CN112714165A - Distributed network cooperation strategy optimization method and device based on combination mechanism - Google Patents
Distributed network cooperation strategy optimization method and device based on combination mechanism Download PDFInfo
- Publication number
- CN112714165A CN112714165A CN202011530099.8A CN202011530099A CN112714165A CN 112714165 A CN112714165 A CN 112714165A CN 202011530099 A CN202011530099 A CN 202011530099A CN 112714165 A CN112714165 A CN 112714165A
- Authority
- CN
- China
- Prior art keywords
- combination
- preset
- coefficient
- cooperation
- distributed network
- 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
Images
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L67/00—Network arrangements or protocols for supporting network services or applications
- H04L67/01—Protocols
- H04L67/10—Protocols in which an application is distributed across nodes in the network
- H04L67/104—Peer-to-peer [P2P] networks
- H04L67/1061—Peer-to-peer [P2P] networks using node-based peer discovery mechanisms
- H04L67/1065—Discovery involving distributed pre-established resource-based relationships among peers, e.g. based on distributed hash tables [DHT]
-
- 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
- H04L41/142—Network analysis or design using statistical or mathematical methods
-
- 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
- Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
- Y02D30/00—Reducing energy consumption in communication networks
- Y02D30/70—Reducing energy consumption in communication networks in wireless communication networks
Landscapes
- Engineering & Computer Science (AREA)
- Computer Networks & Wireless Communication (AREA)
- Signal Processing (AREA)
- Physics & Mathematics (AREA)
- Algebra (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Analysis (AREA)
- Mathematical Optimization (AREA)
- Mathematical Physics (AREA)
- Probability & Statistics with Applications (AREA)
- Pure & Applied Mathematics (AREA)
- Data Exchanges In Wide-Area Networks (AREA)
Abstract
The application discloses a distributed network cooperation strategy optimization method based on a combination mechanism, which comprises the steps of obtaining initial estimation results of a preset node under a plurality of candidate cooperation strategies of a distributed network at a preset moment; assigning a respective initial combining coefficient to each of the initial estimation results; according to a preset combination mechanism, the initial combination coefficient is subjected to self-adaptive adjustment to obtain an optimized combination coefficient which enables the performance of the cooperation strategy to be optimized; and obtaining an optimized estimation result of the preset node by using the optimized combination coefficient and the candidate cooperation strategy. The distributed network cooperation strategy optimization method based on the combination mechanism can overcome the defect of a single cooperation strategy, and enables the distributed network to achieve the best cooperation strategy performance in different application situations. The application also discloses a distributed network cooperation strategy optimization device based on a combination mechanism, a computer device and a computer readable storage medium, which have the same advantages.
Description
Technical Field
The invention belongs to the technical field of self-adaptive signal processing, and particularly relates to a distributed network cooperation strategy optimization method and device based on a combination mechanism, computer equipment and a computer readable storage medium.
Background
In recent years, with the rapid development of artificial multi-agent networks and the deep understanding of human beings on natural multi-agent networks, the adaptive learning technology in the distributed network has become an important concern in the current signal processing field, and has a promoting effect on the problems of signal detection, target positioning and tracking, distributed voice enhancement, spectrum sensing, distributed image processing, biological clustering behavior simulation and the like in the multi-agent networks. In the existing adaptive signal processing field, an adaptive filter is used, the adaptive filter can adapt to or track the non-stationary random change of the external environment through self-learning of real-time stream data under the condition of no priori knowledge, and finally approaches the performance of the optimal filter under a certain criterion, and the adaptive signal processing technology is applied to the fields of communication, control, radar, sonar, earthquake, biomedicine and the like. Compared with the traditional adaptive signal processing technology, the network node not only needs to perform model estimation, signal prediction or anomaly detection according to self-observation data, but also more importantly, in a certain associated sensing area, the node needs to complete a specific global task with other connected neighbor nodes in a self-organized cooperative mode.
Specifically, the collaboration policy in the distributed network includes three types: the method comprises an increment strategy, a consistency strategy and a diffusion strategy, wherein the increment strategy requires information flow in a network to form an annular graph, namely a Hamiltonian loop, although theoretically the increment strategy requires small communication traffic and a simple communication model, establishing the Hamiltonian loop in the network is an NP difficult problem, and in addition, the loop is very sensitive to failure of nodes or links, so the increment strategy is not completely suitable for distributed online adaptive signal processing; in the consistency strategy and the diffusion strategy, each node needs to communicate with the neighbor nodes in real time, and utilizes the information exchange between the node and the neighbor nodes to cooperatively estimate the global target parameters in the network, and each node needs to acquire the information of all the neighbor nodes at each moment, so that the two strategies need more communication resources than an increment strategy, but can fully utilize the cooperation of the nodes in a distributed network structure.
However, the single cooperation strategy adopted in the prior art is limited by its design assumptions, so that the performance is better only in some situations where these assumptions are satisfied, and the performance of the best strategy cannot be achieved in different application situations.
Disclosure of Invention
In order to solve the above problems, the present invention provides a method, an apparatus, a computer device and a computer readable storage medium for optimizing a distributed network cooperation policy based on a combination mechanism, which can overcome the defect of a single cooperation policy, and enable the distributed network to achieve the best performance of the cooperation policy in different application situations.
