l-th of module, then l-th of module when
Between task load in dimension be further simplified as：
In IMA systems, d module is run simultaneously, task load of the IMA systems on time dimension
costtimescaleThe maximum value of each module load is expressed as, i.e.,：
As can be seen from the above equation, for meeting the reconstruct allocation plan of system call demand, system is on time dimension
Task load is unrelated with the practical calculating demand of each subregion task of module, and only related with the distribution of timeslice.When
costtimescaleWhen minimum, system computing capability on time dimension can be obtained and load optimal reconstruct allocation plan.
(2) EMS memory occupation ratio；
IMA systems modules have independent processing unit, each subregion according to Time Slice Circular Scheduling method occupancy at
Element resources are managed, during each subregion task run, other than the time leaf length difference being assigned to, the module memory of occupancy is big
It is small also different.In order to more preferably utilize system resource and improve the robustness of system, need to balance the interior of each module as far as possible
Deposit occupancy ratio.
The least unit of the allocation plan variation of system is a subregion, relative to EMS memory occupation between different subregions task
The difference of amount, the EMS memory occupation amount fluctuation of single partition running can be ignored.It is single since system needs to ensure operation function
The memory source that a subregion occupies does not exceed the module memory upper limit, can pass through the average value table of task run EMS memory occupation amount
Levy the Spatial Dimension processing capacity load of subregion.
As shown in figure 3, the main time frame length of module 1 is T1, EMS memory occupation amount is not in each partition running
Become, and the EMS memory occupation amount between subregion is different, when running to free time piece, the EMS memory occupation amount of module is 0.Module
2 main time frame length is T2, the main time frame length of module r is Tr, the main time frame length of each module is different, point
The timeslice in area is divided and is also not quite similar, and then subregion committed memory in each module except following identical smallest particles and be outside one's consideration
Changing rule also differs.The time leaf length and main time frame length that the number of partitions of each module of IMA systems, subregion occupy
All be it is different, two identical tasks of committed memory amount when it is operated between it is different when, the influence to system load balance
It is different.It is contemplated that IMA systems are in the load capacity on Spatial Dimension, it is still desirable to memory be accounted for according to time attribute
Dosage is weighted.
Task load spatially can be expressed as：
Wherein, M1Total memory size for l-th of module；mf1It is occupied for f partition runnings in l-th of module average interior
Storage.
In order to meet the hardware substitutability of IMA systems, the memory amount of each module is identical, to operational blocks which partition system l, Ml
=M1=M, the task load cost in system space dimensionspacescaleThe maximum value of each module volume load is expressed as, i.e.,：
System is described in the load capacity on Spatial Dimension using the EMS memory occupation of task run, and consideration task is needed to account for
Size and time with memory, therefore alternatively EMS memory occupation ratio is the load energy that system is described in time and space usage dimension
Power.Work as costspacescaleWhen minimum, the optimal reconstruct allocation plan of system load ability can be obtained.
(3) bus bandwidth occupancy；
In IMA systems, inter-partition communication is cached by the channel between port according to sampling or lining up mode and carried out, such as Fig. 4,
Channel connects up the subregion port of application layer, connects the hardware port of hardware layer downwards, can connect a source partition and one
To multiple purpose subregion ports.On time resource, the communication delay of by stages is much smaller than subregion task according to round-robin
The time deviation brought is run, the inter-partition communication time is that the timeslice configuration of the module where subregion determines, is meeting
It unites in the case of requirement of real-time, additional time resource will not be caused to consume.
On space resources, in same module, module is pre-configured with enough cachings and is used for subregion information interaction, not shadow
Ring the space resources that task run occupies.And the subregion communication of intermodule needs to carry out by system bus, it can be total with system
Tape wide occupancy describes the communication capacity of system.In IMA systems, transmitted different from information in distributed system network
Bandwidth Dynamic Allocation, port that certain two subregion is used when being communicated by system bus and the bus bandwidth that can be occupied are
It is preconfigured, so the communication resource of IMA systems spends costspacecommOnly with the information interaction passage quantity of configuration and logical
The preset bandwidth value in road is related, can be expressed as：
Wherein, mckFor the bus bandwidth that channel k is occupied, McTotal amount of bandwidth for system bus.
Since system bus bandwidth distribution is preset, the service efficiency of channel has no effect on system communication overhead, in IMA
Mainly by reducing the number of channels on system bus in system, especially occupy the larger channel of bus bandwidth is to reduce
The communication overhead of system.
(4) evaluation index integrates；
By more than IMA system loads balance evaluation index it can be seen from IMA systems can according to its system structure parameter into
Row static load balancing is analyzed, and does not need to consider by other models the influence of task run generation.That is the load of IMA systems
Equilibrium analysis can be carried out by its AADL structural model, not need to establish individual load balance analysis model.
