CN111262739A - Distributed self-adaptive local diffusion control method under event-triggered communication - Google Patents

Distributed self-adaptive local diffusion control method under event-triggered communication Download PDF

Info

Publication number
CN111262739A
CN111262739A CN202010055205.5A CN202010055205A CN111262739A CN 111262739 A CN111262739 A CN 111262739A CN 202010055205 A CN202010055205 A CN 202010055205A CN 111262739 A CN111262739 A CN 111262739A
Authority
CN
China
Prior art keywords
node
time
current moment
determining
adaptive
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
Application number
CN202010055205.5A
Other languages
Chinese (zh)
Other versions
CN111262739B (en
Inventor
陈枫
郭尊湖
邓舒蔚
刘志锋
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Southwest University
Original Assignee
Southwest University
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Southwest University filed Critical Southwest University
Priority to CN202010055205.5A priority Critical patent/CN111262739B/en
Publication of CN111262739A publication Critical patent/CN111262739A/en
Application granted granted Critical
Publication of CN111262739B publication Critical patent/CN111262739B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • 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
    • H04W—WIRELESS COMMUNICATION NETWORKS
    • H04W4/00—Services specially adapted for wireless communication networks; Facilities therefor
    • H04W4/30—Services specially adapted for particular environments, situations or purposes
    • H04W4/38—Services specially adapted for particular environments, situations or purposes for collecting sensor information

Landscapes

  • Engineering & Computer Science (AREA)
  • Computer Networks & Wireless Communication (AREA)
  • Signal Processing (AREA)
  • Data Exchanges In Wide-Area Networks (AREA)

Abstract

The invention discloses a distributed self-adaptive local diffusion control method under event-triggered communication, which comprises the steps of firstly determining the position information of each node after the network topology changes; then determining neighbor nodes of each node according to the position information of each node and calculating the intermediate estimation value of each node at the current moment, if the trigger condition is met, adopting the intermediate estimation value at the current moment, and if the trigger condition is not met, adopting the intermediate estimation value when the trigger condition is met last time as the intermediate estimation value at the current moment; then, the self-adaptive weight coefficient of each node at the current moment is determined, and finally, the output data of each node is determined. The invention considers the influence of the dynamic network topological structure on the local strategy estimation performance, and obtains the system estimation through self-adaptive local diffusion by means of the traditional ATC algorithm. Meanwhile, an event trigger mechanism is designed to determine whether the node needs to send data to the neighbor node at a certain time, so that the communication cost is effectively reduced.