The invention provides a distributed network cooperation strategy optimization method based on a combined mechanism, which comprises the following steps:
at a preset moment, acquiring an initial estimation result of a preset node under a plurality of candidate cooperation strategies of a distributed network;
assigning a respective initial combining coefficient to each of the initial estimation results;
according to a preset combination mechanism, the initial combination coefficient is subjected to self-adaptive adjustment to obtain an optimized combination coefficient which enables the performance of the cooperation strategy to be optimized;
and obtaining an optimized estimation result of the preset node by using the optimized combination coefficient and the candidate cooperation strategy.
Preferably, in the method for optimizing a distributed network cooperation strategy based on a combination mechanism, the initial combination coefficient includes at least one of a convex combination coefficient and an affine combination coefficient.
Preferably, in the method for optimizing a distributed network cooperation policy based on a combination mechanism, the candidate cooperation policy includes at least one of a plurality of different types of cooperation policies with the same parameter, a plurality of same types of cooperation policies with different parameters, and a plurality of different types of cooperation policies with different parameters.
Preferably, in the method for optimizing a distributed network cooperation policy based on a combination mechanism, when the number of candidate cooperation policies is 2, the preset combination mechanism is at least one of an affine energy normalization mechanism, an affine symbol regression mechanism, a convex energy normalization mechanism and a convex symbol regression mechanism.
Preferably, in the method for optimizing distributed network cooperation policies based on a combination mechanism, when the number of candidate cooperation policies is greater than 2, the preset combination mechanism is at least one of an affine multi-policy mechanism and a convex multi-policy mechanism.
Preferably, in the method for optimizing distributed network cooperation policy based on combination mechanism, when the number of candidate cooperation policies is 2 and the preset combination mechanism is an affine energy normalization mechanism, the combination coefficient is adjusted according to the affine energy normalization mechanismAnd
wherein xk,nIs the input signal of node k at time n, ek,nIs the error signal of node k at time n, i.e. Is the step size parameter, pk,nIs calculated by the following iterative formula:the time smooth coefficient is more than 0 and less than 1, and the value range of epsilon is (0, 0.5)]. While
When the preset combination mechanism is an affine symbol regression mechanism, the combination coefficient is adjusted according to the affine symbol regression mechanismAnd
wherein ek,nIs the error signal of node k at time n, i.e. Is the step size parameter, sgn { x } is a sign function defined as:
When the preset combination mechanism is a convex energy normalization mechanism, the combination coefficient is adjusted according to the convex energy normalization mechanismAnd
xk,nis the input signal of node k at time n, ek,nIs the error signal of node k at time n, i.e. Is the step size parameter, pk,nIs calculated by the following iterative formula:the time smooth coefficient is more than 0 and less than 1, and the value range of epsilon is (0, 0.5)]At αk,nAfter the update is completed, it is limited to [ - α ]+,α+]To do soCalculated by the following formula:
when the preset combination mechanism is a convex symbol regression mechanism, the combination coefficient is adjusted according to the convex symbol regression mechanismAnd
ek,nis the error signal of node k at time n, i.e. Is the step size parameter, sgn { x } is a sign function defined as:
at αknAfter the update is completed, it is limited to [ - α ]+,α+]To do soCalculated by the following formula:
preferably, in the method for optimizing distributed network cooperation policy based on combination mechanism, when the number of candidate cooperation policies is greater than 2 and the preset combination mechanism is an affine multi-policy mechanism, the combination coefficient is adjusted according to the affine multi-policy mechanism
is the step size parameter, the estimation error ek,nCalculated by the following formula:whileDelta value range (0, 0.5)];
When the preset combination mechanism is a convex multi-strategy mechanism, the combination coefficient is adjusted according to the convex multi-strategy mechanism
is the step size parameter, the estimation error ek,nCalculated by the following formula:whileDelta value range (0, 0.5)]In aAfter the update is completed, it is limited to
The invention provides a distributed network cooperation strategy optimization device based on a combined mechanism, which comprises:
the initial estimation result acquisition unit is used for acquiring initial estimation results of a preset node under a plurality of candidate cooperation strategies of the distributed network at a preset moment;
an initial combination coefficient allocation unit, configured to allocate a corresponding initial combination coefficient to each of the initial estimation results;
the self-adaptive adjusting unit is used for carrying out self-adaptive adjustment on the initial combination coefficient according to a preset combination mechanism to obtain an optimized combination coefficient which enables the performance of the cooperation strategy to be optimized;
and the optimized estimation result calculation unit is used for obtaining the optimized estimation result of the preset node by using the optimized combination coefficient and the candidate cooperation strategy.
The invention provides a computer device comprising:
a memory for storing a computer program;
a processor for implementing the steps of the method for optimizing a distributed network cooperation strategy based on a combination mechanism according to any one of the above when the computer program is executed.
The invention provides a computer readable storage medium, which stores thereon a computer program, which when executed by a processor implements the steps of the method for optimizing a distributed network cooperation strategy based on a combination mechanism according to any one of the above.
As can be seen from the above description, according to the combination mechanism-based distributed network cooperation policy optimization method provided by the present invention, initial estimation results of a preset node under multiple candidate cooperation policies of a distributed network are obtained at a preset time, then a corresponding initial combination coefficient is allocated to each initial estimation result, the initial combination coefficients are adaptively adjusted according to a preset combination mechanism to obtain an optimized combination coefficient that optimizes the performance of a cooperation policy, and finally the optimized estimation results of the preset node are obtained by using the optimized combination coefficient and the candidate cooperation policies. The distributed network cooperation strategy optimization device based on the combination mechanism, the computer equipment and the computer readable storage medium have the same advantages as the method.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, it is obvious that the drawings in the following description are only embodiments of the present invention, and for those skilled in the art, other drawings can be obtained according to the provided drawings without creative efforts.