While being pre-configured in of IMA systems ensures system function, the coupling of system module and by stages is reduced, because
It is relatively independent that this, which has obtained each evaluation index, when carrying out index comprehensive, it is no longer necessary to which avoidance system operating status causes
Index effectiveness variation.For the load balance characteristic of more flexible description IMA systems, three kinds of indexs are weighted and are asked
With：
Wherein, coefficientIt can be set by experience and test analysis.
When system total load cost is smaller, show that the load balance characteristic of system deployment scenario is more excellent.
The fitness function f (x) of genetic algorithm has：
Work as costmax- cost (x) ＞ 0, f (x)=costmax- cost (x), otherwise, f (x)=0.
Step 5: the optimal base that Genetic algorithm searching obtains is obtained into IMA system dynamic restructurings configuration side because being decoded
Case.
Genetic algorithm can obtain the optimal genes of individuals of algorithm, wherein wrapping after search convergence or completing regulation algebraically
The software and hardware matching of IMA system dynamic restructuring allocation plans and time scheduling information are included, corresponding software and hardware is obtained by decoding
The time scheduling information in matrix and each module is matched, these information are corresponded on the AADL models of system architecture, can be obtained
To corresponding IMA systems dynamic restructuring allocation plan.
The specific reality of IMA system dynamic restructuring policy optimization methods is described below by way of a simplified flight control system example
Apply process.
In flight control system, " state of flight selector " (State_selector) and " flying control console "
(Operating) be flight control system man-machine interface.It has the function of driver to the setting of working state of system and to system
The instruction function of working condition.System increases steady (pcs, lcs), automatic pilot control/mode (ap_ by increasing steady/control
Manage it) is formed with automatic trim (pap, lap)；Wherein increase steady/automatic increase surely has surely respectively with automatic pilot control/mode
Sensor and GPS are determined as information input interface.Simultaneity factor, which also has, controls line subsystem, mainly state of flight selector
(CDS), start logic, redundancy management, operating system and system BIT compositions.The specific implementation step of this method is as follows.
Step 1: establish the IMA system dynamic restructuring analysis of strategies models based on AADL.
AADL model schematics such as Fig. 5 of the system major software program, operating system start logic, BIT, remaining pipe
Reason increases steady/control and increases steady, automatic trim, automatic pilot control/mode, state of flight and is selected as software application, flies control behaviour
Make platform, state of flight selector, stability sensor and GPS as peripheral hardware, software, the connection relation between are as shown in the figure.
Reliability and real time analysis the attribute data such as table 3, table 4 that system is given are added in AADL models.
3 reliability analysis data of table
4 real time analysis data (unit of table：ms)
It is configured the dynamic restructuring policy mandates of the IMA systems, it, should be by the application point of above 10 basic softwares after system reconfiguration
It is fitted on 5 equivalent IMA system modules, the time configuration of module provides in following configuration space model.
System configuration spatial model is：
F=(d, s, n, T, t0)
Wherein d is the module number of system, and s is the amount of software of system；N is the open ended maximum number of partitions of module；T
For the maximum main time frame period that system allows, t0The granularity of timeslice is divided for module.
Have for the IMA systems：
F=(d, s, n, T, t0)=(5,10,10,10,1)
Step 2: setting IMA systematic genetic Algorithm Analysis algorithms.
A) it encodes
If function f (x1),f(x2,x3,x4), wherein,
x1=1,2 ..., and 10 }, x2=1,2 ..., and 10 }, x3=1,2 ..., and 10 }, x4=1,2 ..., 5 }
Length is used to carry out block code for the string of binary characters of L=sl+dl'=130, wherein,
L=l1+l2+l3=[log2(T/t0)]+[log2n]+[log2D]=11
L'=[log2(T/t0)]=4
Establish bit string space：
SL={ a1,a2,…,ak}
Allocation plan k's is encoded to：
ak=(ak1,ak2,…,aks,ak'1,…,ak'd)
The encoding Factor of hardware is wherein directed to comprising the s encoding Factors and d for software：
aks=(aks1,aks2,…,aksl),ak'd=(ak'd1,ak'd2,…,ak'dl')
Allocation plan is the vectorial a of individualk=(ak1,ak2,…,aks,ak'1,…,ak'd), character string forms are
sk=ak11ak12…ak1lak21…akslak'11ak'12…ak'1l'ak'21…ak'dl'It (turns left and represents successively from low level from the right side
To a high position).B) fitness function
Fitness function f (x),
Work as costmax-cost(x)>0, f (x)=costmax-cost(x)；Otherwise, f (x)=0.
Wherein costmaxIt will be provided in step 4 with the computational methods of cost (x).
