CN110084295A - Control method and control system are surrounded in a kind of grouping of multi-agent system - Google Patents
Control method and control system are surrounded in a kind of grouping of multi-agent system Download PDFInfo
- Publication number
- CN110084295A CN110084295A CN201910318251.7A CN201910318251A CN110084295A CN 110084295 A CN110084295 A CN 110084295A CN 201910318251 A CN201910318251 A CN 201910318251A CN 110084295 A CN110084295 A CN 110084295A
- Authority
- CN
- China
- Prior art keywords
- intelligent body
- grouping
- network
- frame
- agent system
- 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
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/213—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L41/00—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
- H04L41/06—Management of faults, events, alarms or notifications
- H04L41/0677—Localisation of faults
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L41/00—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
- H04L41/12—Discovery or management of network topologies
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L41/00—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
- H04L41/14—Network analysis or design
- H04L41/145—Network analysis or design involving simulating, designing, planning or modelling of a network
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
- Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
- Y02P90/02—Total factory control, e.g. smart factories, flexible manufacturing systems [FMS] or integrated manufacturing systems [IMS]
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Signal Processing (AREA)
- Computer Networks & Wireless Communication (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computing Systems (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Health & Medical Sciences (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Data Exchanges In Wide-Area Networks (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention belongs to computer application technology, discloses a kind of multi-agent system grouping and surround control method and control system, detected by the Sensor Network of known topological structure, determine multi-agent system topological structure；The closer and opposing stationary intelligent body in relative position is merged as one group with simplified control model；Model uses centerized fusion method, and intelligent body is grouped center control intelligent body group of the most intelligent body grouping of interior intelligent body number as Controlling model；Intelligent body grouping after grouping is encoded；Agent encoding in intelligent body grouping；Intelligent body grouping and intelligent body are determined according to encoding, and carry out corresponding control operation.The present invention carries out model foundation and amendment using the Sensor Network of known topological structure, and it is fast to save time, control speed；By the way of grouping, the processing of overall distribution formula, the processing of concentration of local formula are realized；It is controlled by the way of real-time coding, precision is high, and positioning is quasi-.
Description
Technical field
The invention belongs to computer application technologies more particularly to a kind of grouping of multi-agent system to surround control method
And control system.
Background technique
It is exactly to realize that one group of intelligent body passes through distributed network and local message pair that multiple agent, which surrounds target control algorithm,
The control algolithm that target in investigative range is surrounded.The generation of intelligent body concept leads artificial intelligence from early stage people
The research in domain, the mankind no longer need to handle the work of complicated danger in person, but replace the mankind to complete these by " agent Used "
Work is the major reason that artificial intelligence theory is suggested.Can by certain physical entity such as sensor, actuator, robot,
Bion, unmanned plane etc. regard an intelligent body as.With the development of agent theory, multi-agent system theory is also mentioned
Out, i.e., it is associated with by modes such as communications as a net by the intelligent body that a group has sensor, calculating, execution and communication capacity
Network system.For single intelligent body, it can only obtain limited information, so it is highly difficult for solving distributed problem.And it is right
For multi-agent system, by cooperating between intelligent body, the abilities of problems is solved considerably beyond single intelligence
It can body.Compared to single multiagent system, multi-agent system possesses more advantages, is mainly manifested in following several respects: effect
Rate is higher, system performance is more preferable, robustness and fault-tolerance are stronger, price is lower, using and it is easy to operate etc..
In conclusion problem of the existing technology:
In the network system of existing single intelligent body, limited information is obtained, can not be solved the problems, such as distributed.It is existing more
The control method convergence rate of intelligent body is slow, cannot achieve quick control, and in practical application, control precision is not high, and positioning is not
Accurately.
Summary of the invention
In view of the problems of the existing technology, it is grouped the present invention provides a kind of multi-agent system and surrounds control method,
Belong to energy-saving efficient and the invention of technological improvement type.
Multi-agent system grouping provided by the invention surrounds control method and includes:
Step 1 is detected by the Sensor Network of known topological structure, determines multi-agent system topological structure；
Step 2 merges the closer and opposing stationary intelligent body in relative position for one group with simplified control model；
Step 3, model use centerized fusion method, and intelligent body is grouped the most intelligent body point of interior intelligent body number
Center control intelligent body group of the group as Controlling model；
Step 4 encodes the intelligent body grouping after grouping；
Step 5, intelligent body grouping in agent encoding；
Step 6 determines intelligent body grouping and intelligent body according to encoding, and carries out corresponding control operation.