Fig. 1 is a schematic diagram of an embodiment of a distributed network cooperation policy optimization method based on a combination mechanism according to the present invention;
FIG. 2 is a schematic diagram of a combinational architecture of a collaboration policy of a distributed network based on a combinational mechanism;
fig. 3 is a diagram of simulation results of an affine energy normalization mechanism and an affine symbol regression mechanism when M is 2;
fig. 4 is a diagram of simulation results of the convex energy normalization mechanism and the convex symbol regression mechanism when M is 2;
FIG. 5 is a diagram of simulation results for an affine multi-policy mechanism when M > 2;
FIG. 6 is a diagram of simulation results for the convex multi-strategy mechanism when M > 2;
fig. 7 is a schematic diagram of an embodiment of a distributed network cooperation policy optimization apparatus based on a combination mechanism according to the present invention;
fig. 8 is a schematic diagram of an embodiment of a computer device provided in the present invention.
Detailed Description
The core of the invention is to provide a distributed network cooperation strategy optimization method and device based on a combination mechanism, which can overcome the defect of a single cooperation strategy, enable the distributed network to achieve the best performance of the cooperation strategy in different application situations, and can be applied to a diffusion strategy, an increment strategy and a consistency strategy.
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
In the embodiment of the present invention, the diffusion strategy is taken as an example for explanation, but the method provided by the present invention is not limited to this diffusion strategy. The signal model and the related quantity in the embodiment of the invention are as follows: considering a distributed network consisting of N nodes, at each node k an unknown parameter vector of length lx 1 needs to be estimatedAn input vector x of length Lx 1 can be observed at node kk,nAnd a reference signal dk,nAt time n, the data at node k is modeled by a linear modelIs related, wherein zk,nIs additive noise, node k needs to utilize the input vector xk,nAnd a reference signal dk,nTo estimate the unknown parameter vector
Fig. 1 shows an implementation of a distributed network cooperation policy optimization method based on a combination mechanism, where fig. 1 is a schematic diagram of an embodiment of a distributed network cooperation policy optimization method based on a combination mechanism, and the method may include the following steps:
s1: at a preset moment, acquiring an initial estimation result of a preset node under a plurality of candidate cooperation strategies of a distributed network;
taking the candidate cooperation policy as the candidate diffusion policy as an example for explanation, with reference to fig. 2, fig. 2 is a schematic diagram of a combination architecture of cooperation policies of a distributed network based on a combination mechanism, and the step may be to run M candidate diffusion policies S in parallel in the distributed network(1)、S(2)、……、S(M)The node k obtains a candidate diffusion strategy S at the moment n(i)Initial estimation result ofWhere i is 1,2, …, M.
S2: allocating a corresponding initial combination coefficient to each initial estimation result;
specifically, M candidate diffusion strategies S on node k may be performed at time n(1)、S(2)、……、S(M)Initial estimation result ofAssigning initial combining coefficients, respectively
S3: according to a preset combination mechanism, carrying out self-adaptive adjustment on the initial combination coefficient to obtain an optimized combination coefficient for optimizing the performance of the cooperation strategy;
the step can be specifically based on a preset combination mechanism, and the initial combination coefficient is adjusted adaptively The cooperation strategy which leads the tracking performance to be the best corresponds to the optimized combination coefficient.
S4: and obtaining an optimized estimation result of the preset node by using the optimized combination coefficient and the candidate cooperation strategy.
The optimal combination coefficient can be used to calculate the final output result of the distributed network
As can be seen from the above description, in the embodiment of the distributed network cooperation policy optimization method based on the combination mechanism, since the initial estimation results of the preset node under the multiple candidate cooperation policies of the distributed network are obtained at the preset time, then the corresponding initial combination coefficient is allocated to each initial estimation result, the initial combination coefficients are adaptively adjusted according to the preset combination mechanism to obtain the optimized combination coefficient that optimizes the performance of the cooperation policy, and finally the optimized estimation results of the preset node are obtained by using the optimized combination coefficient and the candidate cooperation policies, it can be seen that the multiple candidate cooperation policies are considered here, so that the defect of a single cooperation policy can be overcome, and the distributed network can achieve the best performance of the cooperation policy in different application situations.
Based on the embodiment of the method for optimizing the distributed network cooperation strategy based on the combination mechanism, a specific embodiment of another method provided by the present invention further preferably selects the initial combination coefficient to include at least one of a convex combination coefficient and an affine combination coefficient, that is, only the convex combination coefficient or only the affine combination coefficient may be included, or both of them may be included.
Specifically, the following may be satisfiedAffine combination coefficients of the following conditions:and a convex combination coefficient satisfying the following conditions:and isWhere i is 1,2, …, M.
Also, the candidate cooperation policy may preferably include at least one of a plurality of different types of cooperation policies having the same parameter, a plurality of the same type of cooperation policies having different parameters, and a plurality of different types of cooperation policies having different parameters.