C) initial population is chosen
For example, for IMA system F=(d, s, n, T, t0)=(5,10,10,10,1), software and hardware matching status space number
Amount is calculated according to the second class Stirling number methodTo each software and hardware
Matching configuration possesses the time slice scheduling allocation plan of different number again.Selection is when obtaining initial population, random generation every time
Then 100 state schedules therefrom choose best individual, continuous iteration is until obtaining 100 individuals as initial population.
D) genetic manipulation
In Selecting operation, using roulette wheel selection, the select probability and its fitness value of each individual are proportional.It is right
It shows as not surviving in reliability and the ungratified individual of real-time, mating centering will not be chosen to.It calculates in group
After the select probability of each individual, for assortative mating individual, need to carry out more wheel selections.Each round generate one [0,1] it
Between uniform random number, by the random number, alternatively pointer determines selected individual.By above-mentioned mating pair in crossing operation
Gene swaps, and generates the new assortment of genes, it is expected that beneficial gene can be combined.Employ most common list
Point cross method.In mutation operator, binary system mutation algorithm, mutation probability 0.01 are employed.
E) end condition
Set the algebraically upper limit of iteration as end condition to ensure that system can be completed to search within the suitable time
Journey, by actual verification, for the IMA systems, genetic algorithm algebraically is set as 50 generations i.e. available preferably search result.
Step 3: the reliability of computer system configurations is used as individual screening conditions with real-time
Reliability and real time analysis that Petri net model carries out IMA system reconfiguration strategies are established, according to AADL system knots
Structure model conversation represents the library institute that component C1 to C10 is converted, L1 to L10 to Petri network figure such as Fig. 6, wherein C1 to C10
The transition that connector L1 to L10 is converted are represented, search obtains test path such as table 5.
5 test path of table
Weight average principle is shifted according to path, obtains the mean transferred probability in each path, and then calculate each test path
Path reliability such as table 6：
6 test path reliability of table
| Number | Transition probability | Path reliability |
| P1 | 0.031250 | 0.790 |
| P2 | 0.078125 | 0.900 |
| P3 | 0.062500 | 0.922 |
| P4 | 0.125000 | 0.922 |
| P5 | 0.078125 | 0.890 |
| P6 | 0.125000 | 0.900 |
| P7 | 0.125000 | 0.922 |
| P8 | 0.125000 | 0.922 |
| P9 | 0.031250 | 0.860 |
Architecture is brought into using above-mentioned result of calculation and calculates total formula, and the architecture reliability of calculating is：
RSA=0.7074085/0.78125=0.9055
The RELIABILITY INDEX R of systemo=0.9000, RSA＞ RoMeet the reliability requirement of system.
According to test path by the run time such as table for performing time and each test path of triggered time calculating of component
7：
7 test path run time of table
For all paths, the run time of IMA reconstruct configuration strategies should be less than longest run time limit.
The individual of requirement is unsatisfactory for for reliability and real-time, should be excluded in progeny population.It can by above-mentioned calculating
To obtain, reliability and real-time all meet in the allocation plan, can continue the fitness analysis of individual.
It analyzes to obtain individual adaptation degree Step 4: carrying out load balance to allocation plan
Extraction system configuration information and the communication information such as the following table 8, table 9 from system reconfiguration strategy AADL models：
(chronomere is configured in 8 system hardware and software of table:s)
9 system communication configuration of table
And the memory amount of each module of system is identical, there is Ml=M=1000, system bus bandwidth total amount Mc=3000.
According to information above, three load balance characteristics of computer system configurations scheme：
(1) run time ratio
Task loads of the system module l on time dimension can be expressed as：
WhereinThe working time total in a main time frame for module.
The task load of each module such as table 10：
10 module design task time load of table
Task load of the system on time dimension is expressed as the maximum value of each module load, has：
costtimescale=max (l)tcost=0.9
(2) EMS memory occupation ratio
The memory amount of each module is identical, Ml=M=1000, the task load such as table 11 of each module：
11 module design task space of table loads
Task load in system space dimension is expressed as the maximum value of each module volume load, i.e.,：
costspacescale=max (l)scost=0.85
(3) bus bandwidth occupancy
The communication resource of IMA systems spends the preset bandwidth value only with the information interaction passage quantity of configuration and channel to have
It closes, the communication in module same in table is cast out, remaining every bandwidth chahnel addition can obtain：
Wherein mckThe bus bandwidth occupied for channel k.
(4) evaluation index integrates
It sums to three kinds of indexs：
Cost=costtimescale+costspacescale+costspacecomm=0.9+0.85+0.8=2.55
When system total load cost is smaller, show that the load balance characteristic of system deployment scenario is more excellent.
The fitness function of genetic algorithm has f (x)=costmax- cost (x)=3-2.55=0.45.
