CN103200598A - Wireless sensor network clustering method based on particle swarm optimization - Google Patents

Abstract

The invention relates to a wireless sensor network clustering method based on particle swarm optimization and belongs to the field of topology control of wireless sensor networks. The wireless sensor network clustering method is characterized in that the number of optimal clusters is dynamically decided according to the current state of a farmland environment monitoring network, factors affecting energy consumption are taken into full consideration, a fitness function is constructed, and optimal cluster heads are selected by using a particle swarm optimization algorithm, so that the cluster heads are uniformly distributed in a network, and the load of the network is equalized; during clustering, the factors affecting the energy consumption are taken into full consideration, and the optimization is carried out by using the particle swarm optimization algorithm, so that the utilization ratio of energy is increased, and the life cycle of the farmland environment monitoring network is prolonged; and the residual energy of nodes is taken into consideration during cluster head selection, so that cluster head nodes have enough energy to execute tasks, such as collection and aggregation, and the security of data is guaranteed.

Description

A kind of network clustering method of wireless sensor based on particle group optimizing

Technical field

The invention belongs to wireless sensor network topology control field, be specifically related to a kind of network clustering method of wireless sensor based on particle group optimizing.The method of this patent design is passed through to optimize sub-clustering, reaches to reduce resource-constrained sensor network energy consumption, prolongs the purpose of network lifetime and raising system extension.

Background technology

Wireless senser has caused increasing concern as the common equipment of perception image data, is widely used in fields such as military affairs, Industry Control, environment measurings, especially at the farm environment detection range significant application value is arranged.The common volume of wireless sensor node is small, adopts the powered battery of finite energy, and the life span that need as far as possible prolong wireless senser is gathered more farm environment and detected data.Wireless sensor network topology control is very big to the performance impact of network, and good topological structure can improve the efficient of MAC agreement and Routing Protocol, for data fusion and target localization etc. provides basis, the life span that is conducive to prolong whole network.

The wireless communication module of sensor node also consumes more energy under idle condition, and free time most of the time, so the communication module of closure sensor node, expense can significantly cut down the consumption of energy.Consider to select some node to open communication module as leader cluster node, close the communication module of non-leader cluster node, set up collection, route and the forwarding that a connected network is responsible for data by leader cluster node, so both guaranteed the data communication in the coverage, also saved lot of energy.Under this Topology Management mechanism, the node in the network can be divided into leader cluster node and ordinary node.Leader cluster node to bunch in ordinary node manage, coordinate bunch interior nodes, be responsible for fusion and the forwarding of data, energy consumption is bigger, so the sub-clustering algorithm is the periodic energy consumption of selecting leader cluster node to come sensor node in the equalizing network usually, reach the prolongation network lifetime, avoid occurring too early the purpose in energy " cavity ".Level type structure has many good qualities, and is responsible for data fusion such as leader cluster node, has reduced data traffic; The cluster structured large scale deployment that is conducive to distributed algorithm has improved the extensibility of system; The long-time communication close module of most of node, the life span that has prolonged network significantly.

People such as Wendi B.Heinzelman have proposed LEACH (Low Energy Adaptive Clustering Hierarchy), it is periodic that LEACH carries out, node is elected bunch head in turn, makes sensor node energy consumption relative equilibrium, thus the life span that has effectively prolonged network.But node is just according to the elected bunch head of certain probability among the LEACH, the neighbor node of not considering dump energy, the node of node distribute and and the base station between factors such as distance, the selection randomness of bunch head is big, the skewness of bunch head, so the good energy consumption of balanced whole network not, the capacity usage ratio of whole network is lower.

Proposed a lot of improvement on the basis of LEACH, such as HEED (Hybrid Energy-Efficient Distributed clustering), improved at LEACH, but with bunch in average Danone as the standard of communication in weighing bunch.The factors such as distance of bunch head and base station are considered in the optimization of energy consumption in but this is only emphasized bunch, and a bunch head distribution neither be too even, causes energy consumption higher relatively.