In step 4, coding mode are as follows: the grouping of all intelligent bodies is placed in two dimensional vector space, establishes Laplce
The boundary of matrix, ensemble space is the boundary of matrix, establishes rectangular coordinate system in space with lower left corner boundary, intelligent body is grouped in
Coordinate in matrix is the intelligent body block encoding.
In step 5, it is grouped interior agent encoding priority successively are as follows: high and low, upper and lower, left and right, it is strongest to calculate power
Centered on intelligent body, encoded using relative position.
Further, step 2 merges the closer and opposing stationary intelligent body in relative position for one group with simplified control
In model,
Intelligent body evaluation model is chosen, by comparing original to the nearlyr intelligent body in position and opposing stationary intelligent body, is calculated
The prediction score value of the opposing stationary every frame of intelligent body, and the frame level fractional marks that will acquire are vector X, intelligent body totalframes is labeled as
N；
The length of window of sliding window is winLen, carries out slide window processing to the frame level mass fraction of acquisition, that is, after handling
The frame level score of n-th frame is the mean value of the frame level score of [n-winLen+1, n] frame, by the frame level fractional marks after slide window processing
For vector WX；
It is ranked up WX is ascending, and is WX ' by the result queue after sequence, take the average value of worst p% frame,
It as the quality metric score value of entire intelligent body topological structure sequence, is ranked up, the smallest p% frame mean value is final measurement
As a result；
All frame level scores that intelligent body evaluation model is calculated successively carry out slide window processing, it may be assumed that
Wherein, winLen indicates length of window when sliding window filters, and is the parameter for needing to adjust, and X (t) indicates t frame
Mass fraction, WX (n) then indicate the mass fraction of the n-th frame after slide window processing；
Prediction frame level score is merged with time slot worst time-domain information fusion method using based on intra-frame trunk, most
Whole prediction score:
Wherein, p% is parameter to be adjusted, and N is intelligent body topological structure totalframes, and WX ' (t) indicates ascending and be ranked up
T-th of frame level score afterwards, OMwinPoolingFor the final appraisal results of the quality of the intelligent body.
Further, the final appraisal results of the quality of the intelligent body of acquisition are uniformly processed, construct unified control
Model.
Further, the final appraisal results of the quality of the intelligent body of acquisition are uniformly processed including data decimation, number
Data preprocess and data transformation.
Further, the data transformation includes that data convert dimensionality reduction, and the data transformation dimensionality reduction is from data initial characteristics
In the feature that is extracted by clustering, cut down data dimension.
Further, constructing unified Controlling model process includes: to choose with reference to warning information parameter vector, establishes reference number
Arrange X0,
X0={ X0(k) | k=1,2 ..., n }=(X0(1),X0(2),…,X0(n))
Wherein k indicates moment, X0Indicate that warning information, n indicate warning information parameter vector intrinsic dimensionality；
Secondly, it is assumed that have m relatively fault warning information data, ordered series of numbers X is compared in foundationi
Xi={ Xi(k) | k=1,2 ..., n }=(Xi(1),Xi(2),…,Xi(n)) i=1,2 ..., m
Then, it establishes and compares ordered series of numbers XiTo reference sequence X0In the incidence coefficient ζ at k momenti(k)
Wherein, w1For the corresponding weight of parameters, it is adjusted and determines according to the network attribute of user；Wherein ρ
For resolution ratio, ρ ∈ [0 ,+∞)；ρ is bigger, and resolution is bigger；ρ is smaller, and resolution is smaller；
Compare ordered series of numbers X finally, calculatingiTo reference sequence X0The degree of association
Further, constructing unified Controlling model further comprises by association analysis to intelligent body warning information and net
Failure establishes mapping relations between network, constructs fault location model, carries out network failure positioning by BP neural network；Including such as
Lower process:
Firstly, obtaining m dimension alarm vector Qn=(s1,s2,s3…sm) and n dimension fault vectors On=(p1,p2,p3…pm), and
It is inputted simultaneously by multiple network nodes, makes system that there is parallel organization and parallel processing capability, input is carried out real-time
Dynamic processing；
Secondly, BP network, is that the weight of each connection assigns the value in specified range, while referring to for each neuron node
Determine threshold value；
Again, group input alarm sample machine objective result is supplied to network, and calculates the connection weight of neural network node
Value, the input and output value of threshold value and each implicit layer unit；
Then, just each layer error: the error of output layer unit is calculated using object vector and network real output value, and is tied
The output of each unit of hidden layer is closed to correct connection weight and threshold value, carries out back-propagation amendment；
Finally, training sample vector sum is trained heterogeneous network system, network is transported after the complete whole samples of training
Fault warning information is tieed up, trained BP network is inputted and carries out network failure positioning.