Step 5: the optimal genetic decoding that Genetic algorithm searching obtains is obtained into IMA system dynamic restructuring allocation plans
Genetic algorithm can obtain the optimal genes of individuals of algorithm, wherein wrapping after search convergence or completing regulation algebraically
The software and hardware matching of IMA system dynamic restructuring allocation plans and time scheduling information are included, corresponding software and hardware is obtained by decoding
The time scheduling information in matrix and each hardware module is matched, these information are corresponded on the AADL models of system architecture, it can
To obtain corresponding IMA systems dynamic restructuring allocation plan.
Assuming that Current protocols are the optimal case that genetic algorithm obtains, available information is shown in from scheme individual UVR exposure
12 table 13 of table：
12 software configuration information of table
| Component is numbered | Member position | Time leaf length | Memory size |
| C1 | M3 | 4 | 500 |
| C2 | M1 | 4 | 500 |
| C3 | M2 | 2 | 300 |
| C4 | M2 | 2 | 200 |
| C5 | M3 | 3 | 200 |
| C6 | M4 | 2 | 300 |
| C7 | M4 | 2 | 200 |
| C8 | M2 | 5 | 350 |
| C9 | M5 | 4 | 400 |
| C10 | M5 | 3 | 350 |
13 module configuration information of table
| Module is numbered | Main time frame length |
| M1 | 6 |
| M2 | 10 |
| M3 | 8 |
| M4 | 5 |
| M5 | 10 |
These information are corresponded on the AADL models of system architecture, corresponding IMA systems dynamic restructuring can be obtained and matched
It is as shown in Figure 7 to put AADL models.Such as Fig. 7, component C1 to C10 is located at respectively in the subregion on five module M1 to M5, component it
Between connection relation by system task demand determine and Fig. 5 in it is identical.
Claims (3)
1. a kind of IMA system dynamic restructuring policy optimization methods based on genetic algorithm, it is characterised in that：Include the following steps：
Step 1: establish the IMA system dynamic restructuring analysis of strategies models based on AADL；
IMA system structures and dynamic restructuring process are modeled using AADL language, and added in IMA system dynamic restructurings
System reliability, real-time and load balance analysis attribute, establish the IMA system dynamic restructuring analysis of strategies moulds based on AADL
Type；For fail-safe analysis, basic reliability and task transition probability including each software and hardware of system；For real time analysis,
Response and run time including system software, task transition probability and origin identification；It is analyzed for load balance, needs to provide
The time scheduling mechanism of IMA modules specifies main time frame length and timeslice unit that each IMA modules allow, additionally needs
The top layer of communication interface and channel is configured between specifying IMA system partitionings, i.e. the configuration under different software and hardware matching relationships；
Step 2: setting IMA systematic genetic Algorithm Analysis algorithms；
IMA systems are analyzed using genetic algorithm, the matching of the software and hardware of IMA systems and time configuration are compiled first
Code；The coding using binary coding, believe by software number, hardware number, timeslice quantity comprising system configuration and sequence
Breath；
Using genetic algorithm, using the different characteristics of IMA systems as individual Rule of judgment come when generating and screening population, for
Reliability and real time analysis employ existing Stochastic Petri Net Model and analysis method is analyzed；In load balance point
In analysis, for the task load and traffic load of system, accounted on time and Spatial Dimension with run time ratio and memory respectively
With the task load than describing system；For traffic load, it is proposed that bus bandwidth occupancy is described in system space dimension
Traffic load；In summary specificity analysis obtains the basis for estimation of IMA system dynamic restructuring policy optimizations；
Roulette wheel selection, single-point interior extrapolation method or binary system mutation algorithm are selected in operatings of genetic algorithm；When terminating search, really
Surely the algebraically upper limit searched for is as end condition；
Step 3: the reliability of computer system configurations is used as individual screening conditions with real-time；
By system hardware and software matching and time configuration codes, the AADL allocation models of IMA systems is obtained, while in IMA systems
In the more specificity analysis basic models of AADL, system software and the corresponding reliability of hardware and real-time attribute have been contained, has been led to
Cross AADL model conversations for Petri net model and on the basis of Petri net model calculate IMA systems under current-configuration can
It is compared by property and real-time index, and with design permissible value, lethal gene to be used as to judge individual in genetic algorithm
Whether survive；
It analyzes to obtain individual adaptation degree Step 4: carrying out load balance to allocation plan；
In the configuration of IMA system hardware and softwares and dynamic restructuring, allocation plan is firstly the need of the reliability and real-time for meeting system
Demand；On the basis of these necessary conditions are met, static load balancing analysis is carried out for IMA systems, passes through run time
Than with task processing capacity of the EMS memory occupation than describing IMA systems, the bus bandwidth occupied by IMA intermodule communications describes it
Communication capacity, using the evaluation index as IMA system deployment scenarios；In summary fitness of three kinds of evaluation indexes as individual
Function；
Step 5: the optimal genetic decoding that Genetic algorithm searching obtains is obtained into IMA system dynamic restructuring allocation plans；
Genetic algorithm obtains the optimal genes of individuals of algorithm, including IMA systems after search convergence or completing regulation algebraically
The software and hardware matching of dynamic restructuring allocation plan and time scheduling information, by decoding obtain corresponding software and hardware match matrix and
Time scheduling information in each module, these information are corresponded on the AADL models of system architecture, obtain corresponding IMA systems
Dynamic restructuring allocation plan.