The present invention provides another embodiment, which is a further optimization based on the embodiment of the distributed network cooperation policy optimization method based on the combination mechanism, that is, when the number of candidate cooperation policies is 2, the preset combination mechanism is at least one of an affine energy normalization mechanism, an affine symbol regression mechanism, a convex energy normalization mechanism, and a convex symbol regression mechanism.
The four mechanisms are described in detail below:
wherein xk,nIs the input signal of node k at time n, ek,nIs the error signal of node k at time n, i.e. Is the step size parameter, pk,nIs calculated by the following iterative formula:the time smoothing coefficient is 0 < eta < 1, epsilon is a small positive number and is used for avoiding the condition that the denominator is zero, and the value range of epsilon is (0, 0.5)]The smaller the value, the smaller the deviation. While
wherein ek,nIs the error signal of node k at time n, i.e. Is the step size parameter, sgn { x } is a sign function, which is defined as:
xk,nis the input signal of node k at time n, ek,nIs the error signal of node k at time n, i.e. Is the step size parameter, pk,nIs calculated by the following iterative formula:the time smoothing coefficient is 0 < eta < 1, epsilon is a small positive number and is used for avoiding the condition that the denominator is zero, and the value range of epsilon is (0, 0.5)]The smaller the value, the smaller the deviation. More importantly, in alphak,nAfter the update is completed, it needs to be limited to [ - α ]+,α+]To do soCan be calculated by the following formula:
(4) adjusting the combination coefficients according to a convex symbol regression metric mechanismAnd
ek,nis the error signal of node k at time n, i.e. Is the step size parameter, sgn { x } is a sign function, which is defined as:
more importantly, in alphak,nAfter the update is completed, it needs to be limited to [ - α ]+,α+]To do soCan be calculated by the following formula:
in addition, when the number of candidate cooperation strategies is greater than 2, the preset combination mechanism is at least one of an affine multi-strategy mechanism and a convex multi-strategy mechanism, that is, the affine multi-strategy mechanism and the convex multi-strategy mechanism can be adopted independently, or the affine multi-strategy mechanism and the convex multi-strategy mechanism can be adopted simultaneously. The method comprises the following specific steps:
WhereinIs a newly introduced auxiliary parameter, and the iterative update expression of the auxiliary parameter is as follows:
is the step size parameter, the estimation error eknCalculated by the following formula:whileDelta is a small positive number to avoid the condition that the denominator is zero, and the value range of delta is (0, 0.5)]The smaller the value, the smaller the deviation.
WhereinIs a newly introduced auxiliary parameter, and the iterative update expression of the auxiliary parameter is as follows:
is the step size parameter, the estimation error ek,nCalculated by the following formula:whileDelta is a small positive number to avoid the condition that the denominator is zero, and the value range of delta is (0, 0.5)]The smaller the value, the smaller the deviation. More importantly, inAfter the update is completed, it needs to be limited to
The above embodiments are described in more detail below using three specific examples:
the first example is as follows:
consider thatThe non-stationary system identification application with time variation, the distributed network is composed of N-10 nodes, and the regression quantity xk,nAnd noise zk,nAll generated by a zero mean Gaussian distribution, weight coefficient vectorThe variation of (c) is divided into four stationary phases and three transient phases. In the stationary phase, the weight coefficient vectorThe method is characterized in that the method is generated by standard Gaussian distribution, a weight coefficient vector in a transient stage is generated by linear interpolation of weight coefficient vectors in adjacent stationary stages, two ATC diffusion LMS strategies are considered as candidate strategies, and combined matrixes of the two ATC diffusion LMS strategies are matrixes corresponding to an identity matrix and an averaging rule respectively. For the affine energy normalization mechanism, set ε 0.05, η 0.95 andfor affine symbolic regression mechanism, setting
The result is shown in fig. 3, fig. 3 is a graph of the simulation result of the affine energy normalization mechanism and the affine symbol regression mechanism when M is 2, it can be seen from fig. 3(a) that the best candidate policy performance can be achieved at different stages, and the change of the combination coefficient of the node 4 and the node 9 in fig. 3(b) further verifies the effectiveness of the distributed network diffusion policy based on the combination mechanism.
The second example:
consider thatThe non-stationary system identification application with time variation, the distributed network is composed of N-10 nodes, and the regression quantity xk,nAnd noise zk,nAll generated by a zero mean Gaussian distribution, weight coefficient vectorThe variation of (c) is divided into five stationary phases and four transient phases. In the stationary phase, the weight coefficient vectorThe method is characterized in that the method is generated by standard Gaussian distribution, a weight coefficient vector in a transient stage is generated by linear interpolation of weight coefficient vectors in adjacent stationary stages, two ATC diffusion LMS strategies are considered as candidate strategies, and combined matrixes of the two ATC diffusion LMS strategies are matrixes corresponding to an identity matrix and an averaging rule respectively. For the convex energy normalization mechanism, set ε 0.05, η 0.95 andfor convex symbol regression mechanism, setting
The result is shown in fig. 4, fig. 4 is a graph of a simulation result of the convex energy normalization mechanism and the convex symbol regression mechanism when M is 2, it can be seen from fig. 4(a) that the best candidate policy performance can be achieved at different stages, and the validity of the distributed network diffusion policy based on the combination mechanism is further verified by the change of the combination coefficient of the node 4 and the node 9 in fig. 4 (b).
