CN102946626B - Node sleeping scheduling method under environment of several-for-one clustering wireless sensor network - Google Patents

Abstract

The invention discloses a node sleeping scheduling method under an environment of a several-for-one clustering wireless sensor network. The node sleeping scheduling method comprises the following steps: 1) initializing a network; 2) obtaining a data aggregation tree of a wireless sensor network and obtaining a route matrix R; 3) confirming a periodic running method of a cluster head node, constructing a cross-layer optimization model based on energy consumption, and obtaining a sleeping scheduling method for each cluster head node; 4) starting to run and uploading data to a father node; 5) before each node enters into a sleeping state, obtaining the next node sleeping scheduling method for the node; 6) judging if the energy of the cluster head node is used up; and 7) if the energy of the cluster head node in a backbone network is used up, adopting a routing lookup algorithm and judging if the present remained backbone network node can reconstruct the data aggregation tree, if yes, returning to the step 2, and if not, ending. According to the node sleeping scheduling method provided by the invention, the times for switching the node state are extremely reduced and the energy loss of idle monitoring is avoided.

Description

A kind of many-one cluster wireless sensor network environment lower node sleep scheduling method

Technical field

The present invention relates to a kind of many-one cluster wireless sensor network environment lower node sleep scheduling method, belong to wireless sensor network technology field.

Background technology

Wireless sensor network is a kind of wireless network of foundation-free facility, can Real-Time Monitoring, the various environment in perception and collection network distributed areas or the information of monitoring target, and these data are processed, finally information exchange is crossed to self-organizing network and send user to.

Sensor network nodes quantity is huge, and be often distributed in the severe area of environment, inconvenience is charged at any time or is changed battery, but network requires again to have the longer life-span, therefore, how rationally to utilize sensor node energy, the procotol of design low energy consumption, extending sensor network life, is a major issue of wireless sensor network research.

The wireless transmission of wireless sensor node can be divided into idle listening, receives data, send data and four kinds of states of sleep.The power difference consuming when generally idle listening is with transmission is little, is dormant 10 ³~ 10 ⁴doubly.Because the time of data that network environment produces is fixing, a lot of nodes are the state in idle listening for a long time, has consumed a large amount of energy.

In existing sleep scheduling algorithm, core concept is all to reduce the time of node in idle listening state, make node as much as possible in sleep state, but there are two aspect defects: in (1) partial sleep dispatching algorithm, be that the simple node by completing the work of data send and receive is switched to sleep state from idle listening state, can farthest guarantee the node sleep time although it is so, but, state switches frequently, also can consume a lot of energy, total energy consumption not necessarily reduces, and, switching state frequently, node related work device is frequently closed, start, also can reduce to a certain extent the node life-span, (2) in the sleep scheduling algorithm having, also consider the problems referred to above, taked according to transfer of data situation, reduce the mode of state switching times, the energy consumption of reduction state conversion portion, but such planning, can not farthest reduce idle listening, according to planning, some node will can not be switched to sleep state when idle listening.

Summary of the invention

The object of the invention is in order to address the above problem, a kind of many-one cluster wireless sensor network environment lower node sleep scheduling method is proposed, the present invention is guaranteeing under the prerequisite of network data transmission, by the node dormancy in idle listening state, save a large amount of energy consumption, by building the cross-layer optimization model based on energy consumption, obtain leader cluster node sleep scheduling method, reduce network energy consumption, guarantee being uniformly distributed of each node energy consumption in backbone network, prolong network lifetime.

A many-one cluster wireless sensor network environment lower node sleep scheduling method, comprises following step:

Step 1: carry out netinit;

Step 2, according to Routing Algorithm, obtain the data aggregate tree of wireless sensor network, obtain route matrix R;

Step 3, determine the periodic duty method of leader cluster node, build the cross-layer optimization model based on energy consumption, obtain the sleep scheduling method of each leader cluster node;

Step 4, wireless sensor network are according to node sleep dispatching method, and each node brings into operation according to the sleep scheduling method of this node, to father node uploading data;

Step 5, when each node is after father node sends data, before entering sleep state, according to current wireless sensor network state, obtain the input of the cross-layer optimization model based on energy consumption, adopt the cross-layer optimization model based on energy consumption that in step 3, (2) build, obtain this node node sleep dispatching method next time, and the method is informed to father node and the brotgher of node of this node;

Step 6, wireless sensor network judge whether the energy of leader cluster node exhausts, and if so, proceeds to step 7, otherwise, return to step 4;

If there is leader cluster node depleted of energy in step 7 backbone network, adopt Routing Algorithm, whether the backbone network node that judges current residual can rebuild data aggregate tree, if can rebuild data aggregate tree, return to step 2, otherwise backbone network depleted of energy, wireless sensor network stops operation.

The invention has the advantages that:

(1) the present invention has designed the periodic duty method of energy-conservation leader cluster node, node sleep dispatching algorithm is dissolved into the TDMA(Time Division Multiple Access of backbone network, time division multiple access) planning in.According to designed method, can guarantee that, in one-period, each node only wakes up once, after waking up, complete collect bunch in data, collect son node number according to, send the work such as data, Regeneration planning is laggard enters sleep state.Can farthest reduce the number of times of node state conversion like this, on the one hand; On the other hand, in planning according to network routing condition, wakeup time to father node and child node and the brotgher of node has carried out accommodation, can be converted to sleep state by the idle listening state by node as much as possible, has avoided the energy loss of idle listening.

(2) in the present invention, the routing infrastructure of backbone network is the initial conditions of algorithm, and therefore, algorithm can calculate for different routing infrastructures optimal scheme.Meanwhile, the planning in invention is real-time update in network operation process, so algorithm goes for static routing and dynamic routing.

(3) in the present invention, designed the cross-layer optimization model based on energy consumption, in model, considered every factor of physical layer, MAC layer, network layer, the programme of energy consumption optimum can better be calculated, the drawback of having avoided some factor of one-sided consideration to bring according to network actual conditions.

Accompanying drawing explanation

Fig. 1 is the backbone network leader cluster node cycle of operation figure in the present invention;

Fig. 2 is overview flow chart of the present invention;

Embodiment

Below in conjunction with drawings and Examples, the present invention is described in further detail.

The present invention is a kind of many-one cluster wireless sensor network environment lower node sleep scheduling method, design energy-conservation node cycle of operation figure, according to node cycle of operation figure and physical layer, MAC layer, network layer parameter builds the cross-layer optimization model based on energy consumption, guarantee that each backbone network node only wakes up once in one-period, reduce the time of node in idle listening, reduce network energy consumption, by regulating the data aggregate rate of each node, guarantee that energy consumption is uniformly distributed on each node of backbone network, prolong network lifetime, method flow as shown in Figure 2, mainly comprise following step:

Step 1: carry out netinit;

Wireless sensor network environment is carried out to initialization, complete deployment, the sub-clustering of nodes, after deployment, obtain node distributed intelligence, after sub-clustering, have a leader cluster node in each bunch, all leader cluster nodes form backbone network.

SINK(aggregation node in setting data polymerization tree) node is numbered 1, the Random assignment when node deployment of the numbering of backbone network node.

Step 2, according to Routing Algorithm, obtain the data aggregate tree of wireless sensor network, obtain route matrix R;

(for example: OSPF choose existing a kind of Routing Algorithm, Open Shortest Path First ospf), using the distributed intelligence of backbone network node as input, obtain being adapted to the data aggregate tree of wireless sensor network, the summit of tree is SINK node.According to data aggregate tree, obtain route matrix R.

According to route matrix R, obtain father node matrix F N, child node Matrix C N, layer matrix L, fraternal matrix B N, for searching of follow-up energy consumption calculation father node, child node, the number of plies, the brotgher of node.Be specially:

Route matrix R is:

wherein

Wherein: N represents the quantity of leader cluster node in N backbone network, r _ijrepresent that leader cluster node i is to the link break-make situation of leader cluster node j.

Route matrix R has following feature: diagonal is 0, and symmetric position is opposite number, r _i. represent the connected relation of leader cluster node i and other leader cluster nodes, owing to being tree structure, so each leader cluster node only has a father node, have a plurality of child nodes, r _i. in only have one for+1, all the other are 0 or-1, the quantity that-1 quantity is child node.

Father node matrix F N:

The father node numbering fn of leader cluster node i _irepresent, father node matrix is: FN=[fn ₁fn ₂fn _n].

If leader cluster node j is the father node of leader cluster node i, r in route matrix _ij=1, fn _i=j, if leader cluster node i does not have fn of father node _ibe 0.

Child node Matrix C N:

The child node numbering of leader cluster node i is used represent, in the numbering of each child node by incremental order, arrange, child node matrix is $\mathrm{CN}={\left[\begin{array}{cccc}\stackrel{\→}{{\mathrm{cn}}_1}& \stackrel{\→}{{\mathrm{cn}}_2}& \·\·\·& \stackrel{\→}{{\mathrm{cn}}_N}\end{array}\right]}^T,$ Do not have position and other vacant positions of child node all to mend 0.

According to route matrix R, if leader cluster node j is the child node of leader cluster node i, there is r _ij=-1,

Layer matrix L:

The number of plies L of leader cluster node i in convergence tree _irepresent, i.e. L=[L ₁l ₂l _n].

Wherein, setting sink node serial number is 1, is positioned at ground floor, i.e. L ₁=1.

According to child node matrix, can obtain the child node of each leader cluster node, if the number of plies of a leader cluster node is l, the number of plies of its child node is l+1, by that analogy, can calculate the number of plies of all nodes.

Brotgher of node matrix B N:

The brotgher of node numbering of leader cluster node i is used represent, in the numbering of each brotgher of node by incremental order, arrange, $\mathrm{BN}={\left[\begin{array}{cccc}\stackrel{\→}{{\mathrm{bn}}_1}& \stackrel{\→}{{\mathrm{bn}}_2}& \·\·\·& \stackrel{\→}{{\mathrm{bn}}_N}\end{array}\right]}^T,$ Do not have position and other vacant positions of the brotgher of node all to mend 0.

According to father node matrix and child node matrix, each leader cluster node can obtain all child node numberings that have identical father node with this node, is the brotgher of node numbering of this node.

Step 3, determine the periodic duty method of leader cluster node, build the cross-layer optimization model based on energy consumption, obtain the sleep scheduling method of each leader cluster node.

(1) determine the periodic duty method of leader cluster node;

The leader cluster node cycle of operation, the operation method of each leader cluster node was as shown in Figure 1:

1, leader cluster node, in sleep state, when leader cluster node needs transceiving data, switches to operating state by leader cluster node by sleep state;

State conversion (SC) time of leader cluster node is: father node is than the late Δ T that wakes up of first child node, and when guaranteeing that father node can start to receive son node number certificate, this child node has been ready to send data; Between the brotgher of node, child node below will guarantee to complete and send after data in previous child node successively than the late Δ t that wakes up of previous child node, and this child node is ready to send data to father node just.

2, leader cluster node collect bunch in data;

Leader cluster node is collected the perception data of bunch interior nodes in one-period.Bunch interior nodes be dispose in advance, only for perception data and by data upload to leader cluster node, do not possess the function of route and gateway.Bunch interior nodes after corresponding leader cluster node is started working, by the data upload of perception in one-period to leader cluster node, by leader cluster node via the data aggregate tree of backbone network by convergence to SINK node.

If 3 leader cluster nodes have child node, leader cluster node is collected son node number certificate;

In backbone network, leader cluster node is collected the data of its child node in data aggregate tree, and child node is also the leader cluster node in backbone network, and the at this moment effect of backbone network leader cluster node is the data that forward other leader cluster nodes.

4, leader cluster node sends data to father node;

After backbone network leader cluster node collection data are complete, through data aggregate, the data upload that oneself is stored, to father node, is served as the responsibility of data retransmission by father node.Child node will be carried out synchronously with father node before sending data to father node, periodically to father node, sends synchronization request bag (SYN), obtains father node and confirms after bag (ACK), then start data transmission.After data are sent, child node need to be waited for the feedback packet (FB) that father node sends, and judges whether data send successfully, if this secondary data sends unsuccessfully, storage data, wait for that leader cluster node preferentially resends this secondary data next time while waking (switching to operating state by sleep state) up.

5, leader cluster node enters sleep;

Backbone network leader cluster node completes after the collection and transmission work of data, and the energy consumption loss of avoiding unnecessary idle listening to bring, switches to sleep state by node, reduces node energy consumption.

(2) build the cross-layer optimization model based on energy consumption;

1, get parms:

Packet loss matrix P is:

work as r _ij=0 o'clock, p _ij=0

Wherein: p _ijrepresent that leader cluster node i is to the packet loss on the link of leader cluster node j.

Data transmission rate matrix F is:

work as r _ij=0 o'clock, f _ij=0

Wherein: f _ijrepresent that leader cluster node i is to the data transmission rate on the link of leader cluster node j.

Energy consumption matrix E is:

$E={\left[E_{\mathrm{ij}}\right]}_{N\×N}=\left[\begin{array}{cccccc}{E}_{\·1}& E_{\·2}& E_{\·3}& E_{\·4}& E_{\·5}& E_{\·6}\end{array}\right]\left[\begin{array}{cccccc}{E}_{11}& E_{12}& E_{13}& E_{14}& E_{15}& E_{16}\\ E_{21}& E_{22}& E_{23}& E_{24}& E_{25}& E_{26}\\ \·& \·& \·& \·& \·& \·\\ \·& \·& \·& \·& \·& \·\\ \·& \·& \·& \·& \·& \·\\ E_{N1}& E_{N2}& E_{N3}& E_{N4}& E_{N5}& E_{N6}\end{array}\right]$

Wherein: i represents leader cluster node numbering, j represents the energy properties numbering of node, E _i1the state conversion energy consumption that represents leader cluster node i; E _i2the energy consumption that represents data in the receipts bunch of leader cluster node i; E _i3the energy consumption that represents the receipts son node number certificate of leader cluster node i; E _i4the energy consumption that represents the transmission data of leader cluster node i; E _i5the sleep energy consumption that represents leader cluster node i; E _i6the data processing energy consumption that represents leader cluster node i.

The energy consumption of i node is the total energy consumption of all nodes is

Sleep scheduling matrix T:

$T={\left[t_{\mathrm{ij}}\right]}_{N\×4}=\left[\begin{array}{cccc}{t}_{\·1}& t_{\·2}& t_{\·3}& t_{\·4}\end{array}\right]\left[\begin{array}{cccc}{t}_{11}& t_{12}& t_{13}& t_{14}\\ t_{21}& t_{22}& t_{23}& t_{24}\\ \·& \·& \·& \·\\ \·& \·& \·& \·\\ \·& \·& \·& \·\\ t_{N1}& t_{N2}& t_{N3}& t_{N4}\end{array}\right]$

Wherein: i represents leader cluster node numbering, j represents the time attribute numbering of node, t _i1the time that wake up the next time of expression leader cluster node i; t _i2the time of data in receiving the next time that represents leader cluster node i bunch; t _i3the time of receiving son node number certificate next time that represents leader cluster node i; t _i4the time that sends data next time that represents leader cluster node i.

Data aggregate rate matrix AG:

Because network is data aggregate tree structure, easily find, the closer to the backbone network node of SINK node, need the data volume of forwarding larger, energy consumption is also just larger, likely causes exhausting too early near the backbone node energy consumption of SINK node, and then the situation of whole network paralysis.Therefore, the strategy that need to add data aggregate in network, the data volume that need to transmit according to node is different, sets different data aggregate rates, reduce to a certain extent the data volume that the backbone node near SINK node need to transmit, make as far as possible the balanced energy consumption of each node in backbone network.

The data aggregate rate of leader cluster node i is ag _i, i.e. AG=[a _g1ag ₂ag _n].

Basic power consumption:

State is changed primary energy consumption into E _change, the energy consumption of carrying out a data aggregate is E _agg, sending data power consumption is P _send, receiving data power consumption is P _rec, idle listening power consumption is P _idle, sleep power consumption is P _sleep.

Feedback matrix FB:

The last feedback fb that sends data of leader cluster node i _irepresent, when needs resend data, fb _ibe recorded as the data volume that need to resend.Fb during initialization _ibe 0.

FB=[fb ₁fb ₂fb ₃fb _n], wherein

The packet loss that sends data to father node due to node i is therefore fb _i=S _{send, i}probability be s _{send, i}the data volume that expression need to resend.When the data of node i send successfully, node i will be received feedback information ACK, fb _i=0; When the data of node i send unsuccessfully, node i will receive feedback information NAK or can not receive feedback information, fb _i=S _{send, i}, wherein, S _{send, i}the data volume that represents this transmission of node i.

Set other parameters:

T _{brc_i}represent that leader cluster node i can start to receive the moment of son node number certificate;

T _{bs_i}=t _i1+ t _i2+ t _i3represent that leader cluster node i can start to send to father node the moment of data.

S _sYNthe size that represents SYN bag;

S _aCKthe size that represents ACK bag;

S _fBthe size that represents FB bag;

2, derivation

The calculating of each several part energy consumption:

For leader cluster node i, there is E _i=E _i1+ E _i2+ E _i3+ E _i4+ E _i5+ E _i6, represent the energy consumption of node i in one-period, be specifically calculated as follows.

(1) E _i1: state conversion energy consumption

If according to planning before, node i is waken up once in one-period, therefore E _i1=E _change.

(2) E _i2: the energy consumption of data in receiving bunch

Backbone node i just represents in network i bunch, in establishing bunch, has K _iindividual node, the data arrival rate of each node is λ _ik, bunch (bit/s) in, the network cycle of operation is T _{i_ch}, in bunch, be divided into G _iindividual group, every group of operating time is T _{i_ch}/ G _i, the work period T of leader cluster node i namely _i=T _{i_ch}/ G _i.

Node i within a work period need to bunch in the data volume that gathers be if the data transmission rate in bunch during data upload is μ _i(bit/s), node i receive bunch in time of data be

Therefore: $E_{i2}=P_{\mathrm{rec}}\·t_{i2}=P_{\mathrm{rec}}\·\underset{k=1}{\overset{K_i}{\mathrm{\Σ}}}{\mathrm{\λ}}_{\mathrm{ik}}\·\frac{T_{i\_\mathrm{ch}}}{G_i{\mathrm{\μ}}_i}.$

(3) E _i3: the energy consumption of receiving son node number certificate

If node i has M _iindividual child node, according to node cycle of operation figure, can write out the energy consumption that node i receives m child node and be: E _{i3, m}=E _{wait_SYN, m}+ E _{rec_SYN, m}+ E _{aCK, m}+ E _{receive, m}+ E _{fB, m}.

Wherein, E _{wait_SYN, m}for waiting for synchronous energy consumption, E _{rec_SYN, m}for receiving the energy consumption of sync packet, E _{aCK, m}for sending the energy consumption of ACK, E _{receive, m}for receiving the energy consumption of data, E _fBfor sending the energy consumption of feedback.Specifically be calculated as follows:

Wait for synchronous energy consumption: E _{wait_SYN, m}=P _idlet _{wait_SYN, m}

Wherein, $t_{\mathrm{wait}\_\mathrm{SYN},m}=\left\{\begin{array}{cc}{t}_{\mathrm{bs}\_{\mathrm{cn}}_{\mathrm{im}}}-t_{\mathrm{brc}\_i,m}& t_{\mathrm{bs}\_{\mathrm{cn}}_{\mathrm{im}}}>t_{\mathrm{brc}\_i,m}\\ t_{\mathrm{bs}\_{\mathrm{cn}}_{\mathrm{im}}}+S_m(t_{\mathrm{SYN}}\_t_{\mathrm{inter}\_\mathrm{SYN}})-t_{\mathrm{brc}\_i,m}& t_{\mathrm{bs}\_{\mathrm{cn}}_{\mathrm{im}}}<t_{\mathrm{brc}\_i,m}\end{array}\right.$ For waiting for the synchronous time.

Wherein, represent that m child node of node i is ready to send to father node the moment of data, the 1st, 2,3 time attribute that represent respectively m child node of node i, t _sYNthe time that represents node sending/receiving synchronization request bag (SYN), t _{inter_SYN}represent to send the time interval of SYN twice;

represent that node i is ready to receive the moment of m sub-node data, wherein, t _aCKthe time that represents node sending/receiving confirmation bag (ACK), t _{rec, m-1}represent that m-1 child node of node i receives the time of data, t _fBthe time that represents node sending/receiving feedback packet (FB);

m the child node that represents node i carried out the quantity of replying the sync packet of wrapping of not receiving of transmission when synchronous.

Receive the energy consumption of sync packet: wherein, m the child node that represents leader cluster node i is to the data transmission rate on the link of leader cluster node i.

Send the energy consumption of ACK: wherein, represent that leader cluster node i is to the data transmission rate on the link of m the child node of leader cluster node i.

Receive the energy consumption of data:

With represent the data package size receiving,

Send the energy consumption of feedback: $E_{\mathrm{FB}}=P_{\mathrm{send}}\&CenterDot;\frac{S_{\mathrm{FB}}}{f_{i,{\mathrm{cn}}_{\mathrm{im}}}}.$

(4) E _i4: the energy consumption that sends data

According to node cycle of operation figure, can write out:

E _i4＝E _SYN+E _{inter_SYN}+E _{rec_ACK}+E _send+E _{rec_FB}

Wherein, E _sYNfor sending the energy consumption of synchronization request bag (SYN), E _{inter_SYN}for synchronization request interval energy consumption, E _{rec_ACK}for the energy consumption of confirmation of receipt bag (ACK), E _sendfor sending the energy consumption of data, E _{rec_FB}for receiving the energy consumption of feedback packet (FB).Specifically be calculated as follows:

The father node of node i is fn _i, i is fn _ic child node, the moment that node i is ready for sending data is t _{bs_i}, the moment that father node is prepared its data of reception is have:

Send the energy consumption of synchronization request bag: E _sYN=P _send(s _i+ 1) t _sYN

Wherein, represent to send the time of synchronization request bag (SYN), the data transmission rate of the link of expression from leader cluster node i to father node, the quantity of the sync packet that does not obtain reply sending in expression node i request synchronizing process.

Synchronization request interval energy consumption: E _{inter_SYN}=P _idles _it _{inter_SYN}.

Receive the energy consumption of ACK: E _aCK=P _rect _aCK, wherein

Send the energy consumption of data:

Here the data that send comprise two parts, the data that gather in bunch and the data of collecting from child node.If it is unsuccessful that the last data of node send, need the last data of preferential transmission, then send the data of this collection.

The data that gather in bunch and the data of collecting from child node need to be entered data aggregate, sent again, and established after data aggregate, the data volume that node i need to send to father node is $S_{\mathrm{send},i}={\mathrm{fb}}_i+{\mathrm{ag}}_i(\underset{k=1}{\overset{K_i}{\mathrm{\&Sigma;}}}{\mathrm{\&lambda;}}_{\mathrm{ik}}\&CenterDot;T_i+\underset{m=1}{\overset{M_i}{\mathrm{\&Sigma;}}}{S}_{\mathrm{send},{\mathrm{cn}}_{\mathrm{im}}}).$

The energy consumption that sends data is $E_{\mathrm{send}}=P_{\mathrm{send}}\&CenterDot;t_{\mathrm{send},i}=P_{\mathrm{send}}\&CenterDot;\frac{S_{\mathrm{end},i}}{f_{i,f{n}_i}}.$

Receive the energy consumption of feedback: E _fB=P _rec. _tFb, wherein

(5) E _i5: sleep energy consumption

Each node, when carrying out next sleep scheduling method, wouldn't be deleted current sleep scheduling matrix T, because T will be as the foundation of planning next time.Current matrix is T, and matrix is next time T ', complete planning enter sleep state before, make T=T ', T ' empties.

Node i length of one's sleep in next cycle is t _sleep=t _i1'-(t _i1+ t _i2+ t _i3+ t _i4).

The sleep energy consumption of node i in next cycle is E _i5=P _sleept _sleep.

(6) E _i6: data processing energy consumption

Data processing energy consumption refers to that node carries out the energy loss in data aggregate process, and each node carries out data aggregate one time in one-period, therefore E _i6=E _agg.

To sum up, can calculate the total energy consumption of network in one-period is:

$E_{\mathrm{total}}=\underset{i=1}{\overset{N}{\mathrm{\&Sigma;}}}\underset{j=1}{\overset{6}{\mathrm{\&Sigma;}}}{E}_{\mathrm{ij}}$

$=\underset{i=1}{\overset{N}{\mathrm{\&Sigma;}}}(E_{i1}+E_{i2}+E_{i3}+E_{i4}+E_{i5}+E_{i6})$

$=\underset{i=1}{\overset{N}{\mathrm{\&Sigma;}}}[E_{\mathrm{change}}+E_{\mathrm{agg}}+P_{\mathrm{rec}}\&CenterDot;\underset{k=1}{\overset{K_i}{\mathrm{\&Sigma;}}}{\mathrm{\&lambda;}}_{\mathrm{ik}}\&CenterDot;\frac{T_{i\_\mathrm{ch}}}{G_i{\mathrm{\&mu;}}_i}+\underset{m=1}{\overset{M_i}{\mathrm{\&Sigma;}}}(P_{\mathrm{idle}}\&CenterDot;t_{\mathrm{wait}\_\mathrm{SYN},m}+P_{\mathrm{rec}}\&CenterDot;\frac{S_{\mathrm{SYN}}}{f_{{\mathrm{cn}}_{\mathrm{im}},i}}+P_{\mathrm{send}}\&CenterDot;\frac{S_{\mathrm{ACK}}}{f_{i,c{n}_{\mathrm{im}}}}$

$+P_{\mathrm{rec}}\&CenterDot;\frac{S_{\mathrm{send},{\mathrm{cn}}_{\mathrm{im}}}}{f_{{\mathrm{cn}}_{\mathrm{im}},i}}+P_{\mathrm{send}}\&CenterDot;\frac{S_{\mathrm{FB}}}{f_{i,{\mathrm{cn}}_{\mathrm{im}}}})+(P_{\mathrm{send}}\&CenterDot;(s_i+1)t_{\mathrm{SYN}}+P_{\mathrm{idle}}\&CenterDot;s_i\&CenterDot;t_{\mathrm{inter}\_\mathrm{SYN}}+P_{\mathrm{rec}}\&CenterDot;\frac{S_{\mathrm{ACK}}}{f_{{\mathrm{fn}}_i,i}}$

$+P_{\mathrm{send}}\&CenterDot;\frac{{\mathrm{fb}}_i+{\mathrm{ag}}_i(\underset{k=1}{\overset{K_i}{\mathrm{\&Sigma;}}}{\mathrm{\&lambda;}}_{\mathrm{ik}}\&CenterDot;T_i+\underset{m=1}{\overset{M_i}{\mathrm{\&Sigma;}}}{S}_{\mathrm{send},{\mathrm{cn}}_{\mathrm{im}}})}{f_{i,{\mathrm{fn}}_i}}+P_{\mathrm{rec}}\&CenterDot;\frac{S_{\mathrm{FB}}}{f_{{\mathrm{fn}}_i,i}})+P_{\mathrm{sleep}}\&CenterDot;t_{\mathrm{sleep}}]$

$=\underset{i=1}{\overset{N}{\mathrm{\&Sigma;}}}[E_{\mathrm{change}}+E_{\mathrm{agg}}+P_{\mathrm{rec}}(\underset{k=1}{\overset{K_i}{\mathrm{\&Sigma;}}}{\mathrm{\&lambda;}}_{\mathrm{ik}}\&CenterDot;\frac{T_{i\_\mathrm{ch}}}{G_i{\mathrm{\&mu;}}_i}+\underset{m=1}{\overset{M_i}{\mathrm{\&Sigma;}}}(\frac{S_{\mathrm{SYN}}}{f_{{\mathrm{cn}}_{\mathrm{im}},i}}+\frac{S_{\mathrm{send},{\mathrm{cn}}_{\mathrm{im}}}}{f_{{\mathrm{cn}}_{\mathrm{im}},i}})+\frac{S_{\mathrm{ACK}}}{f_{{\mathrm{fn}}_i,i}}+\frac{S_{\mathrm{FB}}}{f_{{\mathrm{fn}}_i,i}})$

$+P_{\mathrm{send}}(\underset{m=1}{\overset{M_i}{\mathrm{\&Sigma;}}}(\frac{S_{\mathrm{ACK}}}{f_{i,{\mathrm{cn}}_{\mathrm{im}}}}+\frac{S_{\mathrm{FB}}}{f_{i,{\mathrm{cn}}_{\mathrm{im}}}})+(s_i+1)\frac{S_{\mathrm{SYN}}}{f_{i,{\mathrm{fn}}_i}}+\frac{{\mathrm{fb}}_i+{\mathrm{ag}}_i(\underset{k=1}{\overset{K_i}{\mathrm{\&Sigma;}}}{\mathrm{\&lambda;}}_{\mathrm{ik}}\&CenterDot;T_i+\underset{m=1}{\overset{M_i}{\mathrm{\&Sigma;}}}{S}_{\mathrm{send},{\mathrm{cn}}_{\mathrm{im}}})}{f_{i,{\mathrm{fn}}_i}})$

$+P_{\mathrm{idle}}(\underset{m=1}{\overset{M_i}{\mathrm{\&Sigma;}}}{t}_{\mathrm{wait}\_\mathrm{SYN},m}+s_i\&CenterDot;t_{\mathrm{inter}\_\mathrm{SYN}})+P_{\mathrm{sleep}}\&CenterDot;t_{\mathrm{sleep}}]$

The model that can be optimized is:

minE _total

$s.t.|t_{\mathrm{il}}-(t_{{\mathrm{cn}}_{i1}}+\mathrm{\&Delta;T})|\&le;\mathrm{\&delta;}$

$|t_{i1}-(t_{{\mathrm{bn}}_{i1}}+\mathrm{\&Delta;t})|\&le;\mathrm{\&delta;}.$

0＜ag _i＜1(0＜i≤N)

$M_i{\stackrel{\&OverBar;}{\mathrm{\&lambda;}}}_i<f_{i,{\mathrm{fn}}_i}$

In constraints, represent waking up the time difference constantly of father node and its first child node, the time difference of waking the moment up that represents leader cluster node and its first brotgher of node, δ < < 1,0 < ag _i< 1 (0 < i≤N) represents the scope of the data aggregate rate of leader cluster node i, the data arrival rate that represents leader cluster node i is less than this node to the data transmission rate of father node transmission data, wherein, and M _ibunch interior nodes quantity that represents leader cluster node i place bunch, bunch interior nodes average data arrival rate that represents leader cluster node i place bunch, represent that leader cluster node i is to the data transmission rate of the link of father node.

According to constraints, solve E _totalminimum value, obtains sleep scheduling matrix T, by sleep scheduling matrix T determine receive time of waking up the next time of each leader cluster node, next time bunch in time, time of receiving son node number certificate next time of data, send the time of data next time.

(3) obtain each leader cluster node sleep scheduling method first;

By the cross-layer optimization model based on energy consumption obtaining in step (2), determined the time in the periodic duty method of leader cluster node in step (1), obtained the sleep scheduling method first of each node, each leader cluster node sleep scheduling method is first completed by SINK node, and informs successively other leader cluster nodes in wireless sensor network.

Step 4, wireless sensor network are according to node sleep dispatching method, and each node brings into operation according to the sleep scheduling method of this node, to father node uploading data.

Step 5, when each node is after father node sends data, before entering sleep state, according to current wireless sensor network state, obtain the input (change of network state is mainly the variation of data volume and the variation of adjacent node sleep scheduling method that current each leader cluster node need to send) of the cross-layer optimization model based on energy consumption, adopt the cross-layer optimization model based on energy consumption that in step 3, (2) build, obtain this node node sleep dispatching method next time, and the method is informed to father node and the brotgher of node of this node.

Step 6, wireless sensor network judge whether the energy of leader cluster node exhausts, and if so, proceeds to step 7, otherwise, return to step 4.

If there is leader cluster node depleted of energy in step 7 backbone network, adopt Routing Algorithm, whether the backbone network node that judges current residual can rebuild data aggregate tree (is worked as Routing Algorithm, when having output, be judged as backbone network node and can rebuild data aggregate tree), if can rebuild data aggregate tree, return to step 2, otherwise backbone network depleted of energy, wireless sensor network stops operation.

Claims (2)

1. a many-one cluster wireless sensor network environment lower node sleep scheduling method, comprises following step:

Step 1: carry out netinit;

Wireless sensor network environment is carried out to initialization, complete deployment, the sub-clustering of nodes, after deployment, obtain node distributed intelligence, after sub-clustering, have a leader cluster node in each bunch, all leader cluster nodes form backbone network;

In setting data polymerization tree, SINK node is numbered 1, the Random assignment when node deployment of the numbering of backbone network node;

Step 2, according to Routing Algorithm, obtain the data aggregate tree of wireless sensor network, obtain route matrix R;

Using the distributed intelligence of backbone network node as input, according to Routing Algorithm, obtain being adapted to the data aggregate tree of wireless sensor network, the summit of tree is SINK node, according to data aggregate tree, obtains route matrix R;

According to route matrix R, obtain father node matrix F N, child node Matrix C N, layer matrix L, fraternal matrix B N, be specially:

Route matrix R is:

wherein

Wherein: N represents the quantity of leader cluster node in N backbone network, r _ijrepresent that leader cluster node i is to the link break-make situation of leader cluster node j;