Description

Distributed self-adaptive local diffusion control method under event-triggered communication
Technical Field
The invention relates to a wireless sensor network technology, in particular to a distributed self-adaptive local diffusion control method under event-triggered communication.
Background
In recent years, Wireless Sensor Networks (WSNS) have been widely used in different fields such as power grids and environmental monitoring. In a wireless sensor network, methods for estimating unknown parameters of the network are many, and centralized solution and distributed solution are concerned by researchers due to better estimation performance. In a centralized solution, each node in the network transmits its data to the central node. In this solution, the central node has to undertake many computational tasks. Once the central node is damaged or the energy is exhausted, the whole sensor network can be crashed. Unlike centralized solutions, in distributed solutions, local nodes cooperate with each other to complete the estimation task of the network. Thus, it is more robust than the centralized solution in case of link failure.
In the prior art, distributed solutions include incremental, consensus and diffusion solutions. In the flooding scheme, nodes communicate with each other in a broadcast manner. Thus, the scheme does not rely on any central control mechanism or protocol, which makes the distributed algorithm more robust to communication link impairments or node impairments. Because of these advantages, the flooding scheme has found wide application in wireless sensor networks. The Diffusion least mean-square (DLMS) is an important Diffusion scheme, and provides a simple and effective method for implementing distributed adaptive filtering in a network.
However, in wireless sensor networks, nodes tend to be limited in computational power and power. When a node performs a distributed task in a wireless network, the most power consuming action is data transmission. Therefore, reducing inter-node communication is a necessary condition for ensuring long-term stability of the network, and many algorithms attempt to reduce inter-node communication, such as local flooding or data selection.
However, in the existing data selection scheme based on estimation error, data is divided into meaningless data and error data, and data redundancy is reduced by setting a proper threshold value, so as to achieve the purpose of reducing network communication cost. The invention also provides a local diffusion scheme based on a Least Mean Square (LMS) algorithm, and each node uses a subset of intermediate estimators to carry out parameter estimation, thereby reducing the communication cost of the network. However, the conventional local flooding scheme is to reduce the estimated performance of the algorithm at the cost of reduced traffic. Meanwhile, the influence of the network topology change on the estimation performance and the waste of communication resources caused by data redundancy are not considered. In actual environments, the topology of the network may change at any time (e.g., marine environments). The dynamic network topology usually means that the neighbor nodes of each node change every moment, and the estimation performance of the traditional local flooding strategy is usually affected. This means that conventional algorithms are not suitable for dynamic network topology environments.
Disclosure of Invention
Aiming at the problems existing in the current research, the invention aims to provide a local diffusion method suitable for a dynamic network topology environment, and data redundancy in the dynamic network topology is reduced as much as possible.
In order to achieve the purpose, the invention adopts the following specific technical scheme:
a distributed self-adaptive local diffusion control method under event-triggered communication is characterized by comprising the following steps:
s1: acquiring current-time input data and last-time output data of each node, and determining position information of each node at the current time according to the position information of each node at the last time and the moving distance;
s2: determining neighbor nodes of each node according to the position information of each node at the current moment, and determining an intermediate estimation value of each node at the current moment according to the adaptive weight coefficient of each node at the previous moment, the input data and the output data of each neighbor node;
s3: judging whether a triggering condition is met or not according to the intermediate estimation value of each node at the current moment, if the triggering condition is met, adopting the intermediate estimation value of the current moment, and if the triggering condition is not met, adopting the intermediate estimation value of the last time when the triggering condition is met as the intermediate estimation value of the current moment;
s4: determining the self-adaptive weight coefficient of each node at the current moment according to the intermediate estimation value of each node at the current moment and the intermediate estimation value of each neighbor node at the current moment obtained in the step S3;
s5: and determining the output data of each node according to the input data of each node and the self-adaptive weight coefficient of the current moment.
Optionally, there are N nodes in the whole network, and moving in a two-dimensional plane, the position of the nth time of node k is represented as: (a)k,n,bk,n) K ∈ {1,2, …, N }; then in step S1 the following is followed:
Figure BDA0002372566550000031
determining the position information of each node at the current moment, wherein
