CN108966310A - Cluster head based on space compression optimizes election algorithm - Google Patents
Cluster head based on space compression optimizes election algorithm Download PDFInfo
- Publication number
- CN108966310A CN108966310A CN201810649463.9A CN201810649463A CN108966310A CN 108966310 A CN108966310 A CN 108966310A CN 201810649463 A CN201810649463 A CN 201810649463A CN 108966310 A CN108966310 A CN 108966310A
- Authority
- CN
- China
- Prior art keywords
- node
- cluster
- data
- particle
- sensing
- 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.)
- Pending
Links
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04W—WIRELESS COMMUNICATION NETWORKS
- H04W40/00—Communication routing or communication path finding
- H04W40/24—Connectivity information management, e.g. connectivity discovery or connectivity update
- H04W40/32—Connectivity information management, e.g. connectivity discovery or connectivity update for defining a routing cluster membership
Landscapes
- Engineering & Computer Science (AREA)
- Computer Networks & Wireless Communication (AREA)
- Signal Processing (AREA)
- Mobile Radio Communication Systems (AREA)
Abstract
The invention discloses a kind of, and the cluster head based on space compression optimizes election algorithm, N number of sensing node is randomly dispersed in a square region, sink node is located at the center of whole network, sensor network enters circular flow, circulation carries out the foundation of cluster first every time, using the optimum position of cluster head election algorithm search leader cluster node, sensing node selection is apart from nearest leader cluster node voluntarily cluster；Then carry out data transmission, sensing node data are transmitted to leader cluster node, space compression data are obtained based on space compression principle compressed data, and space compression data are transmitted to sink node；Data reconstruction is finally carried out, sink node carries out real-time reconstruction to data.The present invention realizes that the cluster head distribution in network is relatively uniform and further extends network lifecycle, solves leader cluster node and elects the technical issues of influencing network reliability at random.
Description
Technical field
The invention belongs to wireless sensor network fields, and in particular to a kind of cluster head optimization election calculation based on space compression
Method.
Background technique
Wireless sensor network plays increasingly important role, at present in a variety of intelligent Services such as monitoring, management
Wireless sensor network has been applied to multiple fields, including military, industry, environmental monitoring and medical services etc..Benefit
With compressed sensing (Compressed Sensing, CS) technology mining wireless sensor network transmission signal space correlation.And
The application premise calls acquired original signal of CS technology has certain correlation in certain domains, and limit CS technology applies model
It encloses.
For election of cluster head mode, due to traditional research contents be from reduce volume of transmitted data angle,
And have ignored the structure problem of network itself.From the angle of network structure, traditional election of cluster head scheme is using the method recycled
Averaging network energy consumption, but the selection of cluster head is random selection, wherein the probability that each sensing node becomes cluster head is all the same.?
After completing certain cycle-index, the more sensing node of the dump energy sensing node less with dump energy still has phase
Same probability is elected as cluster head.If the less node of dump energy is elected as cluster head, the death of the node can be accelerated, while
The reliability for influencing network, shortens network lifecycle.Moreover, traditional election of cluster head scheme not can guarantee cluster head yet
Position, the scale of cluster is uncontrolled.
In order to effectively reduce the energy consumption of network, realizes the effective control being distributed to cluster, extend the Life Cycle of network
The optimisation technique of compressed sensing technology and election of cluster head mode is combined together by phase, this programme consideration.
Summary of the invention
It is an object of the invention to reduce the energy consumption of network, realizes the effective control being distributed to cluster, propose a kind of base
Optimize election algorithm in the cluster head of space compression, optimizes wireless sensor network by introducing two kinds of constrained parameters of energy and distance
Election of cluster head method, and go deep into excavate perception data spatial coherence, realize network in cluster head distribution it is relatively uniform with
And further extend network lifecycle, it solves leader cluster node and elects the technical issues of influencing network reliability at random.
The present invention adopts the following technical scheme that a kind of cluster head optimization election algorithm based on space compression saves N number of perception
Point is randomly dispersed in a square region, and convergence (sink) node is located at the center of whole network, sensor network
Into circular flow, each circular flow the following steps are included:
1) foundation of cluster: optimizing cluster head election algorithm using particle swarm algorithm, is searched for by the cluster head election algorithm of optimization
The optimum position of leader cluster node, sensing node selection is apart from nearest leader cluster node voluntarily cluster；
2) data are transmitted: sensing node data being transmitted to leader cluster node, are obtained based on space compression principle compressed data
Space compression data, and space compression data are transmitted to sink node；
3) data reconstruction: sink node carries out real-time reconstruction to data.
It preferably, will include sensing node dump energy and location information after sensing node determines position in the step 1)
Data packet LMIt is sent to sink node, sink node obtains ENERGY E consumed by sensing node transmission dataCMAnd sensing node
The distance between sink node di-sink, the cluster head election algorithm search leader cluster node of optimization is executed in sink node most
Best placement, and the optimum position of leader cluster node is broadcasted to whole network.