The third example:
consider thatThe non-stationary system identification application with time variation, the distributed network is composed of N-10 nodes, and the regression quantity xk,nAnd noise zk,nAll generated by a zero mean Gaussian distribution, weight coefficient vectorThe variation of (c) is divided into four stationary phases and three transient phases. In the stationary phase, the weight coefficient vectorThe method is generated by standard Gaussian distribution, the weight coefficient vector of a transient stage is generated by linear interpolation of the weight coefficient vectors of adjacent stationary stages, three ATC diffusion LMS strategies are considered as candidate strategies, and a candidate strategy S(1)The combined matrix of (2) is an identity matrix and has a smaller network step length; candidate policy S(2)The combined matrix of (a) is a matrix corresponding to the average rule; candidate policy S(3)The combined matrix of (a) is an identity matrix and has a larger network step size, and considering the affine multi-policy mechanism, the result is as shown in fig. 5, fig. 5 is a simulation result diagram of the affine multi-policy mechanism when M > 2, and it can be seen from fig. 5(a) that the best candidate policy performance can be achieved at different stages, and the change of the combination coefficient of the node 9 in fig. 5(b) further verifies the effectiveness of the distributed network diffusion policy based on the combined mechanism.
The fourth example:
consider thatThe non-stationary system identification application with time variation, the distributed network is composed of N-10 nodes, and the regression quantity xk,nAnd noise zk,nAll generated by a zero mean Gaussian distribution, weight coefficient vectorThe variation of (c) is divided into five stationary phases and four transient phases. In the stationary phase, the weight coefficient vectorThe method is generated by standard Gaussian distribution, the weight coefficient vector of a transient stage is generated by linear interpolation of the weight coefficient vectors of adjacent stationary stages, three ATC diffusion LMS strategies are considered as candidate strategies, and a candidate strategy S(1)The combined matrix of (2) is an identity matrix and has a smaller network step length; candidate policy S(2)The combined matrix of (a) is a matrix corresponding to the average rule; candidate policy S(3)The combined matrix of (2) is an identity matrix, and has a larger network step length, and a convex multi-strategy mechanism is considered.
The result is shown in fig. 6, fig. 6 is a simulation result diagram of the M > 2 time-varying multi-policy mechanism, and it can be seen from fig. 6(a) that the best candidate policy performance can be achieved at different stages, and the change of the node 9 combination coefficient in fig. 6(b) further verifies the effectiveness of the distributed network flooding policy based on the combination mechanism.
In summary, the method provided by the present invention adopts multiple mechanisms to adaptively adjust the combination coefficients, which has the advantages that the combination mechanism is utilized to overcome the defect of a single diffusion strategy, and different candidate strategies are selected to enable the combination strategy to achieve the performance of the best strategy in different application situations.
Fig. 7 shows an embodiment of a distributed network cooperation policy optimization apparatus based on a combination mechanism, where fig. 7 is a schematic diagram of an embodiment of a distributed network cooperation policy optimization apparatus based on a combination mechanism, where the apparatus includes:
initial estimation result obtainingAn obtaining unit 701, configured to obtain, at a preset time, an initial estimation result of a preset node under multiple candidate cooperation policies of a distributed network, specifically, M candidate diffusion policies S may be run in parallel in the distributed network(1)、S(2)、……、S(M)The node k obtains a candidate diffusion strategy S at the moment n(i)Initial estimation result ofWherein i is 1,2, …, M;
an initial combination coefficient assigning unit 702, configured to assign a corresponding initial combination coefficient to each initial estimation result, specifically, M candidate diffusion strategies S on the node k at time n(1)、S(2)、……、S(M)Initial estimation result ofAssigning initial combining coefficients, respectively
The adaptive adjustment unit 703 is configured to perform adaptive adjustment on the initial combination coefficient according to a preset combination mechanism to obtain an optimized combination coefficient that optimizes the performance of the cooperation strategy, and specifically, may adaptively adjust the initial combination coefficient based on the preset combination mechanismThe cooperation strategy which leads the tracking performance to be the best corresponds to the optimized combination coefficient;
an optimized estimation result calculating unit 704, configured to obtain an optimized estimation result of the preset node by using the optimized combination coefficient and the candidate cooperation policy, and specifically, may calculate a final output result of the distributed network by using the optimized combination coefficient
Another embodiment of the present invention is the device as described aboveOn the basis of the embodiment of the distributed network cooperation strategy optimization device based on the combination mechanism, the initial combination coefficient allocation unit is further specifically used for allocating a corresponding convex combination coefficient and/or affine combination coefficient to each initial estimation result. That is, only the convex combination coefficient may be used, only the affine combination coefficient may be used, or both of them may be included. Specifically, the affine combination coefficient may satisfy the following condition:and a convex combination coefficient satisfying the following condition:and isWhere i is 1,2, …, M.
Moreover, the initial estimation result obtaining unit is specifically configured to obtain, at a preset time, an initial estimation result of a preset node under multiple different types of cooperation policies with the same parameter and/or multiple same types of cooperation policies with different parameters and/or multiple different types of cooperation policies with different parameters.
A further specific embodiment of the apparatus provided by the present invention is based on the embodiment of the distributed network cooperation policy optimization apparatus based on the combination mechanism, and the adaptive adjustment unit may be specifically configured to, when the number of candidate cooperation policies is 2, adaptively adjust the initial combination coefficient according to an affine energy normalization mechanism and/or an affine symbol regression metric mechanism and/or a convex energy normalization mechanism and/or a convex symbol regression metric mechanism, to obtain an optimized combination coefficient that optimizes the cooperation policy performance. That is, at least one of the four mechanisms can be selected, and the specific four mechanisms are as described above and will not be described herein again.