Implement the more of the multi-agent system grouping encirclement control method another object of the present invention is to provide a kind of
Control system is surrounded in multiagent system grouping.
In conclusion advantages of the present invention and good effect are as follows:
The present invention carries out model foundation and amendment using the Sensor Network of known topological structure, and it is fast to save time, control speed；
The present invention realizes the processing of overall distribution formula, the processing of concentration of local formula by the way of grouping；By the way of real-time coding into
Row control, precision is high, and positioning is quasi-.Intelligent body is controlled using neural metwork training center, improves convergence rate.
The present invention merges the closer and opposing stationary intelligent body in relative position for one group in simplified control model,
Intelligent body evaluation model is chosen, by comparing original to the nearlyr intelligent body in position and opposing stationary intelligent body, is calculated opposing stationary
The prediction score value of the every frame of intelligent body, and the frame level fractional marks that will acquire are vector X, intelligent body totalframes is labeled as N；Sliding window
The length of window of mouth is winLen, carries out slide window processing to the frame level mass fraction of acquisition, that is, the frame level score of n-th frame after handling
It is vector WX by the frame level fractional marks after slide window processing for the mean value of the frame level score of [n-winLen+1, n] frame；By WX by
Small to being ranked up greatly, and by the result queue after sequence is WX ', the average value of worst p% frame is taken, as entire intelligent body
The quality metric score value of topological structure sequence, is ranked up, and the smallest p% frame mean value is final measurement results；By intelligent body
All frame level scores that evaluation model calculates successively carry out slide window processing, it may be assumed that
Wherein, winLen indicates length of window when sliding window filters, and is desirable
The parameter of adjusting, X (t) indicate that the mass fraction of t frame, WX (n) then indicate the mass fraction of the n-th frame after slide window processing；Make
Prediction frame level score is merged with time slot worst time-domain information fusion method with based on intra-frame trunk, final prediction point
Number:
Wherein, p% is parameter to be adjusted, and N is the total frame of intelligent body topological structure
Number, WX ' (t) indicate it is ascending be ranked up after t-th of frame level score, OMwinPoolingFor the intelligent body quality most
Whole evaluation result.
The present invention considers contacting for frame and interframe, using the data of each frame of sliding window average value processing, so that estimation accuracy
It greatly promotes.
By being optimized to network parameter, realizes network self-healing, realize network intelligence O&M.The present invention is dug using data
BP neural network method in pick excavates network O&M data and warning information, in conjunction with user's perception information, to network
Failure carry out intelligent positioning, accurate judgement network failure, promoted network quality, present invention reduces operation cost, improve system
System performance, it is ensured that the network operation is efficient, safe and stable.
Detailed description of the invention
Fig. 1 is that control method flow chart is surrounded in multi-agent system grouping provided by the invention.
Specific embodiment
In order to further understand the content, features and effects of the present invention, the following examples are hereby given, and cooperate attached drawing
Detailed description are as follows.
Existing single intelligent body intelligently obtains limited information, can not solve the problems, such as distributed.Existing multiple agent
Control method convergence rate is slow, cannot achieve quick control, and in practical application, control precision is not high, position inaccurate.
In order to solve the above technical problems, being explained in detail with reference to the accompanying drawing to structure of the invention.
As shown in Figure 1, multi-agent system grouping encirclement control method provided in an embodiment of the present invention includes:
S101 is detected by the Sensor Network of known topological structure, determines multi-agent system topological structure.
S102 merges the closer and opposing stationary intelligent body in relative position for one group with simplified control model.