Summary of the invention

The present invention proposes a kind of network clustering method of wireless sensor based on particle group optimizing, in the farm environment monitoring was used, wireless senser adopted powered battery, and the life span that need as far as possible prolong transducer is gathered more farmland Monitoring Data.The method is carried out sub-clustering by particle group optimizing, solved wireless sensor network cluster head skewness, thereby energy consumes the unbalanced problems of bringing such as network lifetime weak point, thereby effectively utilize energy, improve autgmentability and the stability of network, prolong the life cycle of network.

For achieving the above object, technical scheme provided by the invention, a kind of network clustering method of wireless sensor based on particle group optimizing, comprise bunch head select, bunch formation and transfer of data three phases, specific as follows:

1) bunch head is selected: at first detect network conditions according to current farm environment, dynamically determine optimum bunch number K; Consider that then the factor such as the neighbor node that influence energy consumption distribute and quantity, residue energy of node and and the base station between distance etc., through serious analysis and experimental summary repeatedly, structure adaptive value function f, use particle group optimizing to select optimal solution, the base station is broadcast to wireless sensor network with the leader cluster node collection of optimal particle correspondence.Before bunch head was selected first, node need be broadcasted the state information of oneself, such as information such as position, dump energies; In running subsequently, take the mode of " incidentally " to carry node status information, significantly reduce energy consumption.Wherein optimum bunch number K adopts formula (1), and the adaptive value function adopts formula (2).

2) bunch formation: after each node was received the broadcast message of base station, the node that becomes bunch head was by CSMA sign indicating number broadcasting self information, after non-leader cluster node is received this broadcast message, sent to bunch hair of peak signal according to the power that receives signal to join request; After bunch head is received adding information, distribute a tdma slot to give this node, this node transmits data in this tdma slot.

3) transfer of data: bunch head is collected the perception information that bunch interior nodes is gathered, and converges the compression back and adopts single order not have line model to be sent to the base station; The state information of node simultaneously certainly is sent to the base station such as the mode of information such as position, dump energy by " incidentally ".Because the application scenarios of this invention is the base station is positioned at the perception regional center and the perception zone is little, so adopt the single-hop mode to communicate simplified model.

Advantage of the present invention:

1) detects network condition according to current farm environment, dynamically determine optimum bunch quantity; Optimum bunch quantity is dynamically to change along with the operation of network, is not unalterable, and we remain an optimum bunch quantity, have reduced energy consumption, have improved the performance of farm environment detection system.

2) particle swarm optimization algorithm is incorporated in the sub-clustering optimization; Particle group optimizing is a kind of bionical optimization algorithm, has characteristics such as simple, efficient, multidimensional, and is simultaneously easy to implement, optimizes performance height and fast convergence rate.We take full advantage of the good characteristic of particle swarm optimization algorithm and handle the sub-clustering optimization problem, select optimum bunch head.

3) take into full account the factor that influences energy consumption, structure adaptive value function; The present invention considers to influence the various factors of energy consumption, as information of neighbor nodes, residue energy of node and with distance of base station etc.Bunch head collection of adaptive optimal control value particle correspondence is the optimal selection when the prevariety head, this selection taken into full account the position of neighbor node and quantity, node dump energy and with the factors such as distance of base station, energy consumption and transmission energy consumption in having optimized bunch, the leader cluster node dump energy also is optimized simultaneously.

4) the dynamic adjustment of particle cluster algorithm parameter; In conjunction with the sub-clustering optimization problem, we dynamically adjust the parameter in the particle cluster algorithm; Simultaneously the weight of each component of adaptive value function is dynamically adjusted, such as in the starting stage, because node energy is more, in emphasis is optimized bunch with the transmission energy consumption, along with the operation of network, residue energy of node is fewer and feweri, and the proportion that this moment, dump energy accounted for increases relatively.

Description of drawings

Fig. 1 is the wireless sensor network illustraton of model

Fig. 2 is the particle swarm optimization algorithm operational flow diagram

Fig. 3 be in the network survival node number and running time graph of a relation

Fig. 4 be in the network dump energy and running time graph of a relation

Embodiment

We detect the wireless sensor network model in conjunction with farm environment, as shown in Figure 1, concrete enforcement of the present invention are elaborated.

In the farm environment perception surveyed area of 100m * 100m, 100 sensor nodes of random arrangement, the base station is positioned at (50,50) m place, and concrete model is:

(1) base station is positioned at farm environment surveyed area center

(2) sensor node random distribution and energy constraint

(3) sensor node has control function of power, can be according to the distance sending power consumption

(4) each sensor node is known the positional information of oneself

(5) node static or with respect to the base station motion extremely slow

(6) sensor node has three classes, respectively collecting temperature, humidity and CO ₂Concentration

1, bunch head is selected

The present invention at first calculates optimum bunch quantity according to the state information of current farm environment detection wireless sensor network, uses particle swarm optimization algorithm to select optimum bunch head then.Detailed process is as follows:

1) calculates optimum bunch quantity

$K=\sqrt{\frac{N}{2\mathrm{\π}}}\frac{M}{d_{\mathrm{toBS}}}---\left(1\right)$

Wherein N is the node sum, and M is the perception region area, d _ToBSBe the average distance between farm environment surveyed area and base station;

2) structure adaptive value function

f＝αf ₁+βf ₂+γf ₃ (2)

${\mathrm{<figure-callout\; id="1"\; label="f"\; filenames="HSA00000884373400012.png"\; state="\{\{state\}\}">f</figure-callout>}}_1=\underset{k=1}{\overset{K}{\mathrm{\&Sigma;}}}\underset{{\&ForAll;n}_i\&Element;C_{p,k}}{\mathrm{\&Sigma;}}d(n_i,{\mathrm{CH}}_{p,k})/\left|C_{p,k}\right|---\left(3\right)$

$f_2=\underset{i=1}{\overset{N}{\mathrm{\&Sigma;}}}E\left(n_i\right)/\underset{k=1}{\overset{K}{\mathrm{\&Sigma;}}}E\left({\mathrm{CH}}_{p,k}\right)---\left(4\right)$

$f_3=\frac{1}{K}\underset{k=1}{\overset{K}{\mathrm{\&Sigma;}}}d({\mathrm{CH}}_{p,k},\mathrm{BS})---\left(5\right)$

α+β+γ＝1，0≤α，β，γ≤1

Wherein K is candidate's bunch number, n _iBe sensor node i, CH _{P, k}Be a bunch CH who comprises among the particle p _k, | C _{P, k}| be to belong to C among the particle p _kBunch the quantity of node, d is Euclidean distance between the two, E is dump energy, α, beta, gamma are the weight of each component, α is changed to 0.1, β by 0.8, and to be changed to 0.8, γ by 0.1 be 0.1.

3) operation particle swarm optimization algorithm, as shown in Figure 2, detailed process is:

Step 1: an initialization m particle, each particle comprise K bunch header, and the structure of each particle is: p={x ₁, y ₁x ₂, y ₂x ₃, y ₃... x _K, y _K, namely each particle is the 2K dimension;

X wherein _l, y _lBe first candidate bunch header that comprises among the particle p, x _K, y _KBe K the candidate's bunch header that comprises among the particle p.

Step 2: the adaptive value of calculating each particle

For each node n _i, calculate this node and all a candidates bunch CH _{P, k}Between apart from d (n _i, CH _{P, k})

If $d(n_i,{\mathrm{CH}}_{p,k})=\underset{\&ForAll;k=\mathrm{1,2},...,K}{\min }\left\{d(n_i,{\mathrm{CH}}_{p,k})\right\},$ Then with node n _iJoin a candidate bunch CH _{P, k}Middle using formula (2) calculates the adaptive value of each particle

Step 3: note the individual adaptive optimal control value of particle and global optimum's adaptive value

Step 4: particle position and speed are upgraded in using formula (6) and (7)

$v_{\mathrm{id}}^{k+1}={\mathrm{\&omega;v}}_{\mathrm{id}}^k+{\mathrm{<figure-callout\; id="1"\; label="c"\; filenames="HSA00000884373400012.png"\; state="\{\{state\}\}">c</figure-callout>}}_1\mathrm{\&xi;}(p_{\mathrm{id}}^k-x_{\mathrm{id}}^k)+c_2\mathrm{\&eta;}(p_{\mathrm{gd}}^k-x_{\mathrm{id}}^k)---\left(6\right)$

$x_{\mathrm{id}}^{k+1}=x_{\mathrm{id}}^k+v_{\mathrm{id}}^{k+1}---\left(7\right)$

X wherein _IdThe position of representing the d dimension component of i particle, v _IdThe velocity amplitude of representing the d dimension component of i particle, p _iBe the particle of individual adaptive optimal control value correspondence, p _gBe the particle of global optimum's adaptive value correspondence, ω is inertia weight, c ₁With c ₂Be the study factor, ξ and η are [0,1] interval interior pseudo random numbers.We get m=20, x _Max=v _Max=100, ω is 0.9～0.4, c ₁Be 2.5～0.5, c ₂Be 0.5～2.5.

Step 5: restriction particle's velocity and position are mapped to each candidate bunch head in the new particle on the nearest node

Step 6: repeating step 2 to step 5 until maximum iteration time.

4) the leader cluster node collection of adaptive optimal control value particle correspondence is current optimum bunch head collection, and the base station is broadcast to farm environment with this information and detects wireless sensor network

2, bunch formation

After receiving bunch header of base station broadcast, the node that becomes bunch head adopts CSMA sign indicating number broadcasting self information, after non-leader cluster node is received this broadcast message, send to bunch hair of peak signal according to the power that receives signal to join request; After bunch head is received adding information, distribute a tdma slot to give this node, this node transmits data in this tdma slot.

3, transfer of data

Bunch head is collected bunch interior nodes and is gathered a farm environment information, converges to merge the back and adopt single order not have line model to be sent to the base station; When collecting the data of bunch interior nodes collection, information such as node current state can be adopted mode one block movement of " incidentally ", the energy consumption of having saved the node broadcasts oneself state so greatly.

Node sends lbit data process apart from d, and the energy that consumes is:

$E_{\mathrm{Tx}}(l,d)=\left\{\begin{array}{c}l\&times;E_{\mathrm{elec}}+l\&times;{\mathrm{\&epsiv;}}_{\mathrm{fs}}\&times;d^2,d\&le;d_o\\ l\&times;E_{\mathrm{elec}}+l\&times;{\mathrm{\&epsiv;}}_{\mathrm{mp}}\&times;d^4,d>d_o\end{array}\right.$

Node receives the lbit data, and the energy that consumes is:

E _Rx(l)＝l×E _elec

E wherein _ElecBe sensor node circuit energy consumption, d _oFor distance threshold and

ε _FsBe free space channel model power amplifier coefficient of energy dissipation, ε _MpBe multipath fading channel model power amplifier coefficient of energy dissipation.

Be correctness and the validity of checking this method, we verify to this method that in the farm environment monitoring is used farm environment has three class sensor nodes in using, respectively collecting temperature, humidity and CO ₂Concentration, wireless senser adopts powered battery, and the life span that need as far as possible prolong transducer is gathered more farmland Monitoring Data.The inventive method and LEACH, HEED aspect network life more as shown in Figure 3, the present invention's life span that can obviously prolong network than LEACH, HEED, and the time of first node " death " as can be seen from Figure obviously postpones.This be since the inventive method optimized bunch between energy consumption, bunch head and base station energy consumption and select the relative high node of energy to serve as a bunch head, the energy service efficiency of network is obviously improved, improved the performance of system.

The inventive method and LEACH, HEED algorithm aspect energy efficient more as shown in Figure 4, the present invention can consume still less energy than LEACH, HEED as can be seen from Figure.This is because the inventive method remains optimum bunch quantity, and sub-clustering is even, bunch size stable, be to adopt the adaptive value function of summing up through serious analysis, parameter also gets with experimental summary through serious analysis, finally by particle group optimizing obtain sub-clustering.LEACH algorithm bunch head is selected more at random, and a bunch quantity instability; Balanced in the HEED algorithm is emphasized bunch, consider and factors such as base station distance.

The present invention makes a bunch head be evenly distributed on farm environment and detects in the network by optimizing sub-clustering, the balanced load of network; Take into full account the factor that influences energy consumption during sub-clustering, and used particle swarm optimization algorithm to be optimized, improved energy utilization ratio; The dump energy of node is considered in the selection of bunch head, and leader cluster node has enough energy execution task such as to collect, converge like this, has guaranteed the fail safe of data.This method is simple in a word, and is easy to implement, and the performance of farm environment detection system is improved significantly, and is with good expansibility and robustness.

Claims (5)

1. the network clustering method of wireless sensor based on particle group optimizing is characterized in that wireless senser adopts powered battery, needs the life span of prolonging wireless sensor network as far as possible, and described method step is as follows:

1) bunch head is selected: the base station is according to the situation of current wireless sensor network, dynamically determine optimum bunch number K, move particle swarm optimization algorithm then, an initialization m particle, each particle comprises K candidate's bunch header, calculates the adaptive value of each particle according to the adaptive value function f, after particle cluster algorithm finishes, select adaptive optimal control value particle, the K that this particle a comprises a candidate bunch head is optimum bunch head;

2) bunch formation: the base station is broadcast to wireless sensor network with a bunch header; Sensor node becomes broadcast message of node issue of bunch head after receiving the broadcast message of base station, non-leader cluster node adds corresponding sub-clustering according to the power that receives signal;

3) transfer of data: bunch head is collected the perception information that bunch interior nodes is gathered, and converges to merge the back and adopt single order not have line model to be sent to the base station; When collecting the data of bunch interior nodes collection, information such as node current state can be adopted mode one block movement of " incidentally ", the energy consumption of having saved the node broadcasts oneself state so greatly.

2. according to right 1 described a kind of network clustering method of wireless sensor based on particle group optimizing, it is characterized in that: the number of optimum sub-clustering is:

$K=\sqrt{\frac{N}{2\mathrm{\&pi;}}}\frac{M}{d_{\mathrm{toBS}}}$

Wherein N is the node sum, and M is the perception region area, d _ToBSBe the average distance between surveyed area and base station.

3. according to right 1 described a kind of network clustering method of wireless sensor based on particle group optimizing, it is characterized in that: each particle comprises K candidate's bunch header, and each particle is the 2K dimension, and particle structure is:

p＝{x ₁，y ₁；x ₂，y ₂；x ₃，y ₃...x _K，y _K}

X wherein _l, y _lBe first candidate bunch header that comprises among the particle p, x _K, y _KBe K the candidate's bunch header that comprises among the particle p.

4. according to right 1 described a kind of network clustering method of wireless sensor based on particle group optimizing, it is characterized in that: the adaptive value function is:

f＝αf ₁+βf ₂+γf ₃

$f_1=\underset{k=1}{\overset{K}{\mathrm{\&Sigma;}}}\underset{{\&ForAll;n}_i\&Element;C_{p,k}}{\mathrm{\&Sigma;}}d(n_i,{\mathrm{CH}}_{p,k})/\left|C_{p,k}\right|$

$f_2=\underset{i=1}{\overset{N}{\mathrm{\&Sigma;}}}E\left(n_i\right)/\underset{k=1}{\overset{K}{\mathrm{\&Sigma;}}}E\left({\mathrm{CH}}_{p,k}\right)$

$f_3=\frac{1}{K}\underset{k=1}{\overset{K}{\mathrm{\&Sigma;}}}d({\mathrm{CH}}_{p,k},\mathrm{BS})$

α+β+γ＝1，0≤α，β，γ≤1

N wherein _iBe sensor node i, CH _{P, k}Be a bunch CH who comprises among the particle p _k, | C _{P, k}| be to belong to C among the particle p _kBunch the quantity of node, d is Euclidean distance between the two, E is dump energy, α, beta, gamma are the weight of each component.

5. according to right 4 described a kind of network clustering method of wireless sensor based on particle group optimizing, it is characterized in that described adaptive value function component weight is changed to 0.1, β by 0.8 and is changed to 0.8 by 0.1.

Images