In addition, the adaptive adjustment unit may be further specifically configured to, when the number of candidate cooperation strategies is greater than 2, perform adaptive adjustment on the initial combination coefficient according to an affine multi-strategy mechanism and/or a convex multi-strategy mechanism, so as to obtain an optimized combination coefficient that optimizes the performance of the cooperation strategies. That is, the affine multi-policy mechanism, the convex multi-policy mechanism, or both of them may be separately adopted, and the two mechanisms are also described in detail above, and therefore are not described herein again.
Fig. 8 shows an embodiment of a computer device provided by the present invention, where fig. 8 is a schematic diagram of an embodiment of a computer device provided by the present invention, and the computer device includes:
a memory 801 for storing a computer program;
a processor 802, configured to execute the computer program to implement the steps of the method for optimizing a distributed network cooperation policy based on a combination mechanism according to any of the above embodiments.
In an embodiment of a computer-readable storage medium, a computer program is stored on the computer-readable storage medium, and when executed by a processor, the computer program implements the steps of the method for optimizing a distributed network cooperation policy based on a combination mechanism as provided in any one of the above embodiments.
The embodiments of the foregoing distributed network cooperation policy optimization apparatus, computer device, and computer-readable storage medium based on a combination mechanism provided by the present invention can overcome the defect of a single cooperation policy, so that the distributed network can achieve the performance of the best cooperation policy in different application situations, and can be applied to, but not limited to, a diffusion policy, an incremental policy, and a consistency policy.
The previous description of the disclosed embodiments is provided to enable any person skilled in the art to make or use the present invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the generic principles defined herein may be applied to other embodiments without departing from the spirit or scope of the invention. Thus, the present invention is not intended to be limited to the embodiments shown herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.
Claims (10)
1. A distributed network cooperation strategy optimization method based on a combination mechanism is characterized by comprising the following steps:
at a preset moment, acquiring an initial estimation result of a preset node under a plurality of candidate cooperation strategies of a distributed network;
assigning a respective initial combining coefficient to each of the initial estimation results;
according to a preset combination mechanism, the initial combination coefficient is subjected to self-adaptive adjustment to obtain an optimized combination coefficient which enables the performance of the cooperation strategy to be optimized;
and obtaining an optimized estimation result of the preset node by using the optimized combination coefficient and the candidate cooperation strategy.
2. The combination-mechanism-based distributed network cooperation strategy optimization method according to claim 1, wherein the initial combination coefficient comprises at least one of a convex combination coefficient and an affine combination coefficient.
3. The method of claim 1, wherein the candidate collaboration policies comprise at least one of a plurality of different types of collaboration policies with the same parameters, a plurality of the same type of collaboration policies with different parameters, and a plurality of different types of collaboration policies with different parameters.
4. The method according to claim 1, wherein when the number of candidate cooperation strategies is 2, the preset combination mechanism is at least one of an affine energy normalization mechanism, an affine symbol regression mechanism, a convex energy normalization mechanism and a convex symbol regression mechanism.
5. The method according to claim 1, wherein when the number of candidate collaboration policies is greater than 2, the predetermined combination mechanism is at least one of an affine multi-policy mechanism and a convex multi-policy mechanism.
6. The method as claimed in claim 4, wherein when the number of the candidate cooperation strategies is 2 and the preset combination mechanism is an affine energy normalization mechanism, the combination coefficient is adjusted according to the affine energy normalization mechanismAnd
wherein xk,nIs the input signal of node k at time n, ek,nIs the error signal of node k at time n, i.e. Is the step size parameter, pk,nIs calculated by the following iterative formula:the time smooth coefficient is more than 0 and less than 1, and the value range of epsilon is (0, 0.5)]. While
When the preset combination mechanism is an affine symbol regression mechanism, the combination coefficient is adjusted according to the affine symbol regression mechanismAnd
wherein ek,nIs the error signal of node k at time n, i.e. Is the step size parameter, sgn { x } is a sign function defined as:
When the preset combination mechanism is a convex energy normalization mechanism, the combination coefficient is adjusted according to the convex energy normalization mechanismAnd
xk,nis the input signal of node k at time n, ek,nIs node kError signals at time n, i.e. Is the step size parameter, pk,nIs calculated by the following iterative formula:the time smooth coefficient is more than 0 and less than 1, and the value range of epsilon is (0, 0.5)]At αk,nAfter the update is completed, it is limited to [ - α ]+,α+]To do soCalculated by the following formula:
when the preset combination mechanism is a convex symbol regression mechanism, the combination coefficient is adjusted according to the convex symbol regression mechanismAnd
ek,nis the error signal of node k at time n, i.e. Is the step size parameter, sgn { x } is a sign function defined as:
7. the method as claimed in claim 5, wherein when the number of candidate cooperation strategies is greater than 2 and the preset combination mechanism is an affine multi-strategy mechanism, the combination coefficients are adjusted according to the affine multi-strategy mechanism
is the step size parameter, the estimation error ek,nCalculated by the following formula:whileDelta value range (0, 0.5)];
When the preset combination mechanism is a convex multi-strategy mechanism, the combination coefficient is adjusted according to the convex multi-strategy mechanism
8. A distributed network cooperation strategy optimization device based on a combination mechanism is characterized by comprising the following components:
the initial estimation result acquisition unit is used for acquiring initial estimation results of a preset node under a plurality of candidate cooperation strategies of the distributed network at a preset moment;
an initial combination coefficient allocation unit, configured to allocate a corresponding initial combination coefficient to each of the initial estimation results;
the self-adaptive adjusting unit is used for carrying out self-adaptive adjustment on the initial combination coefficient according to a preset combination mechanism to obtain an optimized combination coefficient which enables the performance of the cooperation strategy to be optimized;
and the optimized estimation result calculation unit is used for obtaining the optimized estimation result of the preset node by using the optimized combination coefficient and the candidate cooperation strategy.