2. a kind of IMA system dynamic restructuring policy optimization methods based on genetic algorithm according to claim 1, feature
It is：Genetic algorithm includes a) coding, b) fitness function, c) initial population design, d) genetic manipulation and end condition；
A) it encodes；
It is determined in IMA system hardware and softwares parameters and top-level task chain and sets the software of only working in each subregion
In the case of, the parameter of IMA system reconfiguration structure problems is represented with time parameter and spatial parameter, and the time parameter includes soft
The timeslice quantity and sequence that part occupies, the spatial parameter refer to the hardware location where software；
Software number, hardware number, timeslice quantity and sequence are discrete variable, have direct language using binary coding
Justice；The free time leaf length that module has should equally be on time piece granularity t0, equally use binary coding representation；
B) fitness function；
The Petri net model of IMA system reconfiguration allocation plans obtained by coding and decoding, and then calculate the reliable of IMA systems
Property and real-time index；Reliability and real-time need to only meet system requirements, not as fitness function evaluation index, only in structure
Consider when building subgroup as lethal gene；
In load balance specificity analysis, directly born by the time dimension task of the partition information computing system contained in coding
It carries, and the task load of Spatial Dimension needs to read the Memory Allocation situation of top layer, and then is carried out by the AADL models of reduction
It calculates；When system load balance spends smaller, reconstruct allocation plan is more excellent；Then fitness function f (x) works as costmax-cost
(x)>0, f (x)=costmax-cost(x)；Otherwise, f (x)=0；
Wherein, costmaxThe maximum load that expression system is likely to be breached is spent, i.e., when selection load spends maximum allocation plan
The load of system is spent；Cost (x) represents that the load of system during option and installment scheme x is spent；
C) initial population is chosen；
When selecting initial population, first the individual of random generation certain amount, then therefrom chooses best individual and is added to initially
In population；This continuous iteration of process, until number of individuals has reached predetermined scale in initial population；
D) Evolution of Population；
In Selecting operation, using roulette wheel selection, in the method, the select probability of each individual and its fitness value into
Ratio；Individual adaptation degree is bigger, and selected probability is higher, and vice versa；It is ungratified for reliability and real-time
Individual shows as not surviving, and will not be chosen to mating centering；It calculates in initial population after the select probability of each individual,
For assortative mating individual, need to carry out more wheel selections, each round generates uniform random number between one [0,1], this is random
Alternatively pointer is selected individual to determine for number；After individual is selected, randomly composition mating pair, for crossover operation below；
In mutation operator, binary system mutation algorithm, mutation probability 0.01 are employed；
E) end condition.