Preferably, the optimum position specific steps of leader cluster node are searched for are as follows:
11) S particle is initialized, each particle includes K random leader cluster nodes；
12) i-th of sensing node is calculated at a distance from leader cluster node k in particle pAnd by i-th of sensing node
It is assigned to apart from nearest leader cluster node；
13) calculating target function value cos t and adaptive value, and be iterated；
Cos t=β1f1+β2f2+β3f3
Wherein, f1Represent sensing node adaptive value at a distance from leader cluster node in cluster；f2The energy for representing leader cluster node adapts to
Value；f3Represent leader cluster node adaptive value at a distance from sink node；β1, β2, β3For each weight coefficient for adapting to value function, β1+β2+
β3=1；CP, kIndicate cluster C in particle pk, | CP, k| indicate cluster C in particle pkThe sensing node quantity possessed, CHP, kIndicate particle
Leader cluster node k, n in piIndicate i-th of sensing node, E (ni) indicate i-th of sensing node dump energy, E (CHP, k) table
Showing the dump energy of the leader cluster node k in particle p, N indicates the quantity of sensing node,Indicate cluster head section in particle p
At a distance between point k and sink node；
14) the individual optimal cluster head group position P of each particle is determinedj=(pj1, pj2..., pjd) and global optimum's cluster head
Group position Po=(po1, po2... pod), update each particle rapidity vjdWith position ljd,
vjd=α vjd+c1r1(pjd-ljd)+c2r2(pod-ljd)
ljd=ljd+vjd
Wherein, the position vector of j-th of particle is expressed as lj=(lj1, lj2... ljd), ljdIt is j-th in d dimension space
The position vector of son, the individual optimal cluster head group position of j-th of particle are the optimum position P that j-th of particle lives throughj=
(pj1, pj2..., pjd), pjdFor the optimum position that j-th of particle in d dimension space lives through, global optimum's cluster head group position is
The optimum position P lived through in all particleso=(po1, po2... pod), podTo be lived through most in all particles in d dimension space
Best placement, the flying speed of j-th of particle are Vj=(vj1, vj2..., vjd), j=1,2 ... S, vjdIt is j-th in d dimension space
The flying speed of particle, α are inertia weight, and r1 and r2 are the random numbers between 0~1, and c1 and c2 are accelerator coefficient；
If particle present position does not have leader cluster node, particle is moved to nearest leader cluster node, the i.e. position of particle
There are location components in vector not in node location, then sets nearest for location components corresponding in the position vector of particle
Node location；
15) step 12) is repeated to step 14), until maximum number of iterations, calculates global optimum's cluster head group position, i.e.,
K optimal leader cluster node position.
Preferably, the dump energy of all sensing nodes is based on single order wireless communication model, specifically:
Send the energy consumption E of dataTAre as follows:
Receive the energy consumption E of dataRAre as follows:
ER=(ERX+EDA)×L
Wherein, L is data packet length, ETXThe energy consumption of every bit, E when to transmit dataRXEvery ratio when to receive data
Special energy consumption, ERXValue and ETXValue it is equal, EfsFor the energy consumption of every bit under free space model, EmpFor multipath
The energy consumption of every bit, E under modelfsWith EmpIt is fixed constant, EDAFor the energy consumption of data fusion, dnodeIndicate two
Distance between a sensing node, dthresholdFor distance threshold.
Preferably, the step 2) specific steps are as follows:
21) cluster Ck∈ { 1,2 ..., K } includes n sensing node, and each sensing node acquires t number in t time slot
According to cluster CkThe perception data x that middle sensing node i is acquired in t time slotK, iAnd it is transmitted to leader cluster node k, leader cluster node k is received
The perception data x that sensing node is sent in all clustersk,
xK, i=[xK, i1 xK, i2 xK, i3 ... xK, it]T
Wherein, xK, i1xK, i2xK, i3And xK, itRespectively indicate cluster CkInterior sensing node i time slot 1, time slot 2, time slot 3 and
The perception data of time slot t acquisition；
xk=[xK, 1 xK, 2 xK, 3 ... xK, n]
Wherein, xK, 1 xK, 2 xK, 3And xK, nRespectively indicate cluster CkInterior sensing node 1, sensing node 2, sensing node 3 with
And the perception data that sensing node n is acquired in t time slot；
x′K, j=[xK, 1j xK, 2j xK, 3j ... xK, nj]T
Wherein, x 'K, j, j ∈ { 1,2 ..., t } expression is in cluster CkThe perception that interior all sensing nodes are acquired in same time slot j
Data, xK, 1j xK, 2j xK, 3jAnd xK, njRespectively indicate cluster CkInterior sensing node 1, sensing node 2, sensing node 3 and perception section
The perception data that point n is acquired in time slot j；
22) using m × n rank observing matrix Φ to matrix x 'kCarry out airspace compression, matrix x 'kFor perception data xkTransposition
Matrix, matrix x 'kIt is compressed to the matrix y of m × tk∈Rm×t, ykFor space compression data；
23) leader cluster node k is by space compression data ykIt is transmitted to sink node.