S103, model use centerized fusion method, and intelligent body is grouped the most intelligent body of interior intelligent body number and is grouped
As the center control intelligent body group of Controlling model, center control intelligent body group is trained using BP neural network.
S104 encodes the intelligent body grouping after grouping.
S105 encodes the intelligent body in intelligent body grouping.
S106, the center control intelligent body group after training is grouped by intelligent body and the coding of intelligent body, control are corresponding
Intelligent body grouping and intelligent body.
In step S103, the training method provided in an embodiment of the present invention based on BP neural network is specifically included:
(1) 3 layers of neural network of one input layer of component, a hidden layer, output layer.
(2) weight of basic BP neural network and the adjustment formula of threshold value are as follows:
wkj(t+1)=wkj(t)+αδkHj
wji(t+1)=wji(t)+αδjIi
θk(t+1)=θk(t)+βδk
θj(t+1)=θj(t)+βδj
In above equation, HjFor the output of hidden layer node.IjFor the signal of input node i input.wkj(t) and wkj(t+
It 1) is respectively hidden node j and the connection weight for exporting node layer k when front and back is trained twice.wji(t) and wji(t+1) before being respectively
The connection weight of input node i and hidden node j when training twice afterwards.θkAnd θjThreshold at respectively output node k and hidden node j
Value.α and β is respectively learning parameter.δkAnd δjRespectively export the error signal of node layer k and hidden node j, calculating formula are as follows:
δk=(Tk-Ok)Ok(1-Ok)
In formula, TkTarget output value for sample in output node, OkAnd HjRespectively sample network output node and
The real output value of hidden node.
Wherein hidden layer exports calculation formula are as follows:
In formula, M is input node number.F is S type activation primitive:
Output layer output is summed using linear weighted function:
In formula, s is hidden node number.
In step S104, coding mode provided in an embodiment of the present invention is specifically included:
The grouping of all intelligent bodies is placed in two dimensional vector space, Laplacian Matrix, the boundary of ensemble space are established
Rectangular coordinate system in space is established with lower left corner boundary in the as boundary of matrix, and the coordinate of intelligent body grouping in a matrix is should
Intelligent body block encoding.
In step S105, agent encoding provided in an embodiment of the present invention is specifically included:
Agent encoding priority is successively in being grouped are as follows: high and low, upper and lower, left and right, to calculate during the strongest intelligent body of power is
The heart is encoded using relative position.
In embodiments of the present invention, it is one that step S102, which merges the closer and opposing stationary intelligent body in relative position,
Group in simplified control model,
Intelligent body evaluation model is chosen, by comparing original to the nearlyr intelligent body in position and opposing stationary intelligent body, is calculated
The prediction score value of the opposing stationary every frame of intelligent body, and the frame level fractional marks that will acquire are vector X, intelligent body totalframes is labeled as
N。
The length of window of sliding window is winLen, carries out slide window processing to the frame level mass fraction of acquisition, that is, after handling
The frame level score of n-th frame is the mean value of the frame level score of [n-winLen+1, n] frame, by the frame level fractional marks after slide window processing
For vector WX.
It is ranked up WX is ascending, and is WX ' by the result queue after sequence, take the average value of worst p% frame,
It as the quality metric score value of entire intelligent body topological structure sequence, is ranked up, the smallest p% frame mean value is final measurement
As a result.
All frame level scores that intelligent body evaluation model is calculated successively carry out slide window processing, it may be assumed that
Wherein, winLen indicates length of window when sliding window filters, and is the parameter for needing to adjust, and X (t) indicates t frame
Mass fraction, WX (n) then indicate the mass fraction of the n-th frame after slide window processing.
Prediction frame level score is merged with time slot worst time-domain information fusion method using based on intra-frame trunk, most
Whole prediction score:
Wherein, p% is parameter to be adjusted, and N is intelligent body topological structure totalframes, and WX ' (t) indicates ascending and be ranked up
T-th of frame level score afterwards, OMwinPoolingFor the final appraisal results of the quality of the intelligent body.
The final appraisal results of the quality of the intelligent body of acquisition are uniformly processed, unified Controlling model is constructed.