3. a kind of IMA system dynamic restructuring policy optimization methods based on genetic algorithm according to claim 1, feature
It is：Static load balancing analysis is carried out for IMA systems in step 4, is described by run time ratio and EMS memory occupation ratio
The task processing capacity of IMA systems, the bus bandwidth occupied by intermodule communication and communication step-length describe its communication capacity, with
It is specific as follows as the evaluation index of IMA system reconfiguration allocation plans：
(1) run time ratio；
In IMA systems, task is first allocated to different subregions, and each subregion is in a main time frame, and distribution is not
Same time window；Scheduling strategy of the subregion based on round-robin can only obtain process resource simultaneously in the time slot of distribution
It is called to perform；
For a general IMA system, system includes d module, wherein having k in l-th of modulelA subregion, l-th of mould
Task load of the block on time dimension is expressed as：
Wherein, ζ1For the maximum operation load value of l-th of module, the calculation processing ability of the module is represented；ωf1For l-th of mould
The mission payload of f subregions on block；
The mission payload of f subregions in l-th of moduleFor：
Wherein, T1Main time frame length for l-th of module；The time length of a film being assigned to for f subregions in l-th of module
Degree；
So the representation formula of task load of l-th of module on time dimension is converted into：
NoteThe working time total in a main time frame for l-th of module, then l-th of module is in time dimension
On task load be further simplified as：
In IMA systems, d module is run simultaneously, task load cost of the IMA systems on time dimensiontimescaleTable
The maximum value of each module load is shown as, i.e.,：
Found out by above formula, for meeting the reconstruct allocation plan of system call demand, task load of the system on time dimension
It is unrelated with the practical calculating demand of each subregion task of module, and it is only related with the distribution of timeslice, work as costtimescaleWhen minimum,
It obtains system computing capability on time dimension and loads optimal reconstruct allocation plan；
(2) EMS memory occupation ratio；
Task load spatially is expressed as：
Wherein, M1Total memory size for l-th of module；mf1The average memory occupied for f partition runnings in l-th of module
Amount；
In order to meet the hardware substitutability of IMA systems, the memory amount of each module is identical, to operational blocks which partition system l, Ml=M1
=M, the task load cost in system space dimensionspacescaleThe maximum value of each module volume load is expressed as, i.e.,：
Work as costspacescaleWhen minimum, the optimal reconstruct allocation plan of system load ability is obtained；
(3) bus bandwidth occupancy；
The communication resource of IMA systems spends costspacecommOnly with the information interaction passage quantity of configuration and the pre-set bandwidths of channel
It is worth related, is expressed as：
Wherein, mckFor the bus bandwidth that channel k is occupied, McTotal amount of bandwidth for system bus；
(4) evaluation index integrates；
Summation is weighted to three kinds of indexs：
Wherein, coefficientIt is set by experience and test analysis；
When system total load cost is smaller, show that the load balance characteristic of system deployment scenario is more excellent；The adaptation of genetic algorithm
Degree function f (x) has：
Work as costmax- cost (x) ＞ 0, f (x)=costmax- cost (x), otherwise, f (x)=0.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810003355.4A CN108268710B (en) | 2018-01-03 | 2018-01-03 | Genetic algorithm-based IMA system dynamic reconstruction strategy optimization method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810003355.4A CN108268710B (en) | 2018-01-03 | 2018-01-03 | Genetic algorithm-based IMA system dynamic reconstruction strategy optimization method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108268710A true CN108268710A (en) | 2018-07-10 |
| CN108268710B CN108268710B (en) | 2021-07-09 |
Family
ID=62773281
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810003355.4A Active CN108268710B (en) | 2018-01-03 | 2018-01-03 | Genetic algorithm-based IMA system dynamic reconstruction strategy optimization method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108268710B (en) |
Cited By (14)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109814920A (en) * | 2019-01-24 | 2019-05-28 | 华东师范大学 | A kind of software configuration complexity decomposition apparatus |
| CN110263371A (en) * | 2019-05-13 | 2019-09-20 | 北京航空航天大学 | IMA dynamic restructuring process configuration path generating method based on AADL |
| CN110298060A (en) * | 2019-04-30 | 2019-10-01 | 哈尔滨工程大学 | It is a kind of based on gas turbine Stute space model identification method cold between improving expert inquiry method |
| CN110414097A (en) * | 2019-07-12 | 2019-11-05 | 北京旋极信息技术股份有限公司 | IMA system resource configuration verification method and system, computer readable storage medium |