Preferably, using m × n rank observing matrix Φ to matrix x 'kCarry out airspace compression specifically,
yk=Φ x 'k
=Φ [x 'K, 1 x′K, 2 x′K, 3 ... x′K, t]
=Φ Ψ [θK, 1 θK, 2 θK, 3 ... θK, t]
=Φ Ψ θk=Θ θk
=[yK, 1 yK, 2 yK, 3 … yK, t]
Wherein, Φ is the Gaussian matrix of independent zero-mean, Ψ ∈ Rn×nFor to set matrix, x 'K, j=Ψ θK, j；θK, j∈RnFor
Coefficient column vector, yK, 1 yK, 2 yK, 3And yK, tRespectively indicate cluster CkInterior all sensing nodes the moment 1, the moment 2, the moment 3 and
The compressed data of moment t.
Preferably, data reconstruction is carried out using compression sampling matching pursuit algorithm in the step 3)
Invent it is achieved the utility model has the advantages that the present invention propose it is a kind of based on space compression cluster head optimization election algorithm, it is excellent
Change wireless sensor network cluster head electoral machinery, and go deep into excavating the spatial coherence of perception data, realizes the cluster in network
Head distribution is relatively uniform and further extends network lifecycle, solves leader cluster node and elects influence network reliability at random
Technical problem；The present invention determines optimal cluster head position by defining multiple adaptation value functions, has efficiently controlled the scale of cluster
And keep the distribution of cluster head more reasonable；The stronger advantage of unrestricted and computing capability using sink node energy, in sink node
It executes cluster head election algorithm and broadcasts final result to whole network；It is dynamically selected based on energy and distance model optimization algorithm
Cluster head makes network in circular flow each time, can obtain the smallest energy consumption；It is sufficiently excavated with compressive sensing theory
The spatial coherence of perception data, and space compression is realized at cluster head and sensing node, reach reduction redundant data transmissions
Purpose；Under identical network environment, other scheme methods are compared, there is lower energy consumption and life cycle advantage, have
Effect extends Network morals.
Detailed description of the invention
Fig. 1 is the floor map of the cluster head optimization election algorithm network model of the invention based on space compression；
Fig. 2 is the stereoscopic schematic diagram of the cluster head optimization election algorithm network model of the invention based on space compression；
Fig. 3 is in election of cluster head mode and two kinds of traditional election of cluster head modes of the invention about residue of network organization gross energy
The comparison schematic diagram changed over time；
Fig. 4 be in election of cluster head mode of the invention and two kinds of traditional election of cluster head modes surviving node quantity with circulation
The comparison schematic diagram of number variation；
Fig. 5 be election of cluster head mode of the invention in two kinds of traditional election of cluster head modes in life cycle compared with show
It is intended to；
Fig. 6 is flow chart of the invention.
Specific embodiment
Below according to attached drawing and technical solution of the present invention is further elaborated in conjunction with the embodiments.
The present invention adopts the following technical scheme that, a kind of cluster head based on space compression optimizes election algorithm, sensing network packet
Bottom sensing layer, middle layer and convergence layer three-decker are included, includes N number of sensing node, K cluster, each cluster in the sensing layer of bottom
Separately include 1 leader cluster node and I sensing node；It include at least one intermediate node in middle layer, it will be in the sensing layer of bottom
K cluster be divided at least one cluster group, the quantity of cluster group and the number of intermediate node are equal, and convergence layer includes sink section
Point, as shown in Figure 1, 2, the cluster head based on space compression optimize election algorithm, and sensor network enters circular flow, recycle every time
Specific step is as follows for operation, as shown in Figure 6:
1) foundation of cluster: optimizing cluster head election algorithm using particle swarm algorithm, is searched for by the cluster head election algorithm of optimization
The optimum position of leader cluster node, sensing node selection is apart from nearest leader cluster node voluntarily cluster；
Sensing node determines the data packet L comprising sensing node dump energy and location information behind positionMIt is sent to
Sink node, sink node obtain ENERGY E consumed by sensing node transmission dataCMBetween sensing node and sink node
Distance di-sink, the stronger advantage of and computing capability unrestricted using sink node energy executes the cluster of optimization in sink node
The optimum position of head election algorithm search leader cluster node, and the optimum position of leader cluster node is broadcasted to whole network.