The final appraisal results of the quality of the intelligent body of acquisition are uniformly processed including data decimation, data prediction
It is converted with data.The data transformation includes that data convert dimensionality reduction, and the data transformation dimensionality reduction is to lead to from data initial characteristics
The feature that clustering extracts is crossed, data dimension is cut down.
Constructing unified Controlling model process includes: to choose with reference to warning information parameter vector, establishes reference sequence X0,
X0={ X0(k) | k=1,2 ..., n }=(X0(1),X0(2),…,X0(n))
Wherein k indicates moment, X0Indicate that warning information, n indicate warning information parameter vector intrinsic dimensionality；
Secondly, it is assumed that have m relatively fault warning information data, ordered series of numbers X is compared in foundationi
Xi={ Xi(k) | k=1,2 ..., n }=(Xi(1),Xi(2),…,Xi(n)) i=1,2 ..., m
Then, it establishes and compares ordered series of numbers XiTo reference sequence X0In the incidence coefficient ζ at k momenti(k)
Wherein, w1For the corresponding weight of parameters, it is adjusted and determines according to the network attribute of user；Wherein ρ
For resolution ratio, ρ ∈ [0 ,+∞)；ρ is bigger, and resolution is bigger；ρ is smaller, and resolution is smaller；
Compare ordered series of numbers X finally, calculatingiTo reference sequence X0The degree of association
Constructing unified Controlling model further comprises by association analysis between intelligent body warning information and network
Failure establishes mapping relations, constructs fault location model, carries out network failure positioning by BP neural network；Including following mistake
Journey:
Firstly, obtaining m dimension alarm vector Qn=(s1,s2,s3…sm) and n dimension fault vectors On=(p1,p2,p3…pm), and
It is inputted simultaneously by multiple network nodes, makes system that there is parallel organization and parallel processing capability, input is carried out real-time
Dynamic processing；
Secondly, BP network, is that the weight of each connection assigns the value in specified range, while referring to for each neuron node
Determine threshold value；
Again, group input alarm sample machine objective result is supplied to network, and calculates the connection weight of neural network node
Value, the input and output value of threshold value and each implicit layer unit；
Then, just each layer error: the error of output layer unit is calculated using object vector and network real output value, and is tied
The output of each unit of hidden layer is closed to correct connection weight and threshold value, carries out back-propagation amendment；
Finally, training sample vector sum is trained heterogeneous network system, network is transported after the complete whole samples of training
Fault warning information is tieed up, trained BP network is inputted and carries out network failure positioning.
The above is only the preferred embodiments of the present invention, and is not intended to limit the present invention in any form,
Any simple modification made to the above embodiment according to the technical essence of the invention, equivalent variations and modification, belong to
In the range of technical solution of the present invention.
Claims (10)
1. control method is surrounded in a kind of multi-agent system grouping, which is characterized in that control is surrounded in the multi-agent system grouping
Method processed includes:
Step 1 is detected by the Sensor Network of known topological structure, determines multi-agent system topological structure；
Step 2 merges the closer and opposing stationary intelligent body in relative position for one group with simplified control model；
Step 3, Controlling model use centerized fusion method, and intelligent body is grouped the most intelligent body point of interior intelligent body number
Center control intelligent body group of the group as Controlling model；
Step 4 encodes the intelligent body grouping after grouping；
Step 5, intelligent body grouping in carry out agent encoding；
Step 6 determines intelligent body grouping and intelligent body according to encoding, and carries out corresponding control operation.
2. control method is surrounded in multi-agent system grouping as described in claim 1, which is characterized in that in step 4, coding staff
Formula are as follows: the grouping of all intelligent bodies is placed in two dimensional vector space, establishes Laplacian Matrix, the boundary of ensemble space is
Rectangular coordinate system in space is established with lower left corner boundary in the boundary of matrix, and the coordinate of intelligent body grouping in a matrix is the intelligence
Body block encoding.
3. control method is surrounded in multi-agent system grouping as described in claim 1, which is characterized in that in step 5, in grouping
Agent encoding priority is successively are as follows: high and low, upper and lower, left and right, centered on calculating the strongest intelligent body of power, using opposite position
Set coding.