9. A computer device, comprising:
a memory for storing a computer program;
a processor for implementing the steps of the method for optimizing a distributed network cooperation strategy based on a combined mechanism according to any one of claims 1 to 7 when executing the computer program.
10. A computer-readable storage medium, characterized in that the computer-readable storage medium has stored thereon a computer program, which when executed by a processor implements the steps of the method for distributed network cooperation policy optimization based on combinational mechanism according to any one of claims 1 to 7.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202011530099.8A CN112714165B (en) | 2020-12-22 | 2020-12-22 | Distributed network cooperation strategy optimization method and device based on combination mechanism |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202011530099.8A CN112714165B (en) | 2020-12-22 | 2020-12-22 | Distributed network cooperation strategy optimization method and device based on combination mechanism |
Publications (2)
Publication Number | Publication Date |
---|---|
CN112714165A true CN112714165A (en) | 2021-04-27 |
CN112714165B CN112714165B (en) | 2023-04-04 |
Family
ID=75545219
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202011530099.8A Active CN112714165B (en) | 2020-12-22 | 2020-12-22 | Distributed network cooperation strategy optimization method and device based on combination mechanism |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN112714165B (en) |
Citations (16)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US6285977B1 (en) * | 1993-04-20 | 2001-09-04 | Mitsubishi Denki Kabushiki Kaisha | Multi-agents cooperating method and system |
CN101488188A (en) * | 2008-11-10 | 2009-07-22 | 西安电子科技大学 | SAR image classification method based on SVM classifier of mixed nucleus function |
CN103248693A (en) * | 2013-05-03 | 2013-08-14 | 东南大学 | Large-scale self-adaptive composite service optimization method based on multi-agent reinforced learning |
CN103869698A (en) * | 2012-12-18 | 2014-06-18 | 江南大学 | Sampling control method of multi-intellectual body system consistency |
US20160205697A1 (en) * | 2015-01-14 | 2016-07-14 | Futurewei Technologies, Inc. | Analytics-Assisted, Multi-agents, Self-Learning, Self-managing, Flexible and Adaptive Framework for Intelligent SON |
EP3188373A1 (en) * | 2015-12-30 | 2017-07-05 | Mitsubishi Electric R&D Centre Europe B.V. | Fast distributed clustering for cooperation in telecommunications networks |
CN107045655A (en) * | 2016-12-07 | 2017-08-15 | 三峡大学 | Wolf pack clan strategy process based on the random consistent game of multiple agent and virtual generating clan |
CN109511277A (en) * | 2018-08-01 | 2019-03-22 | 东莞理工学院 | The cooperative method and system of multimode Continuous action space |
CN110261877A (en) * | 2019-06-26 | 2019-09-20 | 南京航空航天大学 | A kind of air-ground coordination vision navigation method and device for scheming optimization SLAM based on improvement |
CN110276698A (en) * | 2019-06-17 | 2019-09-24 | 国网江苏省电力有限公司淮安供电分公司 | Distribution type renewable energy trade decision method based on the study of multiple agent bilayer cooperative reinforcing |
AU2019101446A4 (en) * | 2019-11-23 | 2020-01-23 | Southwest University | A fast distributed strategy for large-scale machine learning |
CN110782011A (en) * | 2019-10-21 | 2020-02-11 | 辽宁石油化工大学 | Networked multi-agent system distributed optimization control method based on reinforcement learning |
CN111161751A (en) * | 2019-12-25 | 2020-05-15 | 声耕智能科技(西安)研究院有限公司 | Distributed microphone pickup system and method under complex scene |
CN111901167A (en) * | 2020-07-24 | 2020-11-06 | 中国科学院重庆绿色智能技术研究院 | Event self-adaptive clustering method based on time-varying multitask network |
CN111934917A (en) * | 2020-07-24 | 2020-11-13 | 重庆邮电大学 | Grouping consistency control method of heterogeneous multi-agent system based on trust node |
CN111988185A (en) * | 2020-08-31 | 2020-11-24 | 重庆邮电大学 | Multi-step communication distributed optimization method based on Barzilai-Borwein step length |
-
2020
- 2020-12-22 CN CN202011530099.8A patent/CN112714165B/en active Active
Patent Citations (16)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US6285977B1 (en) * | 1993-04-20 | 2001-09-04 | Mitsubishi Denki Kabushiki Kaisha | Multi-agents cooperating method and system |
CN101488188A (en) * | 2008-11-10 | 2009-07-22 | 西安电子科技大学 | SAR image classification method based on SVM classifier of mixed nucleus function |
CN103869698A (en) * | 2012-12-18 | 2014-06-18 | 江南大学 | Sampling control method of multi-intellectual body system consistency |
CN103248693A (en) * | 2013-05-03 | 2013-08-14 | 东南大学 | Large-scale self-adaptive composite service optimization method based on multi-agent reinforced learning |
US20160205697A1 (en) * | 2015-01-14 | 2016-07-14 | Futurewei Technologies, Inc. | Analytics-Assisted, Multi-agents, Self-Learning, Self-managing, Flexible and Adaptive Framework for Intelligent SON |
EP3188373A1 (en) * | 2015-12-30 | 2017-07-05 | Mitsubishi Electric R&D Centre Europe B.V. | Fast distributed clustering for cooperation in telecommunications networks |
CN107045655A (en) * | 2016-12-07 | 2017-08-15 | 三峡大学 | Wolf pack clan strategy process based on the random consistent game of multiple agent and virtual generating clan |
CN109511277A (en) * | 2018-08-01 | 2019-03-22 | 东莞理工学院 | The cooperative method and system of multimode Continuous action space |
CN110276698A (en) * | 2019-06-17 | 2019-09-24 | 国网江苏省电力有限公司淮安供电分公司 | Distribution type renewable energy trade decision method based on the study of multiple agent bilayer cooperative reinforcing |
CN110261877A (en) * | 2019-06-26 | 2019-09-20 | 南京航空航天大学 | A kind of air-ground coordination vision navigation method and device for scheming optimization SLAM based on improvement |
CN110782011A (en) * | 2019-10-21 | 2020-02-11 | 辽宁石油化工大学 | Networked multi-agent system distributed optimization control method based on reinforcement learning |
AU2019101446A4 (en) * | 2019-11-23 | 2020-01-23 | Southwest University | A fast distributed strategy for large-scale machine learning |
CN111161751A (en) * | 2019-12-25 | 2020-05-15 | 声耕智能科技(西安)研究院有限公司 | Distributed microphone pickup system and method under complex scene |
CN111901167A (en) * | 2020-07-24 | 2020-11-06 | 中国科学院重庆绿色智能技术研究院 | Event self-adaptive clustering method based on time-varying multitask network |
CN111934917A (en) * | 2020-07-24 | 2020-11-13 | 重庆邮电大学 | Grouping consistency control method of heterogeneous multi-agent system based on trust node |
CN111988185A (en) * | 2020-08-31 | 2020-11-24 | 重庆邮电大学 | Multi-step communication distributed optimization method based on Barzilai-Borwein step length |
Non-Patent Citations (3)
Title |
---|
刘群: "基于策略扩散的群体行为演化机制研究", 《小型微型计算机系统》 * |
席峰等: "基于信息矩阵加权一致策略的分布式Kalman滤波器", 《信息与控制》 * |
王云鹏等: "基于多智能体强化学习的车载自组织网络协作信道分配(英文)", 《FRONTIERS OF INFORMATION TECHNOLOGY & ELECTRONIC ENGINEERING》 * |
Also Published As
Publication number | Publication date |
---|---|
CN112714165B (en) | 2023-04-04 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Zhang et al. | Distributed filtering for switched linear systems with sensor networks in presence of packet dropouts and quantization | |
CN113193556B (en) | Short-term wind power prediction method based on probability prediction model | |
Peng et al. | Input–output data-based output antisynchronization control of multiagent systems using reinforcement learning approach | |
CN113225377B (en) | Internet of things edge task unloading method and device | |
CN112583633B (en) | Distributed optimization method of directed multi-agent network based on rough information | |
WO2021036414A1 (en) | Co-channel interference prediction method for satellite-to-ground downlink under low earth orbit satellite constellation | |
CN113110039B (en) | Finite time distributed aggregation optimization method of multi-agent system | |
CN110278571B (en) | Distributed signal tracking method based on simple prediction-correction link | |
CN115271099A (en) | Self-adaptive personalized federal learning method supporting heterogeneous model | |
Lei et al. | Adaptive multiple non-negative matrix factorization for temporal link prediction in dynamic networks | |
Lu et al. | Diffusion total least-squares algorithm with multi-node feedback | |
CN114818257A (en) | Intelligent micro-grid distributed dynamic tracking method with privacy protection effect | |
CN114708479B (en) | Self-adaptive defense method based on graph structure and characteristics | |
Wen et al. | Variable step-size diffusion normalized sign-error algorithm | |
Wang et al. | Delay-dependent stability analysis for switched stochastic networks with proportional delay | |
CN114326664B (en) | Design method of fault-tolerant controller of nonlinear multi-agent and storage medium | |
Fan et al. | Distributed bias-compensated normalized least-mean squares algorithms with noisy input | |
Chen et al. | Distributed optimization via diffusion adaptation | |
CN111967199B (en) | Agent contribution distribution method under reinforcement learning multi-agent cooperation task | |
CN112714165B (en) | Distributed network cooperation strategy optimization method and device based on combination mechanism | |
CN108205706B (en) | Artificial neural network reverse training device and method | |
Yao et al. | Ensembles of echo state networks for time series prediction | |
Yang et al. | Parallel stochastic decomposition algorithms for multi-agent systems | |
Aduroja et al. | Distributed principal components analysis in sensor networks | |
Van Nguyen et al. | Distributed learning in a multi-agent potential game |
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 |