Search for the optimum position specific steps of leader cluster node are as follows:
11) S particle is initialized, each particle includes K random leader cluster nodes；
12) i-th of sensing node is calculated at a distance from leader cluster node k in particle pAnd by i-th of sensing node
It is assigned to apart from nearest leader cluster node；
13) calculating target function value cos t and adaptive value, and be iterated；
Cos t=β1f1+β2f2+β3f3
Wherein, f1Represent sensing node adaptive value at a distance from leader cluster node in cluster；f2The energy for representing leader cluster node adapts to
Value；f3Represent leader cluster node adaptive value at a distance from sink node；β1, β2, β3For each weight coefficient for adapting to value function, β1+β2+
β3=1；CP, kIndicate cluster C in particle pk, | CP, k| indicate cluster C in particle pkThe sensing node quantity possessed, CHP, kIndicate particle
Leader cluster node k, n in piIndicate i-th of sensing node, E (ni) indicate i-th of sensing node dump energy, E (CHP, k) table
Showing the dump energy of the leader cluster node k in particle p, N indicates the quantity of sensing node,Indicate cluster head section in particle p
At a distance between point k and sink node；
Optimal cluster head position is determined by defining multiple adaptation value functions, has been efficiently controlled the scale of cluster and has been made cluster head
Distribution it is more reasonable；Cluster head is dynamically selected based on energy and distance model optimization algorithm, makes network in circulation fortune each time
When row, the smallest energy consumption can be obtained；
14) the individual optimal cluster head group position P of each particle is determinedj=(pj1, pj2..., pjd) and global optimum's cluster head
Group position Po=(po1, po2... pod), update each particle rapidity vjdWith position ljd,
vjd=α vjd+c1r1(pjd-ljd)+c2r2(pod-ljd)
ljd=ljd+vjd
Wherein, the position vector of j-th of particle is expressed as lj=(lj1, lj2... ljd), ljdIt is j-th in d dimension space
The position vector of son, the individual optimal cluster head group position of j-th of particle are the optimum position P that j-th of particle lives throughj=
(pj1, pj2..., pjd), pjdFor the optimum position that j-th of particle in d dimension space lives through, global optimum's cluster head group position is
The optimum position P lived through in all particleso=(po1, po2... pod), podTo be lived through most in all particles in d dimension space
Best placement, the flying speed of j-th of particle are Vj=(vj1, vj2..., vjd), j=1,2 ... S, vjdIt is j-th in d dimension space
The flying speed of particle, α are inertia weight, and r1 and r2 are the random numbers between 0~1, and c1 and c2 are accelerator coefficient, because being flat
On face, so be two-dimensional space, i.e. d=2；
If particle present position does not have leader cluster node, particle is moved to nearest leader cluster node, the i.e. position of particle
There are location components in vector not in node location, then sets nearest for location components corresponding in the position vector of particle
Node location；
15) step 12) is repeated to step 14), until maximum number of iterations, calculates global optimum's cluster head group position, i.e.,
K optimal leader cluster node position.
The dump energy of all sensing nodes is based on single order wireless communication model, specifically:
Send the energy consumption E of dataTAre as follows:
Receive the energy consumption E of dataRAre as follows:
ER=(ERX+EDA)×L
Wherein, L is data packet length, ETXThe energy consumption of every bit, E when to transmit dataRXEvery ratio when to receive data
Special energy consumption, ERXValue and ETXValue it is equal, EfsFor the energy consumption of every bit under free space model, EmpFor multipath
The energy consumption of every bit, E under modelfsWith EmpIt is fixed constant, EDAFor the energy consumption of data fusion, dnodeIndicate two
Distance between a node, dthresholdFor distance threshold.