4. control method is surrounded in multi-agent system as described in claim 1 grouping, which is characterized in that step 2 is by relative position
Closer and opposing stationary intelligent body merges as one group in simplified control model,
Intelligent body evaluation model is chosen, by comparing original to the nearlyr intelligent body in position and opposing stationary intelligent body, is calculated opposite
The prediction score value of the static every frame of intelligent body, and the frame level fractional marks that will acquire are vector X, intelligent body totalframes is labeled as N；
The length of window of sliding window is winLen, carries out slide window processing to the frame level mass fraction of acquisition, that is, n-th frame after handling
Frame level score be [n-winLen+1, n] frame frame level score mean value, by the frame level fractional marks after slide window processing be vector
WX；
It is ranked up WX is ascending, and is WX ' by the result queue after sequence, take the average value of worst p% frame, as
The quality metric score value of entire intelligent body topological structure sequence, is ranked up, and the smallest p% frame mean value is final measurement knot
Fruit；
All frame level scores that intelligent body evaluation model is calculated successively carry out slide window processing, it may be assumed that
Wherein, winLen indicates length of window when sliding window filters, and is the parameter for needing to adjust, and X (t) indicates the quality of t frame
Score, WX (n) then indicate the mass fraction of the n-th frame after slide window processing；
Prediction frame level score is merged with time slot worst time-domain information fusion method using based on intra-frame trunk, final
Predict score:
Wherein, p% be parameter to be adjusted, N be intelligent body topological structure totalframes, WX ' (t) indicate it is ascending be ranked up after
T-th of frame level score, OMwinPoolingFor the final appraisal results of the quality of the intelligent body.
5. control method is surrounded in multi-agent system grouping as claimed in claim 4, which is characterized in that
The final appraisal results of the quality of the intelligent body of acquisition are uniformly processed, unified Controlling model is constructed.
6. control method is surrounded in multi-agent system grouping as claimed in claim 5, which is characterized in that
The final appraisal results of the quality of the intelligent body of acquisition are uniformly processed including data decimation, data prediction sum number
According to transformation.
7. control method is surrounded in multi-agent system grouping as claimed in claim 6, which is characterized in that
The data transformation includes that data convert dimensionality reduction, and the data transformation dimensionality reduction is from data initial characteristics by cluster point
The feature extracted is analysed, data dimension is cut down.
8. control method is surrounded in multi-agent system grouping as claimed in claim 5, which is characterized in that
Constructing unified Controlling model process includes: to choose with reference to warning information parameter vector, establishes reference sequence X0,
X0={ X0(k) | k=1,2 ..., n }=(X0(1),X0(2),…,X0(n))
Wherein k indicates moment, X0Indicate that warning information, n indicate warning information parameter vector intrinsic dimensionality；
Secondly, it is assumed that have m relatively fault warning information data, ordered series of numbers X is compared in foundationi
Xi={ Xi(k) | k=1,2 ..., n }=(Xi(1),Xi(2),…,Xi(n)) i=1,2 ..., m
Then, it establishes and compares ordered series of numbers XiTo reference sequence X0In the incidence coefficient ζ at k momenti(k)
Wherein, w1For the corresponding weight of parameters, it is adjusted and determines according to the network attribute of user；Wherein ρ is to differentiate
Coefficient, ρ ∈ [0 ,+∞)；ρ is bigger, and resolution is bigger；ρ is smaller, and resolution is smaller；
Compare ordered series of numbers X finally, calculatingiTo reference sequence X0The degree of association
9. control method is surrounded in multi-agent system grouping as claimed in claim 5, which is characterized in that construct unified control mould
Type further comprises establishing mapping relations to failure between intelligent body warning information and network by association analysis, constructs failure
Location model carries out network failure positioning by BP neural network；It comprises the following processes:
Firstly, obtaining m dimension alarm vector Qn=(s1,s2,s3…sm) and n dimension fault vectors On=(p1,p2,p3…pm), and by its
It is inputted simultaneously by multiple network nodes, makes system that there is parallel organization and parallel processing capability, input is moved in real time
State processing；
Secondly, BP network, is that the weight of each connection assigns the value in specified range, while specifying threshold for each neuron node
Value；
Again, group input alarm sample machine objective result is supplied to network, and calculates the connection weight of neural network node, threshold
The input and output value of value and each implicit layer unit；
Then, just each layer error: the error of output layer unit is calculated using object vector and network real output value, and is combined hidden
Connection weight and threshold value are corrected in the output of each unit containing layer, carry out back-propagation amendment；
Finally, training sample vector sum is trained heterogeneous network system, by the event of network O&M after the complete whole samples of training
Hinder warning information, inputs trained BP network and carry out network failure positioning.