Figure BDA0002372566550000032
Representing the lateral travel distance of node k from time n-1 to time n,
Figure BDA0002372566550000033
representing the distance of longitudinal movement of node k from time n-1 to time n.
Optionally, in step S2, it is determined whether the node is a neighboring node through whether direct communication between the two nodes is possible, and an adaptive local diffusion algorithm is adopted when determining the intermediate estimation value of the current time of each node, according to:
Figure BDA0002372566550000034
calculating the intermediate estimated value psi of the node k at the nth timek,n；
Wherein: w is ak,n-1Representing the adaptive weight coefficient of node k at time instant n-1,
Figure BDA0002372566550000035
represents the neighbor node set of node k at time n, dl,nOutput data, x, of the represented i neighbor nodes at time nl,nOf the input data of the represented i neighbor nodes at time n,
Figure BDA0002372566550000036
denotes xl,nRank of (c), coefficient cl,kRepresents the corresponding elements in the N × N non-negative real matrix C, and has:
cl,k＝0，if
Figure BDA0002372566550000037
C1＝1,1TC＝1T1 represents a unit vector of N × 1;
μkthe weight coefficient representing node k compensates the parameter.
Optionally, the triggering mechanism E is set in step S3k,nTo correct the intermediate estimate of node k at time n, wherein:
Ek,n:
Figure BDA0002372566550000038
Figure BDA0002372566550000041
variables of
Figure BDA0002372566550000042
Rho is a positive scalar, and rho is more than 0 and less than 1;
Figure BDA0002372566550000043
representing the intermediate estimate value the last time the trigger condition was satisfied.
Optionally, in step S4 as
Figure BDA0002372566550000044
Determining the adaptive weight coefficient w of the node k at the nth momentk,nWherein the coefficient al,kRepresents the corresponding element in the NxN non-negative real matrix A, and has: a isl,k＝0，if
Figure BDA0002372566550000045
A1＝1,1TA＝1TAnd 1 denotes a unit vector of N × 1.
Optionally, in step S4, the method includes:
Figure BDA0002372566550000046
determining the adaptive weight coefficient w of the node k at the nth momentk,nWherein the coefficient ak,kAnd al,kRepresenting the corresponding element in the NxN non-negative real matrix A and having al,k＝0，if
Figure BDA0002372566550000047
A1＝1,1TA＝1T1 denotes a unit vector of Nx 1, Hl，nIs an L multiplied by L diagonal matrix, the diagonal line of which has M1 and L-M0, L is the number of neighbor nodes of a node k, M is the number of intermediate estimation items selected from the L neighbor nodes, ILAn L × L identity matrix is represented.
Optionally, the value of M is determined according to the current network scale and the number of neighbor nodes.
Optionally, with
Figure BDA0002372566550000048
As a function of local cost, to
Figure BDA0002372566550000049
And as a global cost function, determining intermediate variables involved in the steps S2-S4 through global optimization, and finally obtaining the adaptive weight coefficient of each node at the current moment under the global optimal condition.
Optionally, in step S5 as
Figure BDA00023725665500000410
Determines the output data of the node k at the nth time,
Figure BDA00023725665500000411
rank, w, of input data for node k at time noRepresents the global optimal solution of the network at the nth moment, vk,nNoise at time n of node k.
Optionally, the node is a wireless sensor node.
The invention has the following remarkable effects:
(1) the method considers the influence of a dynamic network topological structure on the local strategy estimation performance, introduces a brand new local cost function, and obtains system estimation through self-adaptive local diffusion by means of a traditional ATC algorithm.
(2) In order to reduce the communication resource waste caused by data redundancy in the self-adaptive local diffusion, the invention designs an event trigger mechanism to determine whether the node needs to send data to the neighbor node at a certain moment, thereby effectively reducing the communication cost.
Drawings
The invention will be further described with reference to the accompanying drawings and examples, in which:
FIG. 1 is a control flow diagram of the present invention;
FIG. 2 is a schematic diagram of dynamic network node location information change;
FIG. 3 is a control diagram of an event trigger mechanism according to the present invention;
FIG. 4 is a waveform of input signals and noise in a simulation experiment;
FIG. 5 is a diagram of a dynamic network topology at different points in time;
FIG. 6 is a graph comparing the impact of the intermediate estimation subset dimension M on estimation performance;
FIG. 7 is a graph comparing the impact of the number of neighbors on a node on the estimated performance;
FIG. 8 is a graph comparing the estimated performance effect of a dynamic network and a non-dynamic network;
fig. 9 is a comparison of the estimated effect of different algorithms.
Detailed Description
In order to make the technical problems, technical solutions and advantages of the present invention more apparent, the following detailed description is given with reference to the accompanying drawings and specific embodiments, it being understood that the specific embodiments described herein are merely illustrative of the present invention and are not intended to limit the present invention.
As shown in fig. 1, a distributed adaptive local flooding control method under event-triggered communication according to the present invention includes the following steps:
s1: acquiring current-time input data and last-time output data of each node, and determining position information of each node at the current time according to the position information of each node at the last time and the moving distance;
the network model usually adopts a wireless sensor network, and when system research is carried out, the system model is usually defined first, and communication between nodes is an undirected graph
Figure BDA0002372566550000061
Comprises a group of nodes N and a group of sets
Figure BDA0002372566550000062
For example, if (i, j) e ξ, this means that node i can exchange information with node j
Figure BDA0002372566550000063
And (4) showing. Each node has an input vector
Figure BDA0002372566550000064
And an output signal
Figure BDA0002372566550000065
They can be linked by the following linear model:
Figure BDA0002372566550000066
wherein v isk,nIs space-independent zero-mean Gaussian noise with variance expressed as
Figure BDA0002372566550000067
n is a time index, T represents the rank of the matrix or vector, woRepresenting the optimal solution for the network.
As shown in fig. 2, for a dynamic network, if there are N nodes in the whole network and move in a two-dimensional plane, the position of the nth time of node k is represented as: (a)k,n,bk,n) K ∈ {1,2, …, N }; then the following can be followed:
Figure BDA0002372566550000068
determining the position information of each node at the current moment, wherein
Figure BDA0002372566550000069
Representing the lateral travel distance of node k from time n-1 to time n,
Figure BDA00023725665500000610
representing the distance of longitudinal movement of node k from time n-1 to time n.
The invention uses a distributed diffusion least mean square strategy, and the optimal estimation w is sought by minimizing a global cost functiono：
Figure BDA0002372566550000071
In distributed estimation, a global optimal estimate is obtained by finding a local optimal estimate. They can be related by the following linear relationship:
Figure BDA0002372566550000072
the local cost function of the network can be expressed as:
Figure BDA0002372566550000073
wherein c isl,kRepresents the corresponding elements in the N × N non-negative real matrix C, and has:
Figure BDA0002372566550000074
1 represents a unit vector of N × 1;
to solve cost function equation (4), as an embodiment, an adaptive-recombination (ATC) diffusion LMS algorithm may be directly utilized. The ATC comprises the following specific processes:
Figure BDA0002372566550000075
Figure BDA0002372566550000076
(recombination)
Coefficient al,kRepresents the corresponding element in the NxN non-negative real matrix A, and has:
Figure BDA0002372566550000077
1 denotes an N × 1 unit vector.
Based on the above analysis, specific implementations of step S2 and step S4 can be obtained, that is:
s2: determining neighbor nodes of each node according to the position information of each node at the current moment, and determining an intermediate estimation value of each node at the current moment according to the adaptive weight coefficient of each node at the previous moment, the input data and the output data of each neighbor node;
in specific implementation, whether the node is a neighbor node is determined through whether direct communication can be performed between two nodes, and a self-adaptive local diffusion algorithm is adopted when the intermediate estimation value of each node at the current moment is determined, according to the following steps:
Figure BDA0002372566550000081
calculating the intermediate estimated value psi of the node k at the nth timek,n；
Wherein: w is ak,n-1Representing the adaptive weight coefficient of node k at time instant n-1,
Figure BDA0002372566550000082
represents the neighbor node set of node k at time n, dl,nOutput data, x, of the represented i neighbor nodes at time nl,nOf the input data of the represented i neighbor nodes at time n,
Figure BDA0002372566550000083
denotes xl,nRank of (c), coefficient cl,kRepresents the corresponding elements in the N × N non-negative real matrix C, and has:
cl,k＝0，if
Figure BDA0002372566550000084
C1＝1,1TC＝1T1 represents a unit vector of N × 1;
μkthe weight coefficient representing node k compensates the parameter.
In order to avoid the waste of communication resources due to data redundancy, the present invention designs an event trigger mechanism, which can determine whether each node sends the current intermediate estimate to its neighbor, i.e. the content defined in step S3, as follows:
s3: judging whether a triggering condition is met or not according to the intermediate estimation value of each node at the current moment, if the triggering condition is met, adopting the intermediate estimation value of the current moment, and if the triggering condition is not met, adopting the intermediate estimation value of the last time when the triggering condition is met as the intermediate estimation value of the current moment;
in specific implementation, the trigger mechanism E is set in step S3k,nTo correct the intermediate estimate of node k at time n, wherein:
Figure BDA0002372566550000085
Figure BDA0002372566550000086
rho is a positive scalar, and rho is more than 0 and less than 1;
Figure BDA0002372566550000087
representing the intermediate estimate value the last time the trigger condition was satisfied.
S4: determining the self-adaptive weight coefficient of each node at the current moment according to the intermediate estimation value of each node at the current moment and the intermediate estimation value of each neighbor node at the current moment obtained in the step S3;
as a second embodiment, in order to adapt to the dynamic change of the network topology, the present invention redesigns an adaptive local diffusion method, that is, by:
Figure BDA0002372566550000091
determining the adaptive weight coefficient w of the node k at the nth momentk,nWherein the coefficient ak,kAnd al,kRepresenting the corresponding element in the NxN non-negative real matrix A and having al,k＝0，if
Figure BDA0002372566550000092
A1＝1,1TA＝1T1 denotes a unit vector of Nx 1, Hl,nIs an L multiplied by L diagonal matrix, the diagonal line of which has M1 and L-M0, L is the number of neighbor nodes of a node k, M is the number of intermediate estimation items selected from the L neighbor nodes, ILThe value of M is determined according to the current network scale and the number of neighbor nodes, and specifically:
Figure BDA0002372566550000093
s5: and determining the output data of each node according to the input data of each node and the self-adaptive weight coefficient of the current moment. Specifically, in step S5, the following steps are performed:
Figure BDA0002372566550000094
determines the output data of the node k at the nth time,
Figure BDA0002372566550000095
rank, w, of input data for node k at time noRepresents the global optimal solution of the network at the nth moment, vk,nNoise at time n of node k.
To further understand the technical effects of the present invention, the present embodiment is defined as ET-PDLMS algorithm, and further describes the analysis of the mean square performance and communication cost performance.
For ease of analysis, we set the following conditions:
one, input vector xk,nIndependent in space-time, the covariance of the input vector is expressed as:
Figure BDA0002372566550000101
two, noise vector vk,nAre independent in space and time. We have E [ v ]k,n]0 and:
Figure BDA0002372566550000102
step size parameter mukSmall enough that its squared value is negligible.
We define the intermediate estimation error vector as:
Figure BDA0002372566550000103
the estimation error of node k is:
Figure BDA0002372566550000104
bringing formula (1) into formula (6) and subtracting w from both sides simultaneouslyoAnd obtaining an intermediate estimation error vector:
Figure BDA0002372566550000105
the estimated value of the k node is rewritten as follows according to equations (5) and (7):
Figure BDA0002372566550000106
substituting the estimated value in equation (18) into woObtaining:
Figure BDA0002372566550000107
Figure BDA0002372566550000108
if the intermediate estimate of node k is psik,nAt time n satisfies
Figure BDA0002372566550000109
According to equation (6), the intermediate estimate for node k is represented as:
Figure BDA00023725665500001010
therefore, we define the superimposed vector of the intermediate estimation error vectors for all nodes as:
Figure BDA0002372566550000111
to facilitate subsequent analysis, we superimpose the intermediate estimation errors on a matrix
Figure BDA0002372566550000112
Expressed as:
Figure BDA0002372566550000113
we define:
M＝blockdiag{μ1IL,…,μkIL,…,μNIL} (24)
Figure BDA0002372566550000114
Figure BDA0002372566550000115
Figure BDA0002372566550000116
for the estimated error superposition matrix of each node, the following is defined:
we can prove that
Figure BDA0002372566550000117
And
Figure BDA0002372566550000118
Figure BDA0002372566550000119
where 1 is an L by L matrix of 0. Bringing equation (23) into equation (27) yields an estimated error vector:
Figure BDA0002372566550000121
with respect to the average performance,
Figure BDA0002372566550000122
Xnand
Figure BDA0002372566550000123
independent of each other, taking into account the expectations of condition one and condition 2 and taking both sides of equation (29) one can obtain:
Figure BDA0002372566550000124
herein, the
Figure BDA0002372566550000125
According to equation (30), if the algorithm is required to be stable in the mean sense, then the matrix is required
Figure BDA0002372566550000126
And (4) stabilizing. All Q rows add up to 1. Therefore, when the matrix is
Figure BDA0002372566550000127
When stable, equation (30) stabilizes. We have:
Figure BDA0002372566550000128
λmax{. denotes the maximum eigenvalue of the matrix. Matrix array
Figure BDA0002372566550000129
Is the matrix IL-μkRkAnd therefore, when equation (31) is satisfied,
Figure BDA00023725665500001210
it can be seen by analysis that if the algorithm is required to be stable in the mean sense, the step size range is:
Figure BDA00023725665500001211
for mean square performance, the squared weighted euclidean norm of vector b and weighting matrix a is:
Figure BDA00023725665500001212
using this Euclidean norm, we can analyze the mean square stability of the ET-APDLMS algorithm. The squared euclidean norm is taken for both sides of equation (29) and the desired operator is applied when considering a1 and a 2. We have:
Figure BDA00023725665500001213
Figure BDA00023725665500001214
where Σ is a random symmetric non-negative definite matrix. Under the precondition of the condition one and the condition two,
Figure BDA00023725665500001215
and Γ are independent of each other. Thus, we can get:
Figure BDA0002372566550000131
by defining γ ═ vec { E [ Γ ] } and δ ═ vec { Σ }, we modify equation (34) to be in accordance with equation (36):
Figure BDA0002372566550000132
vec {. is a vectorization operator that can superimpose the column vectors of the matrix into a column vector, vecT{. is the transpose of the matrix vectorization operation. From the matrix vectorization operation, we represent vec { } as:
Figure BDA0002372566550000133
meanwhile, we can rewrite γ to be: γ ═ Λ δ (39), where:
Figure BDA0002372566550000134
and
Figure BDA0002372566550000135
considering condition three, we can approximate equation (40) as:
Figure BDA0002372566550000136
from the vector operation vec { } versus matrix lanes, we have:
tr{ATB}＝vecT{B}vec{A} (43)