2) data are transmitted: sensing node data being transmitted to leader cluster node, are obtained based on space compression principle compressed data
Space compression data, and space compression data are transmitted to sink node；
21) cluster Ck∈ { 1,2 ..., K } includes n sensing node, and each sensing node acquires t number in t time slot
According to cluster CkThe perception data x that middle sensing node i is acquired in t time slotK, iAnd it is transmitted to leader cluster node k, leader cluster node k is received
The perception data x that sensing node is sent in all clustersk,
xK, i=[xK, i1 xK, i2 xK, i3 ... xk,it]T
Wherein, xK, i1 xK, i 2xK, i3And xK, itRespectively indicate cluster CkInterior sensing node i time slot 1, time slot 2, time slot 3 with
And the perception data of time slot t acquisition；
Wherein, xK, 1 xK, 2 xK, 3And xK, nRespectively indicate cluster CkInterior sensing node 1, sensing node 2, sensing node 3 with
And the perception data of sensing node n acquisition；
x′K, j=[xK, 1j xK, 2j xK, 3j ... xK, nj]T
Wherein, x 'K, j, j ∈ { 1,2 ..., t } expression is in cluster CkThe perception that interior all sensing nodes are acquired in same time slot j
Data, xK, 1j xK, 2j xK, 3jAnd xK, njRespectively indicate cluster CkInterior sensing node 1, sensing node 2, sensing node 3 and perception section
The perception data that point n is acquired in time slot j；
22) using m × n rank observing matrix Φ to matrix x 'kCarry out airspace compression, matrix x 'kFor perception data xkTransposition
Matrix, matrix x 'kIt is compressed to the matrix y of m × tk∈Rm×t, ykFor space compression data；
Using m × n rank observing matrix Φ to matrix x 'kCarry out airspace compression specifically,
yk=Φ x 'k
=Φ [x 'K, 1 x′K, 2 x′K, 3 ... x′K, t]
=Φ Ψ [θK, 1 θK, 2 θK, 3 ... θK, t]
=Φ Ψ θk=Θ θk
=[yK, 1 yK, 2 yK, 3 ... yK, t]
Wherein, Φ is the Gaussian matrix of independent zero-mean, Ψ ∈ Rn×nFor to set matrix, x 'K, j=Ψ θK, j；θK, j∈RnFor
Coefficient column vector, yK, 1 yK, 2 yK, 3And yK, tRespectively indicate cluster CkInterior all sensing nodes the moment 1, the moment 2, the moment 3 and
The compressed data of moment t.
23) leader cluster node k is by space compression data ykIt is transmitted to sink node；
The spatial coherence of perception data is sufficiently excavated with compressive sensing theory, and is realized at cluster head and sensing node
Space compression achievees the purpose that reduce redundant data transmissions.
3) data reconstruction: sink node carries out real-time reconstruction to data using compression sampling matching pursuit algorithm.
In simulated environment, 100 sensing nodes are randomly dispersed in the region of 100m*100m.The position of sink node
Positioned at (50,50), all sensing node primary powers are 0.05J, ETXFor 50nJ/bit, Efs=10pJ/bit/m2, Emp=
0.0013pJ/bit/m2, EDA=5nJ/bit, the data package size that sensing node is sent are 8000bit, the setting of overall compression rate
It is 0.5.Wireless sensor network circular flow number is 2000.
The application of cluster head optimization election algorithm in the actual process based on space compression, figure are designed based on aforementioned present invention
3 to Fig. 5 utilize three residue of network organization energy, surviving node number and life cycle performance indicators, with traditional based on LEACH agreement
The election of cluster head mode (LEACH) of (lowenergyadaptiveclustering hierarchy) and the sky based on LEACH
Between compression election of cluster head mode (Data spatial compression scheme based on LEACH, SCL) comparison, this
The cluster head based on space compression of invention optimizes election algorithm (spatial compression based selection
Algorithm for cluster heads, SCSA) there is apparent advantage in performance.
As shown in figure 3, LEACH scheme interior joint dump energy curve is most precipitous, i.e., each sensing node is flat in the program
Equal energy consumption is maximum, and SCL scheme is taken second place, and the average energy consumption of each sensing node is minimum in the present invention.Therefore with its other party
Case is compared, and present invention obtains the smallest energy consumption speed, to extend Network morals.
Fig. 4 illustrates each scheme surviving node quantity with the increased situation of change of cycle-index, after circulation 400 times,
Sensing node amount of survival of the invention is significantly more than LEACH and SCL two schemes, this also means that the present invention can be transmitted more
More information content.To say objectively, the present invention can obtain apparent performance advantage.
In order to further show performance advantage of the present invention in terms of life cycle, Fig. 5, which is illustrated, can reflect network life
Three performance indicators in period: the time (First node death, FND) of first node death, half node death
Time (Half node death, HND), the time (Eighty percent node death, END) of 80% node death,
As shown in Figure 5, node life cycle of the invention is considerably longer than LEACH and SCL two schemes.In emulation experiment, English spy is used
The truthful data that your Berkeley laboratory is collected carries out the emulation of reconstruction error, and the average reconstruction error of the method for the present invention is
2.70*10-4, it was demonstrated that design method reconstruction precision with higher of the present invention can satisfy scheme for high reconstruction precision completely
Demand.
In conclusion the above method can reasonably reduce the energy consumption of sensing node in sensor network, guaranteeing height
Effectively extend Network morals under the premise of reconstruction precision.
The above is only a preferred embodiment of the present invention, it is noted that for the ordinary skill people of the art
For member, without departing from the technical principles of the invention, several improvement and deformations can also be made, these improvement and deformations
Also it should be regarded as protection scope of the present invention.