10. a kind of multi-agent system grouping packet implemented the grouping of multi-agent system described in claim 1 and surround control method
Enclose control system.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910318251.7A CN110084295B (en) | 2019-04-19 | 2019-04-19 | Multi-agent system grouping surrounding control method and control system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910318251.7A CN110084295B (en) | 2019-04-19 | 2019-04-19 | Multi-agent system grouping surrounding control method and control system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110084295A true CN110084295A (en) | 2019-08-02 |
| CN110084295B CN110084295B (en) | 2023-05-05 |
Family
ID=67415747
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910318251.7A Active CN110084295B (en) | 2019-04-19 | 2019-04-19 | Multi-agent system grouping surrounding control method and control system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110084295B (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110503195A (en) * | 2019-08-14 | 2019-11-26 | 北京中科寒武纪科技有限公司 | The method and its Related product of task are executed using artificial intelligence process device |
| CN111766783A (en) * | 2020-06-29 | 2020-10-13 | 北京航空航天大学 | Cluster system-oriented formation enclosure tracking method capable of converging in limited time |
| CN113146624A (en) * | 2021-03-25 | 2021-07-23 | 重庆大学 | Multi-agent control method based on maximum angle aggregation strategy |
Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20060221073A1 (en) * | 2005-03-04 | 2006-10-05 | Aiseek Ltd | Techniques for accurately determining visibility of objects from multiple viewpoints in a three-dimensional (3D) environment |
| CN104469833A (en) * | 2015-01-08 | 2015-03-25 | 西安电子科技大学 | Heterogeneous network operation and maintenance management method based on user perception |
| CN105429185A (en) * | 2015-12-08 | 2016-03-23 | 华南理工大学 | Economic dispatching method with robust collaborative consistency |
| CN105589333A (en) * | 2016-02-03 | 2016-05-18 | 电子科技大学 | Grouped surrounding control method of multi-agent system |
| CN105979266A (en) * | 2016-05-06 | 2016-09-28 | 西安电子科技大学 | Interframe relevance and time slot worst based time domain information fusion method |
| CN106502090A (en) * | 2016-09-29 | 2017-03-15 | 电子科技大学 | The positioning of multi-agent system and circular control method in a kind of three dimensions |
| CN108092833A (en) * | 2018-02-28 | 2018-05-29 | 哈尔滨理工大学 | A kind of multiple agent uniformity method for diagnosing faults |
| CN108681320A (en) * | 2018-05-11 | 2018-10-19 | 北京理工大学 | A kind of distributed multi agent real-time fault detection method based on regional cooperative |
| CN109299305A (en) * | 2018-10-30 | 2019-02-01 | 湖北工业大学 | A kind of spatial image searching system based on multi-feature fusion and search method |
-
2019
- 2019-04-19 CN CN201910318251.7A patent/CN110084295B/en active Active
Patent Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20060221073A1 (en) * | 2005-03-04 | 2006-10-05 | Aiseek Ltd | Techniques for accurately determining visibility of objects from multiple viewpoints in a three-dimensional (3D) environment |
| CN104469833A (en) * | 2015-01-08 | 2015-03-25 | 西安电子科技大学 | Heterogeneous network operation and maintenance management method based on user perception |
| CN105429185A (en) * | 2015-12-08 | 2016-03-23 | 华南理工大学 | Economic dispatching method with robust collaborative consistency |
| CN105589333A (en) * | 2016-02-03 | 2016-05-18 | 电子科技大学 | Grouped surrounding control method of multi-agent system |
| CN105979266A (en) * | 2016-05-06 | 2016-09-28 | 西安电子科技大学 | Interframe relevance and time slot worst based time domain information fusion method |
| CN106502090A (en) * | 2016-09-29 | 2017-03-15 | 电子科技大学 | The positioning of multi-agent system and circular control method in a kind of three dimensions |
| CN108092833A (en) * | 2018-02-28 | 2018-05-29 | 哈尔滨理工大学 | A kind of multiple agent uniformity method for diagnosing faults |
| CN108681320A (en) * | 2018-05-11 | 2018-10-19 | 北京理工大学 | A kind of distributed multi agent real-time fault detection method based on regional cooperative |
| CN109299305A (en) * | 2018-10-30 | 2019-02-01 | 湖北工业大学 | A kind of spatial image searching system based on multi-feature fusion and search method |