and:
Figure BDA0002372566550000137
Figure BDA0002372566550000138
the formula (39) and the formula (44) are introduced into the formula (37). We can get:
Figure BDA0002372566550000139
when Λ is stable, equation (22) in the mean square sense stabilizes, and Λ can be approximated as:
Figure BDA00023725665500001310
thus, the stability of formula (47) and formula
Figure BDA00023725665500001311
The stability conditions of (3) were the same. At the same time, the step size μ satisfying the formula (32) is selectedkMaking the algorithm stable in the mean square sense.
For the communication cost, the present example analyzes from the adaptation step and the recombination step, respectively.
Self-adaptation: in this step, each node receives { x } from neighboring nodesl,n,dl,n}，xl,nIs a vector of dimension l, dl,nIs a scalar. Suppose that each node in the network has an average
Figure BDA0002372566550000141
And (4) each neighbor node. The amount of data transmitted at this step is NF (L + 1).
Combining: in this step, parameter estimation is performed using the intermediate estimation subset of the neighbor nodes of node k. Assuming that the data used is an M-dimensional vector (the choice of M is related to H), the remaining L-M dimensional data is replaced with intermediate estimates of node k, so the amount of data transmitted in the combining step is NFM.
In particular by setting conditions
Figure BDA0002372566550000142
To determine the number of M's,
Figure BDA0002372566550000143
where th is a threshold designed according to the network size [.]Is taken as the upper limit function.
The performance of the ET-APDLMS algorithm defined in the present invention in the network is further illustrated by simulation experiments. We have designed a wireless sensor network with 50 nodes. Input x for each nodek,nAnd noise vk,nAs shown in fig. 4. In the simulation process, 50 nodes are randomly placed in a 200 × 200 area, and four indexes of the dynamic network topology are shown in fig. 5.
Fig. 6 shows the effect of dimension M of the intermediate estimation subset on the estimation performance. Step size used in the simulation is muk0.05, each node has an average of 4 neighbor nodes. The results of fig. 6 are the average of 50 independent experimental values, indicating that the estimated performance of the algorithm improves as M increases.
Each node has an average of F neighbor nodes, and the step size used in the simulation is mukThe dimension of the intermediate estimation subset is M4, 0.05. Figure 7 is an average of 50 independent experimental values. Simulation results show that the estimation performance of the algorithm is improved with the increase of the number of the adjacent nodes.
In fig. 8, the PDLMS algorithm in the dynamic network topology is compared with the PDLMS algorithm in the non-dynamic network topology. Researches show that the estimation performance of the PDLMS algorithm is influenced by the dynamic network topology structure.
FIG. 9 compares the estimated performance of the ET-APDLMS algorithm and the PDLMS algorithm. Step size used in the simulation is muk0.05, and 25 as th. Simulation results show that the ET-APDLMS algorithm can adapt to dynamic network topological structure better than the PDLMS algorithm. The estimated performance of the ET-APDLMS algorithm is slightly lower than that of the DLMS algorithm, but previous analysis shows that the communication cost of the ET-APDLMS algorithm is significantly reduced.
In summary, the distributed adaptive local diffusion method under event-triggered communication provided by the present invention fully considers the influence of the dynamic network topology structure on the local policy estimation performance, introduces a brand new local cost function, obtains system estimation through adaptive local diffusion by means of the traditional ATC algorithm, and designs an event triggering mechanism to determine whether a node needs to send data to a neighboring node at a certain time in order to reduce the communication resource waste caused by data redundancy in the adaptive local diffusion, thereby effectively reducing the communication cost.
Finally, it is noted that, in this document, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising an … …" does not exclude the presence of other like elements in a process, method, article, or apparatus that comprises the element.
The above sequence numbers of the embodiments of the present invention are merely for description, and do not represent the advantages and disadvantages of the embodiments, and it is clear to those skilled in the art from the above description of the embodiments that the method of the embodiments can be implemented by software plus a necessary general hardware platform, and certainly can also be implemented by hardware, but the former is a better implementation manner in many cases. Based on such understanding, the technical solutions of the present invention may be embodied in the form of a software product, which is stored in a storage medium (such as ROM/RAM, magnetic disk, optical disk) and includes instructions for enabling a terminal (such as a mobile phone, a computer, a server, an air conditioner, or a network device) to execute the method according to the embodiments of the present invention.
While the present invention has been described with reference to the embodiments shown in the drawings, the present invention is not limited to the embodiments, which are illustrative and not restrictive, and it will be apparent to those skilled in the art that various changes and modifications can be made therein without departing from the spirit and scope of the invention as defined in the appended claims.