Claims (7)
1. the cluster head based on space compression optimizes election algorithm, which is characterized in that N number of sensing node is being randomly dispersed in one just
In rectangular region, sink node is located at the center of whole network, and sensor network enters circular flow, each circular flow
The following steps are included:
1) foundation of cluster: optimizing cluster head election algorithm using particle swarm algorithm, searches for cluster head by the cluster head election algorithm of optimization
The optimum position of node, sensing node selection is apart from nearest leader cluster node voluntarily cluster；
2) data are transmitted: sensing node data being transmitted to leader cluster node, obtain space based on space compression principle compressed data
Compressed data, and space compression data are transmitted to sink node；
3) data reconstruction: sink node carries out real-time reconstruction to data.
2. the abnormal user detection method of the deep neural network according to claim 1 based on minimum risk, feature
It is, sensing node determines the data packet L comprising sensing node dump energy and location information behind position in the step 1)M
It is sent to sink node, sink node obtains ENERGY E consumed by sensing node transmission dataCMWith sensing node and sink node
The distance between di-sink, the optimum position of the cluster head election algorithm search leader cluster node of optimization is executed in sink node, and will
It is broadcasted to whole network the optimum position of leader cluster node.
3. the abnormal user detection method of the deep neural network according to claim 1 or 2 based on minimum risk, special
Sign is, searches for the optimum position specific steps of leader cluster node are as follows:
11) S particle is initialized, each particle includes K random leader cluster nodes；
12) i-th of sensing node is calculated at a distance from leader cluster node k in particle pAnd i-th of sensing node is distributed
To apart from nearest leader cluster node；
13) calculating target function value cost and adaptive value, and be iterated；
Cost=β1f1+β2f2+β3f3
Wherein, f1Represent sensing node adaptive value at a distance from leader cluster node in cluster；f2Represent the energy adaptive value of leader cluster node；
f3Represent leader cluster node adaptive value at a distance from sink node；β1, β2, β3For each weight coefficient for adapting to value function, β1+β2+β3
=1；Cp,kIndicate cluster C in particle pk, | Cp,k| indicate cluster C in particle pkThe sensing node quantity possessed, CHp,kIndicate particle p
In leader cluster node k, niIndicate i-th of sensing node, E (ni) indicate i-th of sensing node dump energy, E (CHp,k) table
Showing the dump energy of the leader cluster node k in particle p, N indicates the quantity of sensing node,Indicate cluster head section in particle p
At a distance between point k and sink node；
14) the individual optimal cluster head group position P of each particle is determinedj=(pj1,pj2,…,pjd) and global optimum's cluster head group position
Set Po=(po1,po2,…pod), update each particle rapidity vjdWith position ljd,
vjd=α vjd+c1r1(pjd-ljd)+c2r2(pod-ljd)
ljd=ljd+vjd
Wherein, the position vector of j-th of particle is expressed as lj=(lj1,lj2,…ljd), ljdFor j-th particle in d dimension space
Position vector, the individual optimal cluster head group position of j-th of particle are the optimum position P that j-th of particle lives throughj=(pj1,
pj2,…,pjd), pjdFor the optimum position that j-th of particle in d dimension space lives through, global optimum's cluster head group position is all
The optimum position P lived through in particleo=(po1,po2,…pod), podFor the optimum bit lived through in all particles in d dimension space
It sets, the flying speed of j-th of particle is Vj=(vj1,vj2,…,vjd), j=1,2 ... S, vjdFor j-th of particle in d dimension space
Flying speed, α is inertia weight, and r1 and r2 are the random numbers between 0~1, and c1 and c2 are accelerator coefficient；
If particle present position does not have leader cluster node, particle is moved to nearest leader cluster node, the i.e. position vector of particle
In there are location components not in node location, then set nearest node for location components corresponding in the position vector of particle
Position；
15) it repeats step 12) and calculates global optimum's cluster head group position up to maximum number of iterations to step 14), i.e., it is optimal
K leader cluster node position.
4. the abnormal user detection method of the deep neural network according to claim 3 based on minimum risk, feature
It is, the dump energy of all sensing nodes is based on single order wireless communication model, specifically:
Send the energy consumption E of dataTAre as follows:
Receive the energy consumption E of dataRAre as follows:
ER=(ERX+EDA)×L
Wherein, L is data packet length, ETXThe energy consumption of every bit, E when to transmit dataRXEvery bit when to receive data
Energy consumption, ERXValue and ETXValue it is equal, EfsFor the energy consumption of every bit under free space model, EmpFor multipath model
Under every bit energy consumption, EfsWith EmpIt is fixed constant, EDAFor the energy consumption of data fusion, dnodeIndicate two senses
Know the distance between node, dthresholdFor distance threshold.