| CN110737441A (en) * | 2019-09-03 | 2020-01-31 | 中国航空无线电电子研究所 | IMA configuration generation method based on micro-service architecture |
| CN111258754A (en) * | 2020-01-09 | 2020-06-09 | 上海依图信息技术有限公司 | Resource processing method, device, medium and system based on time window |
| CN111552585A (en) * | 2020-04-16 | 2020-08-18 | 中国航空无线电电子研究所 | IMA system dynamic reconstruction process configuration path generation method |
| CN112463628A (en) * | 2020-12-11 | 2021-03-09 | 北京航空航天大学 | Self-adaptive evolution method of autonomous unmanned system software based on model base framework |
| CN112488387A (en) * | 2020-11-30 | 2021-03-12 | 重庆邮电大学 | Software product line configuration method based on multi-objective optimization algorithm |
| CN112883526A (en) * | 2021-03-15 | 2021-06-01 | 广西师范大学 | Workload distribution method under task delay and reliability constraints |
| CN112947993A (en) * | 2019-12-31 | 2021-06-11 | 深圳市明源云链互联网科技有限公司 | Method and device for reconstructing system framework, electronic equipment and storage medium |
| CN114330879A (en) * | 2021-12-29 | 2022-04-12 | 蔷薇大树科技有限公司 | Order distribution method and system of multi-dimensional constraint |
| CN116719621A (en) * | 2023-06-01 | 2023-09-08 | 上海聚水潭网络科技有限公司 | Data write-back method, device, equipment and medium for mass tasks |
| CN116880978A (en) * | 2023-07-07 | 2023-10-13 | 郑州轻工业大学 | Real-time system partition scheduling method and scheduler based on model inspection |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102624870A (en) * | 2012-02-01 | 2012-08-01 | 北京航空航天大学 | Intelligent optimization algorithm based cloud manufacturing computing resource reconfigurable collocation method |
| CN104202188A (en) * | 2014-09-01 | 2014-12-10 | 北京航空航天大学 | Method for carrying out AFDX network path optimization by genetic algorithm |
| CN105226702A (en) * | 2015-09-02 | 2016-01-06 | 山东大学 | A kind of alternating current-direct current mixing micro-capacitance sensor dynamic schema and recombination method thereof |
| US20160034809A1 (en) * | 2014-06-10 | 2016-02-04 | Sightline Innovation Inc. | System and method for network based application development and implementation |
| CN105373650A (en) * | 2015-10-15 | 2016-03-02 | 北京航空航天大学 | AADL based IMA dynamic reconfiguration modeling method |
| CN105704255A (en) * | 2016-04-29 | 2016-06-22 | 浙江理工大学 | Server load balancing method based on genetic algorithm |
| CN106920015A (en) * | 2017-04-11 | 2017-07-04 | 东南大学 | Suitable for the most short loop method for dynamically partitioning of power distribution network reconfiguration representation |
-
2018
- 2018-01-03 CN CN201810003355.4A patent/CN108268710B/en active Active
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102624870A (en) * | 2012-02-01 | 2012-08-01 | 北京航空航天大学 | Intelligent optimization algorithm based cloud manufacturing computing resource reconfigurable collocation method |
| US20160034809A1 (en) * | 2014-06-10 | 2016-02-04 | Sightline Innovation Inc. | System and method for network based application development and implementation |
| CN104202188A (en) * | 2014-09-01 | 2014-12-10 | 北京航空航天大学 | Method for carrying out AFDX network path optimization by genetic algorithm |
| CN105226702A (en) * | 2015-09-02 | 2016-01-06 | 山东大学 | A kind of alternating current-direct current mixing micro-capacitance sensor dynamic schema and recombination method thereof |
| CN105373650A (en) * | 2015-10-15 | 2016-03-02 | 北京航空航天大学 | AADL based IMA dynamic reconfiguration modeling method |
| CN105704255A (en) * | 2016-04-29 | 2016-06-22 | 浙江理工大学 | Server load balancing method based on genetic algorithm |
| CN106920015A (en) * | 2017-04-11 | 2017-07-04 | 东南大学 | Suitable for the most short loop method for dynamically partitioning of power distribution network reconfiguration representation |
Non-Patent Citations (3)
| Title |
|---|
| JU, HONGYAN 等: "A Modeling Method of IMA Dynamic Reconfiguration Based on AADL", 《PROCEEDINGS OF THE 2015 FIRST INTERNATIONAL CONFERENCE ON RELIABILITY SYSTEMS ENGINEERING 2015 ICRSE》 * |
| 王世海: "基于AADL体系结构模型的构件***可靠性评估", 《***工程与电子技术》 * |
| 程桢: "模型驱动的IMA***资源配置分析与验证及优化方法研究", 《中国优秀硕士学位论文全文数据库 工程科技Ⅱ辑》 * |
Cited By (23)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109814920B (en) * | 2019-01-24 | 2020-09-11 | 华东师范大学 | Software structure complexity degradation device |
| CN109814920A (en) * | 2019-01-24 | 2019-05-28 | 华东师范大学 | A kind of software configuration complexity decomposition apparatus |
| CN110298060A (en) * | 2019-04-30 | 2019-10-01 | 哈尔滨工程大学 | It is a kind of based on gas turbine Stute space model identification method cold between improving expert inquiry method |
| CN110263371A (en) * | 2019-05-13 | 2019-09-20 | 北京航空航天大学 | IMA dynamic restructuring process configuration path generating method based on AADL |
| CN110263371B (en) * | 2019-05-13 | 2020-10-02 | 北京航空航天大学 | IMA dynamic reconstruction process configuration path generation method based on AADL |
| CN110414097A (en) * | 2019-07-12 | 2019-11-05 | 北京旋极信息技术股份有限公司 | IMA system resource configuration verification method and system, computer readable storage medium |
| CN110737441A (en) * | 2019-09-03 | 2020-01-31 | 中国航空无线电电子研究所 | IMA configuration generation method based on micro-service architecture |