Non-Patent Citations (3)
| Title |
|---|
| DUAN MIN ET AL.: "Distributed Encirclement Control of Multi-agent Systems", 《PROCEEDINGS OF THE 33RD CHINESE CONTROL CONFERENCE》 * |
| ZHI-MIN HAN ET AL.: "Distributed coordination inmulti-agent systems:a graph Laplacian perspective", 《FRONTIERS OF INFORMATION TECHNOLOGY & ELECTRONIC ENGINEERING》 * |
| 魏婷婷: "多智能体***协同包围目标控制算法的研究", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110503195A (en) * | 2019-08-14 | 2019-11-26 | 北京中科寒武纪科技有限公司 | The method and its Related product of task are executed using artificial intelligence process device |
| CN111766783A (en) * | 2020-06-29 | 2020-10-13 | 北京航空航天大学 | Cluster system-oriented formation enclosure tracking method capable of converging in limited time |
| CN111766783B (en) * | 2020-06-29 | 2021-08-24 | 北京航空航天大学 | Cluster system-oriented formation enclosure tracking method capable of converging in limited time |
| CN113146624A (en) * | 2021-03-25 | 2021-07-23 | 重庆大学 | Multi-agent control method based on maximum angle aggregation strategy |
| CN113146624B (en) * | 2021-03-25 | 2022-04-29 | 重庆大学 | Multi-agent control method based on maximum angle aggregation strategy |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110084295B (en) | 2023-05-05 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109492822B (en) | Air pollutant concentration time-space domain correlation prediction method | |
| CN108536123B (en) | Train control on board equipment method for diagnosing faults based on long Memory Neural Networks in short-term | |
| CN110110707A (en) | Artificial intelligence CNN, LSTM neural network dynamic identifying system | |
| CN110084295A (en) | Control method and control system are surrounded in a kind of grouping of multi-agent system | |
| CN109598386A (en) | A kind of accurate analysis method of distribution optimization and system based on deep learning | |
| CN110751318B (en) | Ultra-short-term power load prediction method based on IPSO-LSTM | |
| CN105488456B (en) | Method for detecting human face based on adaptive threshold adjustment rejection sub-space learning | |
| CN110147877A (en) | PM2.5 prediction technique based on the long network in short-term of convolution | |
| CN107423839A (en) | A kind of method of the intelligent building microgrid load prediction based on deep learning | |
| CN109344992B (en) | Modeling method for user control behavior habits of smart home integrating time-space factors | |
| CN107871105A (en) | Face authentication method and device | |
| CN113298191B (en) | User behavior identification method based on personalized semi-supervised online federal learning | |
| CN110197307B (en) | Regional sea surface temperature prediction method combined with attention mechanism | |
| CN109214503A (en) | Project of transmitting and converting electricity cost forecasting method based on KPCA-LA-RBM | |
| CN108520213A (en) | A kind of face beauty prediction technique based on multiple dimensioned depth | |
| CN111222992A (en) | Stock price prediction method of long-short term memory neural network based on attention mechanism | |
| CN109389588A (en) | The method for measuring difference between video successive frame and its convolution characteristic pattern | |
| CN107453921A (en) | Smart city system artificial intelligence evaluation method based on nonlinear neural network | |
| CN116029435B (en) | Environmental comfort early warning system is bred to live pig facility | |
| CN113569928A (en) | Train running state detection data missing processing model and reconstruction method | |
| CN108460462A (en) | A kind of Interval neural networks learning method based on interval parameter optimization | |
| CN115081717A (en) | Rail transit passenger flow prediction method integrating attention mechanism and graph neural network | |
| Liu et al. | A differential evolution flower pollination algorithm with dynamic switch probability | |
| CN109284720A (en) | Measure application of the difference in video Activity recognition between video successive frame and its convolution characteristic pattern | |
| CN111340187A (en) | Network characterization method based on counter attention mechanism |
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 |