5. the abnormal user detection method of the deep neural network according to claim 1 based on minimum risk, feature
It is, the step 2) specific steps are as follows:
21) cluster Ck∈ { 1,2 ..., K } includes n sensing node, and each sensing node acquires t data, cluster in t time slot
CkThe perception data x that middle sensing node i is acquired in t time slotk,iAnd it is transmitted to leader cluster node k, leader cluster node k receives all
The perception data x that sensing node is sent in clusterk,
xk,i=[xk,i1 xk,i2 xk,i3 ... xk,it]T
Wherein, xk,i1xk,i2xk,i3And xk,itRespectively indicate cluster CkInterior sensing node i is in time slot 1, time slot 2, time slot 3 and time slot
The perception data of t acquisition；
xk=[xk,1 xk,2 xk,3 ... xk,n]
Wherein, xk,1xk,2xk,3And xk,nRespectively indicate cluster CkInterior sensing node 1, sensing node 2, sensing node 3 and perception section
The perception data that point n is acquired in t time slot；
x′k,j=[xk,1j xk,2j xk,3j...xk,nj]T
Wherein, x 'k,j, j ∈ 1,2 ..., and t } it indicates in cluster CkThe perception data that interior all sensing nodes are acquired in same time slot j,
xk,1j xk,2j xk,3jAnd xk,njRespectively indicate cluster CkInterior sensing node 1, sensing node 2, sensing node 3 and sensing node n exist
The perception data of time slot j acquisition；
22) using m × n rank observing matrix Φ to matrix x 'kCarry out airspace compression, matrix x 'kFor perception data xkTransposition square
Battle array, matrix x 'kIt is compressed to the matrix y of m × tk∈Rm×t, ykFor space compression data；
23) leader cluster node k is by space compression data ykIt is transmitted to sink node.
6. the abnormal user detection method of the deep neural network according to claim 5 based on minimum risk, feature
It is, using m × n rank observing matrix Φ to matrix x 'kCarry out airspace compression specifically,
yk=Φ x 'k
=Φ [x 'k,1x′k,2x′k,3...x′k,t]
=Φ Ψ [θk,1 θk,2 θk,3 ... θk,t]
=Φ Ψ θk=Θ θk
=[yk,1 yk,2 yk,3 ... yk,t]
Wherein, Φ is the Gaussian matrix of independent zero-mean, Ψ ∈ Rn×nFor to set matrix, x 'k,j=Ψ θk,j；θk,j∈RnFor coefficient
Column vector, yk,1yk,2yk,3And yk,tRespectively indicate cluster CkInterior all sensing nodes are in time slot 1, time slot 2, time slot 3 and time slot t
Space compression data.
7. the abnormal user detection method of the deep neural network according to claim 1 based on minimum risk, feature
It is, data reconstruction is carried out using compression sampling matching pursuit algorithm in the step 3).
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810649463.9A CN108966310A (en) | 2018-06-22 | 2018-06-22 | Cluster head based on space compression optimizes election algorithm |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810649463.9A CN108966310A (en) | 2018-06-22 | 2018-06-22 | Cluster head based on space compression optimizes election algorithm |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN108966310A true CN108966310A (en) | 2018-12-07 |
Family
ID=64491870
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810649463.9A Pending CN108966310A (en) | 2018-06-22 | 2018-06-22 | Cluster head based on space compression optimizes election algorithm |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108966310A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110493809A (en) * | 2019-08-16 | 2019-11-22 | 惠州Tcl移动通信有限公司 | Mobile terminal and its communication data method for detecting abnormality, computer-readable medium |
| CN111542011A (en) * | 2020-04-27 | 2020-08-14 | 中山大学 | Layered wireless sensor network clustering routing method based on particle swarm optimization |
| CN112486053A (en) * | 2020-10-20 | 2021-03-12 | 国佳云为(常州)信息科技有限公司 | Industrial park environment monitoring and management system based on block chain |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101286917A (en) * | 2008-04-25 | 2008-10-15 | 南昌大学 | Heterogeneous wireless sensor network life prolonging method based on dormancy mechanism |
| US20100040281A1 (en) * | 2008-08-12 | 2010-02-18 | Halliburton Energy Services, Inc. | Systems and Methods Employing Cooperative Optimization-Based Dimensionality Reduction |
| KR101072448B1 (en) * | 2010-04-27 | 2011-10-11 | 강원대학교산학협력단 | Wireless sensor network system and clustering method thereof |
| CN103200598A (en) * | 2013-04-25 | 2013-07-10 | 青岛农业大学 | Wireless sensor network clustering method based on particle swarm optimization |
| CN103209436A (en) * | 2013-01-28 | 2013-07-17 | 南开大学 | Multi-parameter information fusion sparse model based on compressive sensing theory |
| CN105464704A (en) * | 2015-12-28 | 2016-04-06 | 安徽理工大学 | Coal mine safety monitoring system with wireless sensor information fusion function |
-
2018
- 2018-06-22 CN CN201810649463.9A patent/CN108966310A/en active Pending
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101286917A (en) * | 2008-04-25 | 2008-10-15 | 南昌大学 | Heterogeneous wireless sensor network life prolonging method based on dormancy mechanism |
