CN103281746B - A kind of wireless sense network route method of quotient topology energy hierarchical Dynamic Programming - Google Patents

Abstract

A kind of wireless sense network route method <b> of quotient topology energy hierarchical Dynamic Programming, the network lifecycle first run is divided quotient set according to sensing net node geographical attribute by </b>, draw alternate divisions scheme, find the empty sub-clustering of the optimum first floor and record bunch head by competition.The each void of the first floor bunch finds respective optimum submanifold in the same way, with this hierarchical until each sub-clustering can not down divide again.Each bunch of last one deck is final effectively sub-clustering, and its each leader cluster node is by forming transmission path with respective father's leader cluster node step by step.Between each partitioning site first judging last round of ground floor taking turns from second, whether integral energy is unbalance.Then call first round same method reconstruct as disequilibrium and make clustering routing.If balance, retain and divide and select new bunch of head, under then judging further, whether one deck balances with each bunch of father, and takes ground floor the same manner Recursion process until the sub-clustering of formation bottom, construct new route.The present invention can construct efficient low-consume route, extend Network morals.

Description

A kind of wireless sense network route method of quotient topology energy hierarchical Dynamic Programming

Technical field

The present invention relates to sensing network routing algorithm, be specifically related to a kind of wireless sense network route method of passing rank game based on energy under the quotient topology of geographical position.

Background technology

To in the research and practice process of the method, the present inventor finds: WSN(wireless sense network) by disposing wireless sensor node in a large number in specific monitored area, by gathering the information of perceived object in specific wireless sensing network overlay area, by the communication of single-hop or multi-hop, to collect and data message after process is supplied to the infrastructure and network support that terminal use .WSN do not need to fix, there is rapid deployment, the features such as survivability is strong, military surveillance can be widely used in, environmental monitoring, medical monitoring, agricultural breeding and other commercial fields, and space exploration and disaster such as to be speedily carried out rescue work at the special dimension.The efficient routing protocol research of WSN is one of the research emphasis in this field always in recent years.

Clustering route protocol is route-type important in WSN, in clustering route protocol, network is divided into bunch (cluster) usually. and so-called bunch, there is the set of network nodes of certain association exactly. each bunch is made up of member (clustermember) in a bunch of head (clusterhead) and multiple bunches, in bunch, member is by bunch head sink information, by bunch head to communicate with base station BS (basestation) clustering routing mechanism there is following advantage (1) the member node most of the time can communication close module, the connected network forming a more last layer by bunch head carrys out the long distance routing forwarding of responsible data. so both ensure that the data communication in original coverage, also network energy is saved to a great extent, (2) bunch head forwards after having merged the data of member node again, decreases data traffic, thus saves network energy, (3) function of member node is fairly simple, complicated routing iinformation need not be safeguarded. this considerably reduce the quantity of network routing control information, decrease the traffic: (4) clustering topology is convenient to management, be conducive to the application of distributed algorithm, fast reaction can be made to system change, there is good extensibility, be applicable to large scale network: (5) and plane road by compared with, more easily overcome sensor node and move the problem brought.One of difficult point of current main cluster-based techniques more difficultly takes into account the harmonious and geographical distribution harmony of energy ezpenditure.

In topology and relevant art of mathematics thereof, a quotient space (quotientspace, also referred to as identification space identificationspace) says it is by equivalent for some points in a given space or " sticking together " intuitively; Determine which point is equivalent by an equivalence relation.This is the common methods from the new space of given spatial configuration.By to initial data manifold structure or induce the varigrained quotient topological space, Theory of Quotient Space is exactly the relation between each quotient topological space of research, comprises the theory of synthesis, comprehensive, decomposition and the problem reasoning in the different quotient space.

Summary of the invention

Object of the present invention: a kind of wireless sense network route method obtained by the laddering energy game strategies of layering between diverse geographic location quotient topology is provided.The method can consider energy and the geographical distribution characteristic of WSN node, all node energy consumption of efficient balance, extends network life cycle.

The solution of the present invention: the quotient set inducing different levels according to sensing net node geographical attribute divides (i.e. sub-clustering), each submanifold interior joint radix of same bunch of regulation is identical simultaneously.Carry out energy game between different levels point cluster node by the thought of hierarchical and determine whether to need to divide (i.e. sub-sub-clustering) and how optimal dividing.After in the end obtaining final effectively sub-clustering, its each leader cluster node is by forming multi-hop transmission path to guarantee that each bunch of information more effectively can be sent to Sink with the leader cluster node of each self-recording father bunch step by step.Between each partitioning site first judging last round of ground floor taking turns from second, whether integral energy is unbalance.Then call first round same method reconstruct as disequilibrium and make clustering routing.If balance, retain and divide and select new bunch of head, under then judging further, whether one deck balances with each bunch of father, and takes ground floor the same manner Recursion process until the sub-clustering of formation bottom, construct new route.Technical solution of the present invention can consider energy and the balanced clustering routing of geographical distribution characteristic structure sense network energy of WSN node, effectively extends network lifecycle.

Concrete steps of the present invention are as follows:

1, according to region attribute of a relation, wireless sensing net topology Wsn is induced N kind alternate divisions scheme, specifies often kind of alternate divisions to be the step of k small grain size submanifold to be:

A () node just no longer changes its geographical position once deployment, base station records the node distributed intelligences such as the dump energy of each node and geographical indicator;

B () passes through a selected geographical distribution attribute, as certain height above sea level, or certain horizontal longitude and latitude, Wsn is divided into K equivalence class node submanifold and Wsn ₁', Wsn ₂' ... Wsn _k';

C () selected different geographical distribution attribute repeats to perform step b for N time and obtains N kind alternate divisions scheme;

2, the competition that energy balance and energy estimate consumption is carried out to this N number of alternate divisions, as obtained optimal dividing i.e. optimum next straton sub-clustering and Wsn ₁', Wsn ₂' ... Wsn _k', then enter the following step 3; Otherwise this wireless sense network Wsn originally as finally effective bunch, then enters the following step 4, concrete steps:

A () selects one of alternate divisions;

B (), in K sub-sub-clustering, all selects some optimums bunch head to every sub-sub-clustering and based on these bunch of head by computational prediction minimal energy consumption value Consume (Wsn_ _i') wherein i ∈ 1,2 ... k};

C () aggregation step b obtains k energy predicting consumption the predict energy Consume (Wsn) consumed is needed to compare with father bunch, then cancelling this division as shown in the formula dividing post consumption energy (1) Suo Shi and being more than or equal to the front consumption of division, consuming as shown in the formula then thinking shown in (2) that this division is effectively alternate divisions before dividing as being less than;

$\mathrm{Consume}\left(\mathrm{Wsn}\right)\&le;\underset{i=1}{\overset{k}{\mathrm{\&Sigma;}}}\mathrm{Consume}\left(\mathrm{Ws}{n_i}^{\&prime;}\right)\&CenterDot;\&CenterDot;\&CenterDot;\left(1\right)$

$\mathrm{Consume}\left(\mathrm{Wsn}\right)>\underset{i=1}{\overset{k}{\mathrm{\&Sigma;}}}\mathrm{Consume}\left(\mathrm{Ws}{n_i}^{\&prime;}\right)\&CenterDot;\&CenterDot;\&CenterDot;\left(2\right)$

(d) repeat step a to step c as there is no any alternate divisions then this father bunch basis as finally effective bunch and enter step 4, if there be m alternate divisions, enter next step e;

E () calculates respective K sub-cluster node dump energy Energe (Wsn of m alternate divisions respectively ₁') wherein i ∈ 1,2 ..., k}

And balanced degree $\mathrm{Balance}(\mathrm{Ws}{n_1}^{\&prime;},\mathrm{Ws}{n_2}^{\&prime;},\&CenterDot;\&CenterDot;\&CenterDot;\mathrm{Ws}{n_k}^{\&prime;})=$

${\mathrm{Log}}_{\frac{1}{e}}\frac{\mathrm{Max}\{\mathrm{Energe}\left(\mathrm{Ws}{n_1}^{\&prime;}\right),\mathrm{Energe}\left(\mathrm{Ws}{n2}^{\&prime;}\right),\&CenterDot;\&CenterDot;\&CenterDot;\mathrm{Energe}\left(\mathrm{Ws}{n_k}^{\&prime;}\right)\}-\mathrm{Min}\{\mathrm{Energe}\left(\mathrm{Ws}{n_1}^{\&prime;}\right),\mathrm{Energe}\left(\mathrm{Wsn}{2}^{\&prime;}\right),\&CenterDot;\&CenterDot;\&CenterDot;\mathrm{Energe}\left(\mathrm{Ws}{n_k}^{\&prime;}\right)\}}{\mathrm{Min}\{\mathrm{Energe}\left(\mathrm{Ws}{n_1}^{\&prime;}\right),\mathrm{Energe}\left(\mathrm{Wsn}{2}^{\&prime;}\right),\&CenterDot;\&CenterDot;\&CenterDot;\mathrm{Energe}\left(\mathrm{Ws}{n_k}^{\&prime;}\right)\}}$

The consumed energy of f energy balance degree Balance () and prediction that () considers each alternate divisions reduces degree Save (), selects optimum sub-clustering.The consumed energy minimizing degree of prediction and discriminant function are as shown in the formula (3) and formula (4).

$\mathrm{Save}(\mathrm{Ws}{n_1}^{\&prime;},\mathrm{Ws}{n_2}^{\&prime;},...\mathrm{Ws}{n_k}^{\&prime;})={\mathrm{Log}}_{1/e}\frac{\mathrm{Consume}\left(\mathrm{Wsn}\right)-{\mathrm{\&Sigma;}}_{i=1}^k\mathrm{consume}\left(\mathrm{Ws}{n_i}^{\&prime;}\right)}{{\mathrm{\&Sigma;}}_{i=1}^k\mathrm{consume}\left(\mathrm{Ws}{n_i}^{\&prime;}\right)}\&CenterDot;\&CenterDot;\&CenterDot;\left(3\right)$

αSave（Wsn ₁’，Wsn ₂’，…Wsn _k’）+βBanance（Wsn ₁’，Wsn ₂’，…Wsn _k’）………（4）

Note α and β is adjustable parameter

3, for selected this K submanifold, it is as follows to enter step 1 step respectively:

A () is for this K selected submanifold Wsn ₁', Wsn ₂' ... Wsn _k', extract wherein any one Wsn _i', i ∈ { 1,2 ... k}

(b) Wsn _i' enter step 1. as the sub-Wsn that lower one deck one is relatively independent

C () all the other K-1 submanifold also enters step a and b respectively and carries out next distinguishing hierarchy;

4, determine all finally after effective bunch, the leader cluster node of effective bunch to give bunch in member distribution T DMA timetable

(a) Sink node broadcast announcement corresponding node who be bunch head and father at different levels bunch head and who be bunch in member;

(b) each effective sub-clustering bunch head according to bunch in member node add up to it and distribute time slot, and broadcast, ordinary node namely can in the time of distributing to a bunch head transmission of information afterwards;

5, a bunch head collects this bunch of information, is implemented to the multi-hop transmission of Sink node by the leader cluster node of its father in cluster process bunch;

A () leader cluster node carries out data fusion after having have collected together the information that this bunch of ordinary node send;

B data are received the leader cluster node of father bunch and send Sink node step by step back to by () leader cluster node;

Note: father bunch is not final effectively sub-clustering, excessive bunch that just produces in clustering process;

6, first judge after entering next round last round of whether exist ground floor divide;

A) according on take turns the whole Wsn of historical record and whether there is the effective K of ground floor equivalence class node submanifold { Wsn ₁', Wsn ₂' ... Wsn _k' divide;

7, just directly bunch head is changed, as finally effective bunch as do not divided; And enter step 4;

If a) do not divided, think that Wsn is the final effectively sub-clustering of epicycle, more renew bunch head and enter step 4;

8, divide as existed and just investigate these with younger brother's father and elder brothers sub-clustering and virtual whether integral energy is unbalance;

A) just this k is investigated a bit with younger brother's father and elder brothers sub-clustering and Wsn according to formula 3 as there is division ₁', Wsn ₂' ... Wsn _k' whether integral energy is unbalance;

If 9 integral energies also keep balance, more whether each bunch that investigates balance exist lower floor's division separately, and enter step 7;

If a) integral energy also keeps balancing, each bunch that investigates balance successively , Wsn ₂' ... Wsn _k'

B) Wsn is determined _i' noly there is lower floor and divide, and by Wsn _i' be denoted as Wsn and enter step 7;

If 10 integral energies no longer balance, more renew bunch head and enter step 1;

If a) integral energy does not keep balancing, change bunch head of Wsn;

B) Wsn changing new bunch of head is substituted into step 1;

11, step 1 is repeated to step 10 until all node energies exhaust.

Beneficial effect of the present invention: the method can consider energy and the geographical distribution characteristic of WSN node, all node energy consumption of efficient balance, extends network life cycle.

Accompanying drawing explanation

Fig. 1 is a typical sensors network diagram of the present invention;

In figure: 1, Sink node 2, bunch and bunch member 3, monitored area;

Fig. 2 is the schematic diagram that wireless sensing net topology of the present invention induces N kind alternate divisions;

In figure: 1, alternate divisions;

Fig. 3 the present invention often takes turns flow chart;

Fig. 4 multi-hop data transmission of the present invention schematic diagram;

In figure: 1 base station (Skin node), 2 bunches, 3 bunches heads, 4, bunch member.

Embodiment

As shown in Figure 1,2,3, 4, be described as follows:

(establishing with k=2, N=8 as example)

1) according to region attribute of a relation, wireless sensing net topology Wsn is induced N kind alternate divisions scheme, specifies often kind of alternate divisions to be the step of two small grain size submanifolds to be:

A () node just no longer changes its geographical position once deployment, base station records the node distributed intelligences such as the dump energy of each node and geographical indicator.

B Wsn is divided into 2 equivalence class node submanifold { Wsn by a selected geographical distribution attribute (as certain height above sea level, or certain horizontal longitude and latitude) by () ₁', Wsn ₂'.

C () selected different geographical distribution attribute repeats to perform step b for 8 times and obtains 8 kinds of alternate divisions schemes.

2) competition that energy balance and energy estimate consumption is carried out to above-mentioned N number of alternate divisions, as obtained optimal dividing (i.e. optimum next straton sub-clustering Wsn1 ', Wsn2 '), then enter the following step 3, otherwise this wireless sense network Wsn is originally as finally effective bunch, then enter the following step 4, concrete steps:

A () selects one of alternate divisions.

B (), in 2 sub-sub-clusterings, all selects some optimums bunch head to every sub-sub-clustering and based on these bunch of head by computational prediction minimal energy consumption value. wherein i ∈ { 1,2}.

C () aggregation step b obtains 2 energy predicting consumption and closes the predict energy that consumes is needed with father's bunch (not dividing) compare, as divide post consumption energy shown in as shown in the formula (1) and be more than or equal to divide before consumption then cancel this division, consume as shown in the formula then thinking shown in (2) that this division is effectively alternate divisions before dividing as being less than.

$\mathrm{Consume}\left(\mathrm{Wsn}\right)\&le;\underset{i=1}{\overset{2}{\mathrm{\&Sigma;}}}\mathrm{Consume}\left(\mathrm{Ws}{n_i}^{\&prime;}\right)\&CenterDot;\&CenterDot;\&CenterDot;\left(1\right)$

$\mathrm{Consume}\left(\mathrm{Wsn}\right)>\underset{i=1}{\overset{2}{\mathrm{\&Sigma;}}}\mathrm{Consume}\left(\mathrm{Ws}{n_i}^{\&prime;}\right)\&CenterDot;\&CenterDot;\&CenterDot;\left(2\right)$

(d) repeat step a to step c as there is no any alternate divisions then this father bunch basis as finally effective bunch and enter step 4, if there be m alternate divisions, enter next step e.

E () calculates respective 2 sub-cluster node dump energy Energe (Wsn of m alternate divisions respectively ₁') wherein i ∈ { 1,2}

And balanced degree $\mathrm{Balance}(\mathrm{Ws}{n_1}^{\&prime;},\mathrm{Ws}{n_2}^{\&prime;})=$

${\mathrm{Log}}_{\frac{1}{e}}\frac{\mathrm{Max}\{\mathrm{Energe}\left(\mathrm{Ws}{n_1}^{\&prime;}\right),\mathrm{Energe}\left(\mathrm{Ws}{n2}^{\&prime;}\right)\}-\mathrm{Min}\{\mathrm{Energe}\left(\mathrm{Ws}{n_1}^{\&prime;}\right),\mathrm{Energe}\left(\mathrm{Wsn}{2}^{\&prime;}\right)\}}{\mathrm{Min}\{\mathrm{Energe}\left(\mathrm{Ws}{n_1}^{\&prime;}\right),\mathrm{Energe}\left(\mathrm{Wsn}{2}^{\&prime;}\right)\}}$

The consumed energy of f energy balance degree Balance () and prediction that () considers each alternate divisions reduces degree Save (), selects optimum sub-clustering.The consumed energy minimizing degree of prediction and discriminant function are as shown in the formula 3 and formula 4.

$\mathrm{Save}(\mathrm{Ws}{n_1}^{\&prime;},\mathrm{Ws}{n_2}^{\&prime;})={\mathrm{Log}}_{1/e}\frac{\mathrm{Consume}\left(\mathrm{Wsn}\right)-{\mathrm{\&Sigma;}}_{i=1}^2\mathrm{consume}\left(\mathrm{Ws}{n_i}^{\&prime;}\right)}{{\mathrm{\&Sigma;}}_{i=1}^2\mathrm{consume}\left(\mathrm{Ws}{n_i}^{\&prime;}\right)}\&CenterDot;\&CenterDot;\&CenterDot;\left(3\right)$

αSave（Wsn ₁’，Wsn ₂’）+βBanance（Wsn ₁’，Wsn ₂’）………（4）

Note α and β is adjustable parameter

3) for selected this K submanifold, it is as follows to enter step 1 step respectively:

A () is for these selected 2 submanifold Wsn ₁', Wsn ₂', extract wherein any one Wsn _i',

(b) Wsn _i' enter step 1. as the sub-Wsn that lower one deck one is relatively independent

C () another 1 submanifold also enters step a and b and carries out next distinguishing hierarchy.

4) determine all finally after effective bunch, the leader cluster node of effective bunch to give bunch in member distribution T DMA timetable

(a) Sink node broadcast announcement corresponding node who be bunch head and father at different levels bunch head and who be bunch in member.

(b) each effective sub-clustering bunch head according to bunch in member node add up to it and distribute time slot, and broadcast, ordinary node namely can in the time of distributing to a bunch head transmission of information afterwards.

5) a bunch head collects this bunch of information, is implemented to the multi-hop transmission of Sink node by the leader cluster node of its father in cluster process bunch.

A () leader cluster node carries out data fusion after having have collected together the information that this bunch of ordinary node send.

B data are received the leader cluster node of father bunch and send Sink node back to step by step by () leader cluster node.Note: father bunch is not final effectively sub-clustering, excessive bunch that just produces in clustering process.

6) first judge after entering next round last round of whether exist ground floor divide.

(a) according on take turns the whole Wsn of historical record and whether there are effective 2 the equivalence class node submanifold { Wsn of ground floor ₁', Wsn ₂' divide.

7) just directly bunch head is changed, as finally effective bunch as do not divided.And enter step 4.

If a () does not divide, think that Wsn is the final effectively sub-clustering of epicycle, more renew bunch head and enter step 4.

8) as existence division just investigates these, with younger brother's father and elder brothers sub-clustering (virtual), whether integral energy is unbalance.

A () divides as existed and just investigates these 2 a bit with younger brother's father and elder brothers sub-clustering (Wsn according to formula 3 ₁', Wsn ₂') whether integral energy is unbalance.

9) if integral energy also keeps balance, more whether each bunch that investigates balance exist lower floor's division separately, and enter step 7.

If a () integral energy also keeps balancing, each bunch that extracts balance successively out

B () determines Wsn _i' noly there is lower floor and divide, and by Wsn _i' be denoted as Wsn and enter step 7.

10) if integral energy no longer balances, more renew bunch head and enter step 1.

If a () integral energy does not keep balancing, change bunch head of Wsn.

B the Wsn changing new bunch of head is substituted into step 1 by ().

11) step 1 is repeated to step 10 until all node energies exhaust.

Claims (1)

1. a wireless sense network route method for quotient topology energy hierarchical Dynamic Programming, is characterized in that:

(1) according to region attribute of a relation, wireless sensing net topology Wsn is induced N kind alternate divisions scheme, specifies often kind of alternate divisions to be the step of k small grain size submanifold to be:

A () node just no longer changes its geographical position once deployment, base station records the node distributed intelligences such as the dump energy of each node and geographical indicator;

B () passes through a selected geographical distribution attribute, as certain height above sea level, or certain horizontal longitude and latitude, Wsn is divided into K equivalence class node submanifold and Wsn ₁', Wsn ₂' ... Wsn _k';

C () selected different geographical distribution attribute repeats to perform step b for N time and obtains N kind alternate divisions scheme;

(2) competition that energy balance and energy estimate consumption is carried out to above-mentioned N number of alternate divisions, as obtained optimal dividing (i.e. next straton sub-clustering and Wsn of optimum ₁', Wsn ₂' ... Wsn _k', then enter the following step (three); Otherwise this wireless sense network Wsn originally as finally effective bunch, then enters the following step (four), concrete steps:

A () selects one of alternate divisions;

B (), in K sub-sub-clustering, all selects some optimums bunch head to every sub-sub-clustering and based on these bunch of head by computational prediction minimal energy consumption value Consume (Wsn_ _i') wherein i ∈ 1,2 ... k};

C () aggregation step b obtains k energy predicting consumption the predict energy Consume (Wsn) consumed is needed to compare with father bunch, then cancelling this division as shown in the formula dividing post consumption energy (1) Suo Shi and being more than or equal to the front consumption of division, consuming as shown in the formula then thinking shown in (2) that this division is effectively alternate divisions before dividing as being less than;

$\mathrm{Consume}\left(\mathrm{Wsn}\right)\&le;\underset{i=1}{\overset{k}{\mathrm{\&Sigma;}}}\mathrm{Consume}\left({{\mathrm{Wsn}}_i}^{\&prime;}\right)...\left(1\right)$

$\mathrm{Consume}\left(\mathrm{Wsn}\right)>\underset{i=1}{\overset{k}{\mathrm{\&Sigma;}}}\mathrm{Consume}\left({{\mathrm{Wsn}}_i}^{\&prime;}\right)...\left(2\right)$

(d) repeat step a to step c as there is no any alternate divisions then this father bunch basis as finally effective bunch and enter step (four), if there be m alternate divisions, enter next step e;

E () calculates respective K sub-cluster node dump energy Energe (Wsn of m alternate divisions respectively ₁') wherein i ∈ 1,2 ..., k}

And balanced degree $\mathrm{Balance}({{\mathrm{Wsn}}_1}^{\&prime;},{{\mathrm{Wsn}}_2}^{\&prime;},...{{\mathrm{Wsn}}_k}^{\&prime;})=$

${\mathrm{Log}}_{\frac{1}{e}}\frac{\mathrm{Max}\{\mathrm{Energe}\left({{\mathrm{Wsn}}_1}^{\&prime;}\right),\mathrm{Energe}\left({\mathrm{Wsn}2}^{\&prime;}\right),...\mathrm{Energe}\left({{\mathrm{Wsn}}_k}^{\&prime;}\right)\}-\mathrm{Min}\{\mathrm{Energe}\left({{\mathrm{Wsn}}_1}^{\&prime;}\right),\mathrm{Energe}\left({\mathrm{Wsn}2}^{\&prime;}\right),...\mathrm{Energe}\left({{\mathrm{Wsn}}_k}^{\&prime;}\right)\}}{\mathrm{Min}\{\mathrm{Energe}\left({{\mathrm{Wsn}}_1}^{\&prime;}\right),\mathrm{Energe}\left({\mathrm{Wsn}2}^{\&prime;}\right),...\mathrm{Energe}\left({{\mathrm{Wsn}}_k}^{\&prime;}\right)\}}$

The consumed energy of f energy balance degree Balance () and prediction that () considers each alternate divisions reduces degree Save (), selects optimum sub-clustering; The consumed energy minimizing degree of prediction and discriminant function are as shown in the formula 3 and formula 4;

$\mathrm{Save}({{\mathrm{Wsn}}_1}^{\&prime;},{{\mathrm{Wsn}}_2}^{\&prime;},...{{\mathrm{Wsn}}_k}^{\&prime;})={\mathrm{Log}}_{1/e}\frac{\mathrm{Consume}\left(\mathrm{Wsn}\right)-{\mathrm{\&Sigma;}}_{i=1}^k\mathrm{consume}\left({{\mathrm{Wsn}}_i}^{\&prime;}\right)}{{\mathrm{\&Sigma;}}_{i=1}^k\mathrm{consume}\left({{\mathrm{Wsn}}_i}^{\&prime;}\right)}...\left(3\right);$

αSave(Wsn ₁’，Wsn ₂’，…Wsn _k’)+βBanance(Wsn ₁’，Wsn ₂’，…Wsn _k’)

………(4)；

Note α and β is adjustable parameter;

(3) for selected this K submanifold, it is as follows to enter step (one) step respectively:

A () is for this K selected submanifold Wsn ₁', Wsn ₂' ... Wsn _k', extract wherein any one Wsn _i', i ∈ { 1,2 ... k}

(b) Wsn _i' enter step () as the sub-Wsn that lower one deck one is relatively independent;

C () all the other K-1 submanifold also enters step a and b respectively and carries out next distinguishing hierarchy;

(4) determine all finally after effective bunch, the leader cluster node of effective bunch to give bunch in member distribution T DMA timetable

(a) Sink node broadcast announcement corresponding node who be bunch head and father at different levels bunch head and who be bunch in member;

(b) each effective sub-clustering bunch head according to bunch in member node add up to it and distribute time slot, and broadcast, ordinary node namely can in the time of distributing to a bunch head transmission of information afterwards;

(5) a bunch head collects this bunch of information, is implemented to the multi-hop transmission of Sink node by the leader cluster node of its father in cluster process bunch;

A () leader cluster node carries out data fusion after having have collected together the information that this bunch of ordinary node send;

B data are received the leader cluster node of father bunch and send Sink node step by step back to by () leader cluster node;

Note: father bunch is not final effectively sub-clustering, excessive bunch that just produces in clustering process;

(6) first judge after entering next round last round of whether exist ground floor divide;

A) according on take turns the whole Wsn of historical record and whether there is the effective K of ground floor equivalence class node submanifold { Wsn ₁', Wsn ₂' ... Wsn _k' divide;

(7) just directly bunch head is changed, as finally effective bunch as do not divided; And enter step (four);

If a) do not divided, think that Wsn is the final effectively sub-clustering of epicycle, more renew bunch head and enter step (four);

(8) divide as existed and just investigate these with younger brother's father and elder brothers sub-clustering and virtual whether integral energy is unbalance;

A) just this k is investigated a bit with younger brother's father and elder brothers sub-clustering and Wsn according to formula 3 as there is division ₁', Wsn ₂' ... Wsn _k' whether integral energy is unbalance;

(9) if integral energy also keeps balance, more whether each bunch that investigates balance exist lower floor's division separately, and enter step (seven);

If a) integral energy also keeps balancing, each bunch that investigates balance successively ∈ { Wsn ₁', Wsn ₂' ... Wsn _k';

B) Wsn is determined _i' noly there is lower floor and divide, and by Wsn _i' be denoted as Wsn and enter step (seven);

(10) if integral energy no longer balances, more renew bunch head and enter step ();

If a) integral energy does not keep balancing, change bunch head of Wsn;

B) Wsn changing new bunch of head is substituted into step ();

(11) repeat step () to step (ten) until all node energies exhaust.