Claims (10)

1. A distributed self-adaptive local diffusion control method under event trigger communication is characterized by comprising the following steps:
s1: acquiring current-time input data and last-time output data of each node, and determining position information of each node at the current time according to the position information of each node at the last time and the moving distance;
s2: determining neighbor nodes of each node according to the position information of each node at the current moment, and determining an intermediate estimation value of each node at the current moment according to the adaptive weight coefficient of each node at the previous moment, the input data and the output data of each neighbor node;
s3: judging whether a triggering condition is met or not according to the intermediate estimation value of each node at the current moment, if the triggering condition is met, adopting the intermediate estimation value of the current moment, and if the triggering condition is not met, adopting the intermediate estimation value of the last time when the triggering condition is met as the intermediate estimation value of the current moment;
s4: determining the self-adaptive weight coefficient of each node at the current moment according to the intermediate estimation value of each node at the current moment and the intermediate estimation value of each neighbor node at the current moment obtained in the step S3;
s5: and determining the output data of each node according to the input data of each node and the self-adaptive weight coefficient of the current moment.
2. The method of distributed adaptive local flooding control under event-triggered communication according to claim 1, characterized in that: there are N nodes in the whole network, and move in the two-dimensional plane, and the position of the nth time of the node k is expressed as: (a)k,n,bk,n) K ∈ {1,2, …, N }; then in step S1 the following is followed:
Figure FDA0002372566540000011
determining the position information of each node at the current moment, wherein
Figure FDA0002372566540000012
Representing the lateral travel distance of node k from time n-1 to time n,
Figure FDA0002372566540000013
representing the distance of longitudinal movement of node k from time n-1 to time n.
3. The method of distributed adaptive local flooding control under event-triggered communication according to claim 1, characterized in that: in step S2, whether the node is a neighbor node is determined by whether direct communication between two nodes is possible, and an adaptive local diffusion algorithm is used when determining the intermediate estimate value of each node at the current time, according to the following:
Figure FDA0002372566540000021
calculating the intermediate estimated value psi of the node k at the nth timek,n；
Wherein: w is ak,n-1Represents the adaptive weight coefficient, N, of the node k at the time instant N-1kRepresents the neighbor node set of node k at time n, dl,nOutput data, x, of the represented i neighbor nodes at time nl,nOf the input data of the represented i neighbor nodes at time n,
Figure FDA0002372566540000022
denotes xl,nRank of (c), coefficient cl,kRepresents the corresponding elements in the N × N non-negative real matrix C, and has:
cl,k＝0，
Figure FDA0002372566540000023
C1＝1,1TC＝1T1 represents a unit vector of N × 1;
μkthe weight coefficient representing node k compensates the parameter.
4. The distributed adaptive local flooding control method under event-triggered communication according to claim 3, characterized in that, in step S3, trigger mechanism E is setk,nTo correct the intermediate estimate of node k at time n, wherein:
Figure FDA0002372566540000024
Figure FDA0002372566540000025
variables of
Figure FDA0002372566540000026
Rho is a positive scalar, and rho is more than 0 and less than 1;
Figure FDA0002372566540000027
representing the intermediate estimate value the last time the trigger condition was satisfied.
5. The method for distributed adaptive local flooding control under event-triggered communication according to claim 4, wherein step S4 is performed in accordance with
Figure FDA0002372566540000028
Determining the adaptive weight coefficient w of the node k at the nth momentk,nWherein the coefficient al,kRepresents the corresponding element in the NxN non-negative real matrix A, and has: a isl,k＝0，
Figure FDA0002372566540000029
A1＝1,1TA＝1TAnd 1 denotes a unit vector of N × 1.
6. The method for distributed adaptive local flooding control under event-triggered communication according to claim 4, wherein in step S4, according to:
Figure FDA0002372566540000031
determining the adaptive weight coefficient w of the node k at the nth momentk,nWherein the coefficient ak,kAnd al,kRepresenting the corresponding element in the NxN non-negative real matrix A and having al,k＝0，
Figure FDA0002372566540000032
A1＝1,1TA＝1T1 denotes a unit vector of Nx 1, Hl,nIs an L multiplied by L diagonal matrix, the diagonal line of which has M1 and L-M0, L is the number of neighbor nodes of a node k, M is the number of intermediate estimation items selected from the L neighbor nodes, ILAn L × L identity matrix is represented.
7. The method according to claim 6, wherein the value of M is determined according to the current network size and the number of neighbor nodes.
8. The method for distributed adaptive local flooding control under event-triggered communication according to claim 6 or 7, characterized in that
Figure FDA0002372566540000033
As a function of local cost, to
Figure FDA0002372566540000034
And as a global cost function, determining intermediate variables involved in the steps S2-S4 through global optimization, and finally obtaining the adaptive weight coefficient of each node at the current moment under the global optimal condition.
9. The method for distributed adaptive local flooding control under event-triggered communication according to claim 8, wherein step S5 is performed in accordance with
Figure FDA0002372566540000035
Determines the output data of the node k at the nth time,
Figure FDA0002372566540000036
rank, w, of input data for node k at time noRepresents the global optimal solution of the network at the nth moment, vk,nNoise at time n of node k.
10. The method of distributed adaptive local flooding control under event-triggered communication of claim 1, wherein the nodes are wireless sensor nodes.
CN202010055205.5A 2020-01-17 2020-01-17 Distributed self-adaptive local diffusion control method under event-triggered communication Active CN111262739B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202010055205.5A CN111262739B (en) 2020-01-17 2020-01-17 Distributed self-adaptive local diffusion control method under event-triggered communication

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202010055205.5A CN111262739B (en) 2020-01-17 2020-01-17 Distributed self-adaptive local diffusion control method under event-triggered communication

Publications (2)

Publication Number Publication Date
CN111262739A true CN111262739A (en) 2020-06-09
CN111262739B CN111262739B (en) 2021-04-06

Family

ID=70950680

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202010055205.5A Active CN111262739B (en) 2020-01-17 2020-01-17 Distributed self-adaptive local diffusion control method under event-triggered communication

Country Status (1)

Country Link
CN (1) CN111262739B (en)

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103152751A (en) * 2013-02-07 2013-06-12 西安邮电大学 Energy-saving transmission adaptive LMS (Least-Mean Squares) distributed detection method for wireless sensor network
CN103945514A (en) * 2014-04-22 2014-07-23 西安交通大学 Diffusion type self-adaption uplink power control method based on fairness
EP2930531B1 (en) * 2014-04-11 2016-06-15 Thales Secondary surveillance radar with high refresh rate
CN110121184A (en) * 2019-04-26 2019-08-13 电子科技大学 A kind of distributed online adaptive diffusion combination coefficient optimization method of nonnegativity restrictions
CN110266363A (en) * 2019-06-26 2019-09-20 电子科技大学 A kind of distributed diffusion self-adapting anti-jamming method based on tensor

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103152751A (en) * 2013-02-07 2013-06-12 西安邮电大学 Energy-saving transmission adaptive LMS (Least-Mean Squares) distributed detection method for wireless sensor network
EP2930531B1 (en) * 2014-04-11 2016-06-15 Thales Secondary surveillance radar with high refresh rate
CN103945514A (en) * 2014-04-22 2014-07-23 西安交通大学 Diffusion type self-adaption uplink power control method based on fairness
CN110121184A (en) * 2019-04-26 2019-08-13 电子科技大学 A kind of distributed online adaptive diffusion combination coefficient optimization method of nonnegativity restrictions
CN110266363A (en) * 2019-06-26 2019-09-20 电子科技大学 A kind of distributed diffusion self-adapting anti-jamming method based on tensor

Non-Patent Citations (4)

* Cited by examiner, † Cited by third party
Title
JAE-WOO LEE等: "Data-Reserved Periodic Diffusion LMS With Low Communication Cost Over Networks", 《IEEE ACCESS》 *
LIMEI HU等: "An Improved Diffusion Affine Projection Estimation Algorithm for Wireless Sensor Networks", 《CIRCUITS, SYSTEMS, AND SIGNAL PROCESSING (2020) 》 *
MARC J. PIGGOTT等: "Stability of distributed adaptive algorithms II: Diffusion algorithms", 《 2015 54TH IEEE CONFERENCE ON DECISION AND CONTROL (CDC)》 *
WEI HUANG等: "Communication-reducing diffusion LMS algorithm over multitask networks", 《INFORMATION SCIENCES》 *

Also Published As

Publication number Publication date
CN111262739B (en) 2021-04-06

Similar Documents

Publication Publication Date Title
CN111176327B (en) Multi-agent system enclosure control method and system
Ren et al. Multi-agent Kalman consensus with relative uncertainty
Xu et al. Anti-attack event-triggered control for nonlinear multi-agent systems with input quantization
Wen et al. Event-based distributed recursive filtering for state-saturated systems with redundant channels
CN110196554B (en) Safety consistency control method of multi-agent system
CN112327633A (en) Method for leadership following multi-agent system consistency with time lag and disturbance
CN110601777B (en) Method for estimating satellite-ground downlink co-channel interference under low-orbit mobile satellite constellation
Tian et al. An adaptive fusion strategy for distributed information estimation over cooperative multi-agent networks
WO2010045971A1 (en) Method and system for the deployment of nodes of a wireless communications network
Cavalcante et al. An adaptive projected subgradient approach to learning in diffusion networks
CN111781822B (en) Privacy protection grouping consistency control method of multi-agent system
CN112788699B (en) Method and system for determining network topology of self-organizing network
Ke et al. Distributed adaptive fault-tolerant consensus of nonlinear multi-agent systems via state-constraint impulsive protocols with time-delay
Tan et al. Output feedback leader-following consensus for nonlinear stochastic multiagent systems: The event-triggered method
Griparic et al. Consensus-based distributed connectivity control in multi-agent systems
CN110225539B (en) Energy-saving method based on distributed estimation of wireless sensor network
CN111262739B (en) Distributed self-adaptive local diffusion control method under event-triggered communication
Wang et al. Dynamic event-triggered boundary control for exponential consensus of multi-agent systems of impulsive PDEs with switching topology
Yang et al. Distributed estimation of graph spectrum
Dimarogonas L 2 gain stability analysis of event-triggered agreement protocols
Wang et al. Distributed two‐stage state estimation with event‐triggered strategy for multirate sensor networks
Wang et al. Knowledge distillation based cooperative reinforcement learning for connectivity preservation in uav networks
Zhu et al. On hybrid consensus-based extended Kalman filtering with random link failures over sensor networks
Hwang et al. Adaptive reinforcement learning in box-pushing robots
Yang et al. Neural-Network-Based Set-Membership Filtering Under WTOD Protocols via a Novel Event-Triggered Compensation 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