| CN110737441B (en) * | 2019-09-03 | 2023-08-01 | 中国航空无线电电子研究所 | IMA configuration generation method based on micro-service architecture |
| CN112947993A (en) * | 2019-12-31 | 2021-06-11 | 深圳市明源云链互联网科技有限公司 | Method and device for reconstructing system framework, electronic equipment and storage medium |
| CN112947993B (en) * | 2019-12-31 | 2021-12-07 | 深圳市明源云链互联网科技有限公司 | Method and device for reconstructing system framework, electronic equipment and storage medium |
| CN111258754A (en) * | 2020-01-09 | 2020-06-09 | 上海依图信息技术有限公司 | Resource processing method, device, medium and system based on time window |
| CN111552585A (en) * | 2020-04-16 | 2020-08-18 | 中国航空无线电电子研究所 | IMA system dynamic reconstruction process configuration path generation method |
| CN112488387A (en) * | 2020-11-30 | 2021-03-12 | 重庆邮电大学 | Software product line configuration method based on multi-objective optimization algorithm |
| CN112488387B (en) * | 2020-11-30 | 2023-08-29 | 重庆邮电大学 | Software product line configuration method based on multi-objective optimization algorithm |
| CN112463628A (en) * | 2020-12-11 | 2021-03-09 | 北京航空航天大学 | Self-adaptive evolution method of autonomous unmanned system software based on model base framework |
| CN112883526A (en) * | 2021-03-15 | 2021-06-01 | 广西师范大学 | Workload distribution method under task delay and reliability constraints |
| CN112883526B (en) * | 2021-03-15 | 2023-04-07 | 广西师范大学 | Workload distribution method under task delay and reliability constraint |
| CN114330879A (en) * | 2021-12-29 | 2022-04-12 | 蔷薇大树科技有限公司 | Order distribution method and system of multi-dimensional constraint |
| CN114330879B (en) * | 2021-12-29 | 2022-09-16 | 蔷薇大树科技有限公司 | Order distribution method and system of multi-dimensional constraint |
| CN116719621A (en) * | 2023-06-01 | 2023-09-08 | 上海聚水潭网络科技有限公司 | Data write-back method, device, equipment and medium for mass tasks |
| CN116719621B (en) * | 2023-06-01 | 2024-05-03 | 上海聚水潭网络科技有限公司 | Data write-back method, device, equipment and medium for mass tasks |
| CN116880978A (en) * | 2023-07-07 | 2023-10-13 | 郑州轻工业大学 | Real-time system partition scheduling method and scheduler based on model inspection |
| CN116880978B (en) * | 2023-07-07 | 2024-04-09 | 郑州轻工业大学 | Real-time system partition scheduling method and scheduler based on model inspection |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108268710B (en) | 2021-07-09 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108268710A (en) | A kind of IMA system dynamic restructuring policy optimization methods based on genetic algorithm | |
| CN113411369B (en) | Cloud service resource collaborative optimization scheduling method, system, medium and equipment | |
| Shanthikumar et al. | A unifying view of hybrid simulation/analytic models and modeling | |
| Wah | File placement on distributed computer systems | |
| CN110673951B (en) | Mimicry scheduling method, system and medium for general operation environment | |
| CN110389820B (en) | Private cloud task scheduling method for resource prediction based on v-TGRU model | |
| Ayoubi et al. | An autonomous IoT service placement methodology in fog computing | |
| CN110471820B (en) | Cloud storage system disk fault prediction method based on cyclic neural network | |
| Teng et al. | Simmapreduce: A simulator for modeling mapreduce framework | |
| CN102523030B (en) | Communication satellite effective load test system simulation platform | |
| CN110249310A (en) | The resource management for virtual machine in cloud computing system | |
| CN108268589B (en) | Aggregate query method, apparatus, computer equipment and the readable medium of time series data | |
| CN101946260A (en) | Modelling computer based business process for customisation and delivery | |
| US8745637B2 (en) | Middleware for extracting aggregation statistics to enable light-weight management planners | |
| CN104050042A (en) | Resource allocation method and resource allocation device for ETL (Extraction-Transformation-Loading) jobs | |
| CN109857535A (en) | The implementation method and device of task priority control towards Spark JDBC | |
| CN106156413A (en) | A kind of Multiscale modelling method for designing towards large-scale distributed comprehensively modularized avionics system DIMA | |
| CN117076077A (en) | Planning and scheduling optimization method based on big data analysis | |
| CN112559122A (en) | Virtualization instance management and control method and system based on electric power special security and protection equipment | |
| CN106027642A (en) | Method and system for determining number of disks of CDN (Content Delivery Network) node | |
| CN111414961A (en) | Task parallel-based fine-grained distributed deep forest training method | |
| CN111984403A (en) | Method and system for distributing and deploying software components in distributed system | |
| CN110263371A (en) | IMA dynamic restructuring process configuration path generating method based on AADL | |
| CN115733888A (en) | Method for supporting multi-Internet-of-things application adaptive unloading under edge environment | |
| Zhang et al. | A Highly Reliable Data Storage Method for Power Data Centers with Hyper-Converged Architectures |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |