CN101394321A - Non-uniform clustering method for cluster wireless sensor network based on energy balance - Google Patents

Abstract

The invention relates to a power balancing-based non-uniform clustering method for a cluster wireless sensor network, relating to wireless communication technologies. The method comprises the steps of dividing a wireless sensor network monitoring region with uniformly distributed nodes into a non-uniform a ring, determining the positions of cluster head nodes according to an equivalent distance method, ensuring a reliability link index deltap of end-to-end communication between the cluster head nodes to cluster nodes, and setting up an energy consumption solving equation. An optimal ring radius vector is obtained by optimizing parameters of the energy consumption solving equation of the cluster head nodes with a steepest gradient descent algorithm based on the idea of 'power-balancing'. The cluster size is finally obtained by adjusting the positions and the transmission power of the cluster head nodes in accordance with the optimal ring radius vector, so as to balance energy consumption among the cluster head nodes and prolong the network life cycle. The method remarkably prolongs the network life cycle, improves the efficiency of the routing protocol and the media control protocol, and provides a support basis for time synchronization, data integration and target positioning technologies.

Description

Based on the non-homogeneous cluster-dividing method of the cluster wireless sensor network of " balancing energy "

Technical field

The present invention relates to wireless communication technology field, is a kind of solution massive wireless sensor " energy cavity " problem, and prolongs the non-homogeneous cluster-dividing method of network life cycle.

Background technology

A kind of ad hoc deployed wireless networks that wireless sensor network is made up of by communication a large amount of microsensor nodes that are deployed in the monitoring of environmental.The node that constitutes sensor network generally includes aggregation node and sensor node.Wireless sensor network makes wireless sensor network become a popular research direction of current wireless network field in the extensive use in fields such as national defense and military, environmental monitoring, Smart Home, building condition monitoring, space exploration, health care, urban transportation.Its appearance is regarded as a revolution in the information perception and the field of collection.Sensor node is a micro embedded system, its disposal ability, storage capacity and communication capacity relative a little less than, by carrying the powered battery of finite energy, and after deployment, be difficult to makeup energy once more.Therefore how to use the energy of sensor node efficiently, be the matter of utmost importance that will consider in the sensor network design life cycle that prolongs network the biglyyest.Because data flow is transmitted with the pattern of many-one (many-to-one) in the wireless sensor network, need bear more traffic load from the sensor node that aggregation node is nearer, therefore these node energy consumption excessive velocities cause occurring the energy cavity around aggregation node.The appearance in energy cavity makes the original overlay area of network missing data can't be delivered to aggregation node, and finish the life cycle of network in advance, and the lot of energy resource is wasted in the network.

There have been some coping mechanisms to propose at the cavity of the energy in cluster wireless sensor network problem.In the network of multi-hop, the energy cavity tends to forming from the near zone of aggregation node.It is a kind of measure that can alleviate energy cavity problem that Lian J and professor NaikK propose place more sensor node from the nearer place of aggregation node, so-called number of nodes non-uniform Distribution method that Here it is, but interstitial content is increased with geometric ratio to interior ring by outer shroud in this method network, causes the node sum in the network to be exponential growth.Therefore, have only when sensor node can be with low-cost large-scale production, the non-homogeneous method of this number of nodes just has feasibility.The Mihail professor L.Sichitiu sensor node of giving chapter and verse is provided with the method for the energy non-uniform Distribution of different primary powers apart from the distance of base station.Energy non-uniform Distribution method solves the problem in energy cavity from adjusting network energy density, can bear a large amount of communication overheads thereby energy density is higher near the base station.The problem that exists is in the environment of random arrangement, and the actual distribution of node is difficult to reach predefined distribution.Luo J and Hubaux J professor P introduce dynamic the problem that solves the energy cavity in the wireless sensor network.If the position of aggregation node can be moved according to certain method in the network, the node around the aggregation node can be avoided forming the energy cavity around aggregation node along with the time constantly changes like this, eliminates the unbalanced situation of node energy consumption.But frequent mobile sink node can cause the frequent variations of network topology, makes the cost of path discovery increase, and has improved the power consumption of whole network.Change the restriction that the aggregation node position is subjected to physical condition simultaneously, in application, exist difficulty.

Based on characteristics such as self-organizing, wireless channel, the node power of cluster wireless sensor network are controlled, make and implemented for the optimal control of its topological structure.Studies have shown that in wireless sensor network and set up network hierarchical structure, can save the energy of system, prolong the life cycle of system, improve the performance of system, be applicable to massive wireless sensor by the control of sub-clustering (clustering) topology.But the problem that exists is based in the wireless sensor network of sub-clustering technology, data also will merge and transmit simultaneously the data from other leader cluster nodes in leader cluster node (cluster head) was not only managed bunch, the communication mode of multi-hop causes the leader cluster node energy consumption nearer apart from aggregation node too fast and lost efficacy between leader cluster node and the aggregation node, causes other node communication terminals in the node of this bunch and the network.Energy consumption between the leader cluster node unbalanced life cycle that greatly reduces sensor network.LEACH (the low energy adaptive clustering hierarchy) algorithm that Heinzelman professor W proposes passes through random rotation selection leader cluster node equiprobably, and energy load is assigned on each node of network fifty-fifty.HEED (the hybrid energy-eficient distributed clustering) algorithm that Younis professor O proposes is at first chosen some candidate's leader cluster nodes according to the dump energy of node with coming probabilistic, competes producing final leader cluster node then with bunch height of intercommunication cost.LEACH and HEED need periodically to upgrade leader cluster node and form new bunch, so that the energy of whole network reaches balance, have many deficiencies such as the cluster expense is big, the network design area is limited, autgmentability is relatively poor.Leader cluster node adopts the single-hop communication mode directly to communicate by letter with aggregation node more in traditional sub-clustering algorithm, and research emphasis mainly concentrates on the energy consumption between the member node in balanced bunch, does not consider the energy consumption equalization problem between the leader cluster node.

Summary of the invention

Technical problem to be solved by this invention is: at the above-mentioned defective that exists in the prior art; design a kind of non-homogeneous cluster-dividing method based on balancing energy (power-balancing) solve on a large scale based in the cluster wireless sensor network of sub-clustering technology between the leader cluster node because the leader cluster node energy consumption excessive velocities of the close aggregation node that the communication mode of multi-hop (hop-to-hop) causes exhausts the empty problem of energy that lost efficacy and produced; energy consumption between the balance leader cluster node, the life cycle of prolongation network.Be defined as exhausting because of the energy content of battery the earliest in the network life span of the node that lost efficacy in the present invention the life cycle of wireless sensor network.

The technical scheme that the present invention addresses the above problem is: by the control of sub-clustering topology wireless sensor network is created as hierarchical structure type network, with the equally distributed radio sensor network monitoring area dividing of node is endless belt heterogeneous, initially determines the position of leader cluster node according to the method for equivalent distances.In taking all factors into consideration in the large-scale cluster wireless sensor network bunch communication with bunch between the interaction of communicating by letter, the reliability of communicating by letter with assurance leader cluster node end-to-end with aggregation node (end-to-end) links, and sets up the energy consumption solving equation of leader cluster node.With leader cluster node energy consumption balance in the wireless sensor network, the problem that makes leader cluster node consume the minimum and maximization network of maximum power dissipation value life cycle reduces the symbol optimization problem of standard by mathematical analysis.Under based on the thought of " balancing energy ", adopt steepest gradient descent algorithm in the numerical analysis to optimize parameter in the energy consumption equation of leader cluster node, obtain the optimal ring radius vector of endless belt.Separate position and the transmitting power of adjusting leader cluster node according to the ring radius vectors, final definite bunch size.Bunch interior nodes is communicated by letter with leader cluster node in the mode of single-hop, and leader cluster node is communicated by letter with aggregation node in the mode of multi-hop, and the leader cluster node in the endless belt can only be selected the own nearest leader cluster node forwarding data of distance in adjacent the last endless belt.

The present invention has following advantage:

(1) different with traditional heuristic (heuristics) cluster-dividing method based on load balancing (load-balancing), the present invention has provided the method that solves based on the unbalanced problem of leader cluster node energy consumption in the wireless sensor network of sub-clustering technology by mathematical analysis, maximization network problem life cycle is reduced the symbol optimization problem of standard, for the solution of this problem provides good theoretical foundation and found the solution scheme.The steepest gradient descent algorithm has good convergence for finding the solution maximization network problem model life cycle Parameter Optimization.

(2) the present invention propose based on the non-homogeneous cluster-dividing method of balancing energy not only effectively balance energy consumption between leader cluster node, and guaranteed that the reliability of communication link between leader cluster node and the aggregation node is connected.Solved traditional based on the too fast problem that lost efficacy in advance of the nearer leader cluster node energy consumption of the balanced cluster-dividing method middle distance aggregation node of load (load-balancing), significant prolongation the life cycle of network.

(3) the layer-stepping topological structure that the present invention is based on the sub-clustering technology can not only improve the efficient of Routing Protocol and medium control protocol, and with close the getting in touch that exist of optimizing overall performance of network, provide support the basis for time synchronized, data fusion and target localization technology simultaneously.

Description of drawings

Fig. 1 is an implementation procedure flow chart of the present invention

Fig. 2 adopts the non-homogeneous sub-clustering topology diagram of the wireless sensor network based on balancing energy of the present invention

Fig. 3 is that the different cluster-dividing method wireless sensor network cluster of n=2 head node dump energy is than simulation curve figure

Fig. 4 is that the different cluster-dividing method wireless sensor network cluster of n=4 head node dump energy is than simulation curve figure

Fig. 5 is the different cluster-dividing method wireless sensor network of n=2 normalization simulation curve life cycle figure

Fig. 6 is the different cluster-dividing method wireless sensor network of n=4 normalization simulation curve life cycle figure

Embodiment

The present invention is created as hierarchical structure type network by the control of sub-clustering topology with wireless sensor network, is endless belt heterogeneous with the equally distributed radio sensor network monitoring area dividing of node, initially determines the position of leader cluster node according to the method for equivalent distances.In taking all factors into consideration in the large-scale cluster wireless sensor network bunch communication with bunch between the interaction of communicating by letter, the reliability that assurance leader cluster node end-to-end with aggregation node (end-to-end) is communicated by letter links, and sets up the energy consumption model of leader cluster node.With leader cluster node energy consumption balance in the wireless sensor network, the problem that makes leader cluster node consume the minimum and maximization network of maximum power dissipation value life cycle reduces the symbol optimization problem of standard by mathematical analysis.Under based on the thought of " balancing energy ", adopt the parameter of the energy consumption model of the steepest gradient descent algorithm optimization leader cluster node in the numerical analysis, obtain optimal ring radius vector.According to the position and the transmitting power of ring radius vectors adjustment leader cluster node, final definite bunch size.The present invention prolongs the life cycle of network by the size adjusted bunch and the energy consumption between the leader cluster node position balance leader cluster node.

Below at accompanying drawing and instantiation, enforcement of the present invention is specified.

Figure 1 shows that implementation procedure flow chart of the present invention:

1. network configuration initialization

By the control of sub-clustering topology wireless sensor network being created as hierarchical structure type network, is endless belt heterogeneous and the position of initially determining leader cluster node according to the principle of equivalent distances with the equally distributed radio sensor network monitoring area dividing of node.

The initialization network node is evenly distributed among the circular monitored area S that radius is R with density p, and aggregation node is placed in circle centre position; Border circular areas S is divided into K+1 endless belt, and the radius along the center, monitored area to peripheral all endless belt is designated as R respectively ₀, R ₁, R ₂..., R _k, the ring radius vectors is r=(r ₁, r ₂, L, r _K) ^TBased on the symmetry of monitored area and the uniformity of node distribution, R is dropped in the position _i=(ρ, θ) | r _I-1＜ρ≤r _iNode in the endless belt is classified as i type bunch; Be positioned at R ₀=(ρ, θ) | 0＜ρ≤r ₀In the endless belt node directly and aggregation node communicate r wherein ₀＜R; Endless belt R _iIn leader cluster node can only with interior adjacent endless belt R _I-1In nearest leader cluster node communicate in the mode of single-hop; R ₁, R ₂..., R _K-1Leader cluster node in the endless belt not only will be managed a bunch interior nodes, merges each node data and transfers to next-hop node, also will bear the data relay task of other endless belt, R _kNode in the endless belt then need not be born the responsibility of data relaying.

2. leader cluster node position initialization

Because node is to be evenly distributed in the symmetrical circular monitored area, so leader cluster node is evenly distributed in C _i=(ρ, θ) | ρ=d _iCircle on, r wherein _I-1＜d _i≤ r _i, 1≤i≤K.

The present invention initially determines the position of leader cluster node by the method for equivalent distances.With certain endless belt R _iInterior nodes sends every Bit data to endless belt R _I-1The mean value equivalence of the energy that interior leader cluster node consumed is for being positioned at the equally distributed C of leader cluster node _i=(ρ, θ) | ρ=d _iNode sends every Bit data to interior adjacent endless belt R on the annulus _I-1The energy that interior leader cluster node consumed is determined equivalent distances d thus _iValue.Make x _iBe R _iLeader cluster node in the ring arrives R with data passes _I-1The required average distance of leader cluster node in the ring:

$x_i=d_i-d_{i-1}=\frac{2{\∫}_{r_{i-1}}^{r_i}(\mathrm{\ρ}-r_{i-1})\mathrm{\ρd\ρ}}{r_i^2-r_{i-1}^2}=\frac{2(r_i^2+r_i{r}_{i-1}+r_{i-1}^2)}{3(r_i+r_{i-1})}-d_{i-1}---\left(1\right)$

$d_i=\frac{2(r_i^2+r_i{r}_{i-1}+r_{i-1}^2)}{3(r_i+r_{i-1})}---\left(2\right)$

3. set up channel model and leader cluster node energy consumption solving equation

In the present invention takes all factors into consideration in the large-scale cluster wireless sensor network bunch communication with bunch between the interaction of communicating by letter, the reliability that assurance leader cluster node end-to-end with aggregation node (end-to-end) is communicated by letter links, and sets up the energy consumption model of leader cluster node.Node is set receives the required energy of every Bit data and be not less than predetermined energy threshold value τ, guarantee that leader cluster node end-to-end with aggregation node (end-to-end) is communicated by letter and reach reliability and link index δ p.

Adopt the Rayleigh attenuation model to describe between the leader cluster node, the channel model of communicating by letter between leader cluster node and the aggregation node, channel gain is $h\left(x\right)=L\left(d_0\right){\left(\frac{x}{d_0}\right)}^{-n}\mathrm{\ζ}.$ Wherein $L\left(d_0\right)=\frac{G_t{G}_r{l}^2}{16{\mathrm{\π}}^2{d}_0^2},$ L (d ₀) be path loss, G _tBe transmitter antenna gain (dBi), G _tReceiving antenna gain, x are the distance between launch terminal and the receiving terminal, and n path loss index, ζ are that normalized stochastic variable is represented the decay power gain, are exponential distribution, P _r(ζ≤t)=1-e ^-t

In cluster wireless sensor network, bunch interior nodes produces data and sends the data to leader cluster node in the mode of single-hop with speed λ bit/s.Leader cluster node produces data with speed μ bit/s and forward the data to aggregation node in the mode of multi-hop by other leader cluster nodes.According to receiving the energy e that every Bit data circuit is consumed _Rx, send every Bit data circuit consumed energy e of institute _Tx, transmitting power function E _Ti(λ _i+ μ _i, Γ) determine endless belt R _iThe energy consumption E of interior leader cluster node _i:

E _i＝e _rx(λ _i+μ _i)+e _tx(λ _i+μ _i)+E _Ti(λ _i+μ _i，Γ) (3)

Transmitting power function E wherein _Ti(λ _i+ μ _i, Γ) relevant with the routing mechanism that sends data bit flow and leader cluster node.In the present invention, leader cluster node is transmitted data by the shortest path route, i.e. R _iAdjacent R in leader cluster node in the endless belt can only be selected _I-1The own nearest leader cluster node of distance is transmitted data in the endless belt.

E _Ti＝e _ti(λ _i+μ _i) (4)

Bring formula (4) into formula (3) and can get R _i, the leader cluster node communication energy consumes in the ring:

E _i＝(λ _i+μ _i)(e _rx+e _tx+e _ti) (5)

E wherein _TiLaunch the energy that every Bit data consumes for leader cluster node, then receive the energy that every Bit data will consume $e_{\mathrm{ri}}=e_{\mathrm{ti}}L\left(d_0\right){\left(\frac{x}{d_0}\right)}^{-n}\mathrm{\ζ}.$ Under the shortest path route, maximum number of links is K in the end-to-end path, for satisfying the reliability index δ p of end-to-end link, e _RiMust be not less than predetermined energy threshold value τ, δ _lBe short chain road reliability index.Therefore launch the required minimum energy e of every Bit data _TiBe shown below:

${\mathrm{\δ}}_l={\mathrm{\δ}}_p^{\frac{1}{K}}=\mathrm{Pr}(e_{\mathrm{ri}}\≥\mathrm{\τ})=\mathrm{Pr}\{\mathrm{\ζ}\≥\frac{\mathrm{\τ}}{e_{\mathrm{ti}}L\left(d_0\right)}{\left(\frac{x_i}{d_0}\right)}^n\}=e^{-\frac{\mathrm{\ι}{x}_i^n}{e_{\mathrm{ti}}L\left(d_0\right)d_0^n}}---\left(6\right)$

$e_{\mathrm{ti}}=-\frac{\mathrm{K\τ}{x}_i^n}{L\left(d_0\right)d_0^n\ln {\mathrm{\δ}}_p}---\left(7\right)$

Wherein $\left\{\begin{array}{c}{\mathrm{<figure-callout\; id="1"\; label="x"\; filenames="A200810070376E00151.png,A200810070376E00181.png"\; state="\{\{state\}\}">x</figure-callout>}}_1={\mathrm{<figure-callout\; id="1"\; label="d"\; filenames="A200810070376E00151.png,A200810070376E00181.png"\; state="\{\{state\}\}">d</figure-callout>}}_1=\frac{2({\mathrm{<figure-callout\; id="1"\; label="r"\; filenames="A200810070376E00151.png,A200810070376E00181.png"\; state="\{\{state\}\}">r</figure-callout>}}_1^2+r_1{r}_0+r_0^2)}{3({\mathrm{<figure-callout\; id="1"\; label="r"\; filenames="A200810070376E00151.png,A200810070376E00181.png"\; state="\{\{state\}\}">r</figure-callout>}}_1+r_0)}\\ x_i=d_i-d_{i-1}=\frac{2(r_i^2+r_i{r}_{i-1}+r_{i-1}^2)}{3(r_i+r_{i-1})}-\frac{2(r_{i-1}^2+r_{i-1}{r}_{i-2}+r_{i-2}^2)}{3(r_{i-1}+r_{i-2})}\end{array}\right.---\left(8\right)$

R _iTotal data traffic in the endless belt is ${\mathrm{\&lambda;}}_{\mathrm{totali}}=\mathrm{\&pi;}(R^2-r_{i-1}^2)\mathrm{\&rho;\&lambda;},$ For ease of simple in actual applications, the present invention gets the angular width degree and is Ring fan be one bunch, and the cluster-dividing method unanimity of each endless belt, then R _iLeader cluster node number in the endless belt is

R _iThe average data flow of each bunch is in the endless belt $\mathrm{\&lambda;}+\mathrm{\&mu;}=\frac{{\mathrm{\&lambda;}}_{\mathrm{totali}}}{N}=\frac{\mathrm{\&pi;}(R^2-r_{i-1}^2)\mathrm{\&rho;\&lambda;}}{N}.$ Can set up R thus _iLeader cluster node energy consumption solving equation in the endless belt:

Obtain leader cluster node energy consumption solving equation in all endless belt,, solving equation obtains optimum endless belt radius vectors.

4. set up maximization network problem solving life cycle equation

Because the primary power of leader cluster node equates that in order to prolong the life cycle of network substantially, the problem that we need solve is the energy consumption of each leader cluster node of balance, and makes the ceiling capacity consumption of leader cluster node drop to minimum.For this reason, set up original equation:

Minimize?max{E ₁，E ₂...E _k} (10)

The present invention is by optimizing ring radius vectors r=(r ₁, r ₂, L, r _K) ^T, adjust the position and the transmitting power of leader cluster node, the size of determining bunch with the energy consumption of this each leader cluster node of balance, and makes the ceiling capacity consumption of leader cluster node drop to minimum.Promptly get the maximization network problem solving life cycle equation of belt restraining condition:

$\left\{\begin{array}{c}{\min }_{\{r_i,...,r_K\}}\left\{\max (E_1\left(r\right),...,E_K\left(r\right)\}\right\}\\ s.t\\ r_0<{\mathrm{<figure-callout\; id="1"\; label="r"\; filenames="A200810070376E00151.png,A200810070376E00181.png"\; state="\{\{state\}\}">r</figure-callout>}}_1<...<r_K=R\end{array}\right.---\left(11\right)$

5. according to leader cluster node energy consumption solving equation iterative optimal ring radius vector

The present invention adopts the steepest gradient descent algorithm in the numerical analysis to calculate the optimal ring radius solution vector, the negative gradient vector components iteration adjustment ring radius vectors of the energy consumption function by step factor coefficient and leader cluster node, make the energy consumption function convergence of leader cluster node, try to achieve optimal ring radius vector.Its concrete steps are as follows:

(1) introduces auxiliary variable t 〉=E _i(r), 1≤i≤K wherein is with maximization network problem life cycle

Reduce the symbol optimization problem (signomial optimal problem) of standard:

$\left\{\begin{array}{c}{\min }_{\{r,t\}}t\\ s.t\\ t^{-1}{E}_i\left(r\right)\&le;1,\\ r_{i-1}{r}_i^{-1}<1,\\ r_K=R.i=1,...,K\end{array}\right.---\left(12\right)$

(2) initialization ring radius vectors r=(r ₁, r ₂, L, r _K) ^T

(3) with r=(r ₁, r ₂, L, r _K) ^TSubstitution leader cluster node energy consumption solving equation calculates the energy consumption values E=(E that each encircles interior leader cluster node ₁, E ₂, L, E _K) ^T

(4) obtain the energy consumption E=(E of leader cluster node ₁, E ₂, L, E _K) ^TIn maximum E _MAXAnd corresponding ring radius r _i

(5) according to formula $r_i^{\left(k\right)}=r_i^{(k-1)}-{\mathrm{\&beta;}}_i^{\left(k\right)}\frac{{\&PartialD;E}_i}{{\&PartialD;r}_i}{|}_{n=r_i^{(k-1)}}$ Revise ring radius vectors value r=(r ₁, r ₂, L, r _K) ^T, wherein Be expression step factor coefficient,

Be that leader cluster node energy consumption function exists $r_i=r_i^{(k-1)}$ The single order local derviation at place is illustrated in $r_i=r_i^{(k-1)}$ The negative gradient vector components of contour surface;

(6) iterate (3) (4), (5), E=(E under the condition that is satisfying (12) formula ₁, E ₂, L, E _K) ^TConvergence;

(7) drawing the ring radius vectors of this moment is the optimal ring vector and separates $r\&prime;={({r^{\&prime;}}_1,{r^{\&prime;}}_2,L,{r^{\&prime;}}_K)}^T.$

6. utilize formula (2) and angular width degree according to the optimal ring vector that obtains

Value is determined the leader cluster node position, utilizes formula (1) to adjust the transmitting power of leader cluster node, definite bunch size.

According to method of the present invention determine leader cluster node position, transmitting power and bunch size, set up the network of layer-stepping topological structure thus, the energy consumption of this network between can the balance leader cluster node prolongs the life cycle of network.

Be the network topology structure figure that the present invention is based on the layer-stepping topological structure of sub-clustering technology as shown in Figure 2, sensor node is evenly distributed among the circular monitored area S that radius is R with density p, only exist an aggregation node to be fixed on the center of monitored area S in the network, aggregation node is not subjected to the constraint of energy; According to the optimal ring radius vector of trying to achieve $r\&prime;={({r^{\&prime;}}_1,{r^{\&prime;}}_2,L,{r^{\&prime;}}_K)}^T$ Adjust K+1 endless belt dividing monitored area S, remember R respectively ₀, R ₁, R ₂..., R _k, correspondingly the node in the network be divided into K type bunch, R is dropped in the position _i=(ρ, θ) | r _I-1＜ρ≤r _iSensor node in the endless belt is classified as i type bunch, is positioned at R ₀Node in the endless belt directly and aggregation node communicate; R _iThe endless belt interior nodes is divided into

Individual bunch, node precognition partition information, leader cluster node is with density Be evenly distributed in C _i=(ρ, θ) | ρ=d _iCircle on, r wherein _I-1＜d _i≤ r _i, 1≤i≤K; Leader cluster node has data-handling capacity and the energy reserve higher than ordinary node, and the radio transmitted power may command, and the primary power of leader cluster node is identical; R _iLeader cluster node in the ring can only be selected R _I-1The own nearest leader cluster node of distance comes Data transmission in the ring; R ₁, R ₂..., R _K-1Leader cluster node in the endless belt is not only managed and is merged bunch interior nodes data and transfer to next-hop node, also bears the data relay task of other endless belt, R _kNode in the endless belt then need not be born the responsibility of data relaying; Bunch interior nodes is communicated by letter with leader cluster node in the mode of single-hop, and leader cluster node is communicated by letter with aggregation node by the multi-hop communication mode.

We compared by experiment the non-homogeneous cluster-dividing method that the present invention is based on balancing energy with based on the even distribution cluster-dividing method of load balancing with based on probability vector p=(p ₁, p ₂, L, p _K) ^TEven cluster-dividing method.Core concept based on the even distribution cluster-dividing method of load balancing is division in the circular monitored area to be waited the endless belt of width, correspondingly node be divided into equal sizes bunch, bunch interior nodes is communicated by letter by the mode of single-hop with leader cluster node, and leader cluster node is given aggregation node by the mode Data transmission of multi-hop.Based on the even cluster-dividing method of probability vector with based on the even distribution cluster-dividing method of load balancing different be that leader cluster node is with probability vector p=(p ₁, p ₂, L, p _K) ^TSelect Data transmission to give aggregation node at random to the leader cluster node or the direct Data transmission of next jumping.The experimental result proof is better than other two kinds of methods based on the non-homogeneous cluster-dividing method of balancing energy greatly in the performance aspect network life cycle and the network energy surplus ratio.

Accompanying drawing 3,4 is respectively path loss coefficient n=2, different cluster-dividing method wireless sensor network cluster head node energy surplus ratio simulation curve figure during n=4.The network energy surplus ratio is meant when finish the life cycle of network, the ratio between the primary power summation of all nodes in energy remaining summation and the network in the network.We can see that when path attenuation index n=2, along with the increase of network radius, when network life cycle finished, employing was lower than 5% based on the leader cluster node energy surplus ratio of the network of the non-homogeneous cluster-dividing method of balancing energy is very for a short time, and tends towards stability; Employing is higher based on the leader cluster node energy surplus ratio of the outer shroud of network of the even cluster-dividing method of probability vector; Employing increases to be with radius based on the network cluster head node energy surplus ratio of the even cluster-dividing method of load balancing and increases progressively trend, and when network life cycle finished, leader cluster node still had the energy near 90% to remain in the network outer shroud.Employing based on the energy surplus ratio of the network of the non-homogeneous cluster-dividing method of balancing energy will be when adopting other two kinds of node location modes the energy surplus ratio of network.When path attenuation index n=4, adopt network performance instability based on the non-uniform Distribution method of probability vector, energy residue in subregion is higher; Employing still increases progressively with the increase of radius based on the leader cluster node energy surplus ratio of the network of the even location mode of load balancing; Employing is significantly less than other two kinds of methods based on the leader cluster node energy surplus ratio of the network of the non-homogeneous cluster-dividing method of balancing energy.

Accompanying drawing 5,6 is respectively n=2 for the path loss coefficient, different cluster-dividing method wireless sensor network cluster head node simulation curve life cycle figure during n=4.Be defined as exhausting because of the energy content of battery the earliest the life span of the leader cluster node that lost efficacy in the present invention the life cycle of wireless sensor network.As can be seen from the figure adopt based on the network of the even cluster-dividing method of load balancing the very fast premature dead of leader cluster node energy consumption, thereby the life cycle of network is shorter near aggregation node; Based on the network lifecycle of the even cluster-dividing method of probability vector is stable but life cycle is shorter, but be better than network based on the even cluster-dividing method of load balancing; Compare with other two kinds of methods, the life cycle that has well prolonged network based on the non-homogeneous cluster-dividing method of balancing energy, and along with tend towards stability the life cycle of the increase network of network radius.

Claims (5)

1, the non-homogeneous cluster-dividing method of a kind of cluster wireless sensor network based on balancing energy, it is characterized in that, by the control of sub-clustering topology wireless sensor network is created as hierarchical structure type network, and its monitored area is divided into endless belt heterogeneous, initially determine the position of leader cluster node according to the method for equivalent distances, guarantee that leader cluster node reaches reliability with the aggregation node end-to-end communication and links index δ p, set up the energy consumption solving equation of leader cluster node, utilize the steepest gradient descent algorithm to find the solution leader cluster node energy consumption equation, determine optimum endless belt radius vectors, adjust the position and the transmitting power of leader cluster node by optimum endless belt radius vectors, definite bunch size, the energy consumption of each leader cluster node of balance.

2, non-homogeneous cluster-dividing method according to claim 1 is characterized in that, the concrete grammar of described definite equivalent distances comprises, with certain endless belt R _iInterior nodes sends every Bit data adjacent endless belt R in it _I-1The mean value equivalence of the energy that interior leader cluster node consumed is endless belt R _iInterior is d apart from the monitored area center radius _iAnnulus on node send every Bit data to interior adjacent endless belt R _I-1The energy that interior leader cluster node consumed is determined equivalent distances d _i

3, non-homogeneous cluster-dividing method according to claim 1 is characterized in that, according to receiving every Bit data circuit consumed energy e of institute _Rx, send every Bit data circuit consumed energy e of institute _Tx, transmitting power function E _Ti(λ _i+ μ _i, Γ), set up the energy consumption solving equation of leader cluster node.

4, non-homogeneous cluster-dividing method according to claim 1, it is characterized in that, bunch interior nodes is communicated by letter with leader cluster node in the mode of single-hop, leader cluster node is communicated by letter with aggregation node in the mode of multi-hop, and the leader cluster node in the endless belt can only be selected the own nearest leader cluster node forwarding data of distance in the adjacent last endless belt.

5, non-homogeneous cluster-dividing method according to claim 1, it is characterized in that, described steepest gradient descent algorithm, the negative gradient vector components iteration adjustment ring radius vectors of the energy consumption function by step factor coefficient and leader cluster node, make the energy consumption function convergence of leader cluster node, try to achieve optimal ring radius vector.

Images