| US20100040281A1 (en) * | 2008-08-12 | 2010-02-18 | Halliburton Energy Services, Inc. | Systems and Methods Employing Cooperative Optimization-Based Dimensionality Reduction |
| KR101072448B1 (en) * | 2010-04-27 | 2011-10-11 | 강원대학교산학협력단 | Wireless sensor network system and clustering method thereof |
| CN103209436A (en) * | 2013-01-28 | 2013-07-17 | 南开大学 | Multi-parameter information fusion sparse model based on compressive sensing theory |
| CN103200598A (en) * | 2013-04-25 | 2013-07-10 | 青岛农业大学 | Wireless sensor network clustering method based on particle swarm optimization |
| CN105464704A (en) * | 2015-12-28 | 2016-04-06 | 安徽理工大学 | Coal mine safety monitoring system with wireless sensor information fusion function |
Non-Patent Citations (2)
| Title |
|---|
| JIASHENG ZHOU: "spatial compression scheme for improving the lifetime of wireless sensor networks", 《IEEE INTERNATIONAL CONFERENCE ONCOMMUNICATION TECHNOLOGY》 * |
| 陈霞: "基于能量均衡的WSN分簇多路径路由算法研究", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110493809A (en) * | 2019-08-16 | 2019-11-22 | 惠州Tcl移动通信有限公司 | Mobile terminal and its communication data method for detecting abnormality, computer-readable medium |
| CN110493809B (en) * | 2019-08-16 | 2023-07-18 | 惠州Tcl移动通信有限公司 | Mobile terminal, communication data anomaly detection method thereof and computer readable medium |
| CN111542011A (en) * | 2020-04-27 | 2020-08-14 | 中山大学 | Layered wireless sensor network clustering routing method based on particle swarm optimization |
| CN111542011B (en) * | 2020-04-27 | 2021-09-24 | 中山大学 | Layered wireless sensor network clustering routing method based on particle swarm optimization |
| CN112486053A (en) * | 2020-10-20 | 2021-03-12 | 国佳云为(常州)信息科技有限公司 | Industrial park environment monitoring and management system based on block chain |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Liu et al. | Average AoI minimization in UAV-assisted data collection with RF wireless power transfer: A deep reinforcement learning scheme | |
| CN108966310A (en) | Cluster head based on space compression optimizes election algorithm | |
| Xia et al. | Federated-learning-based client scheduling for low-latency wireless communications | |
| CN111722643A (en) | Unmanned aerial vehicle cluster dynamic task allocation method imitating wolf colony cooperative hunting mechanism | |
| CN108616302A (en) | Unmanned aerial vehicle multiple coverage model under power control and deployment method | |
| CN106304191B (en) | A kind of data receiver method and device based on cluster structured radio sensor network | |
| CN103052130B (en) | A kind of wireless multimedia sensor network data fusion method based on rough set | |
| CN106341842A (en) | Method and device for transmitting data of wireless sensor network | |
| CN102333372A (en) | Real-time positioning method and system based on radio frequency fingerprints | |
| CN114690799A (en) | Air-space-ground integrated unmanned aerial vehicle Internet of things data acquisition method based on information age | |
| CN103024857A (en) | Clustering control method applied to wireless sensor networks | |
| CN105426964B (en) | A kind of satellite imagery and the Joint Task planing method that transmits | |
| CN104902565A (en) | Distributed wireless sensor network three-dimensional multi-dimensional scaling (MDS) location method | |
| Tuba et al. | Energy efficient sink placement in wireless sensor networks by brain storm optimization algorithm | |
| CN103428806B (en) | Joint relay selection in a kind of reliable collaboration communication and Poewr control method | |
| Zhang et al. | Fetching ecosystem monitoring data in extreme areas via a drone-enabled internet of remote things | |
| Yunjie | Wireless sensor monitoring system of Canadian Poplar Forests based on Internet of Things | |
| CN114513805A (en) | Wireless modeling method and system | |
| CN115545106A (en) | AoI sensitive data collection method and system in multiple unmanned aerial vehicles | |
| Zhao et al. | Adaptive multi-UAV trajectory planning leveraging digital twin technology for urban IIoT applications | |
| CN102802166A (en) | Improved Zigbee network lamination method | |
| Zhou et al. | Task-Oriented Wireless Communications for Collaborative Perception in Intelligent Unmanned Systems | |
| CN115119174A (en) | Unmanned aerial vehicle autonomous deployment method based on energy consumption optimization in irrigation area scene | |
| CN102892174A (en) | Balanced energy-saving routing method based on smart antenna and dynamic virtual cluster | |
| CN102413178A (en) | Safe automatic clustering method of wireless sensor network nodes |
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 | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20181207 |
|
| RJ01 | Rejection of invention patent application after publication |