Father node matrix F N:

The father node numbering fn of leader cluster node i _irepresent, father node matrix is: FN=[fn ₁fn ₂fn _n]; If leader cluster node j is the father node of leader cluster node i, r in route matrix _ij=1, fn _i=j, if leader cluster node i does not have fn of father node _ibe 0;

Child node Matrix C N:

The child node numbering of leader cluster node i is used represent, in the numbering of each child node by incremental order, arrange, child node matrix is do not have position and other vacant positions of child node all to mend 0; According to route matrix R, if leader cluster node j is the child node of leader cluster node i, there is r _ij=-1,

Layer matrix L:

The number of plies L of leader cluster node i in convergence tree _irepresent, i.e. L=[L ₁l ₂l _n]; Wherein, setting sink node serial number is 1, is positioned at ground floor, i.e. L ₁=1; According to child node matrix, obtain the child node of each leader cluster node, if the number of plies of a leader cluster node is l, the number of plies of its child node is l+1, by that analogy, obtains the number of plies of all nodes;

Brotgher of node matrix B N:

The brotgher of node numbering of leader cluster node i is used represent, in the numbering of each brotgher of node by incremental order, arrange, do not have position and other vacant positions of the brotgher of node all to mend 0; According to father node matrix and child node matrix, each leader cluster node obtains all child node numberings that have identical father node with this node, is the brotgher of node numbering of this node;

Step 3, determine the periodic duty method of leader cluster node, build the cross-layer optimization model based on energy consumption, obtain the sleep scheduling method of each leader cluster node;

(1) determine the periodic duty method of leader cluster node;

The operation method of each leader cluster node is:

1, leader cluster node, in sleep state, when leader cluster node needs transceiving data, switches to operating state by leader cluster node by sleep state;

The state conversion time of leader cluster node is: father node is than the late Δ T that wakes up of first child node, and when guaranteeing that father node can start to receive son node number certificate, this child node has been ready to send data; Between the brotgher of node, child node below will guarantee to complete and send after data in previous child node successively than the late Δ t that wakes up of previous child node, and this child node is ready to send data to father node just;

2, leader cluster node collect bunch in data;

Leader cluster node is collected the perception data of bunch interior nodes in one-period; Bunch interior nodes after corresponding leader cluster node is started working, by the data upload of perception in one-period to leader cluster node, by leader cluster node via the data aggregate tree of backbone network by convergence to SINK node;

If 3 leader cluster nodes have child node, leader cluster node is collected son node number certificate;

In backbone network, leader cluster node is collected the data of its child node in data aggregate tree;

4, leader cluster node sends data to father node;

After backbone network leader cluster node collection data are complete, through data aggregate, the data upload that oneself is stored is to father node, before child node sends data to father node, to carry out synchronously with father node, periodically to father node, send synchronization request bag, obtain father node and confirm after bag, then start data transmission; After data are sent, child node need to be waited for the feedback packet that father node sends, and judges whether data send successfully, if this secondary data sends unsuccessfully, storage data, wait for that leader cluster node preferentially resends this secondary data next time while waking up;

5, leader cluster node enters sleep;

Backbone network leader cluster node completes after the collection and transmission work of data, and node is switched to sleep state;

(2) build the cross-layer optimization model based on energy consumption;

1, get parms:

Packet loss matrix P is:

work as r _ij=0 o'clock, p _ij=0

Wherein: p _ijrepresent that leader cluster node i is to the packet loss on the link of leader cluster node j;

Data transmission rate matrix F is:

work as r _ij=0 o'clock, f _ij=0

Wherein: f _ijrepresent that leader cluster node i is to the data transmission rate on the link of leader cluster node j;

Energy consumption matrix E is:

$E={\left[E_{\mathrm{ij}}\right]}_{N\&times;N}=\left[\begin{array}{cccccc}{E}_{\&CenterDot;1}& E_{\&CenterDot;2}& E_{\&CenterDot;3}& E_{\&CenterDot;4}& E_{\&CenterDot;5}& E_{\&CenterDot;6}\end{array}\right]=\left[\begin{array}{cccccc}{E}_{11}& E_{12}& E_{13}& E_{14}& E_{15}& E_{16}\\ E_{21}& E_{22}& E_{23}& E_{24}& E_{25}& E_{26}\\ .& .& .& .& .& .\\ .& .& .& .& .& .\\ .& .& .& .& .& .\\ E_{N1}& E_{N2}& E_{N3}& E_{N4}& E_{N5}& E_{N6}\end{array}\right]$

Wherein: i represents leader cluster node numbering, j represents the energy properties numbering of node, E _i1the state conversion energy consumption that represents leader cluster node i; E _i2the energy consumption that represents data in the receipts bunch of leader cluster node i; E _i3the energy consumption that represents the receipts son node number certificate of leader cluster node i; E _i4the energy consumption that represents the transmission data of leader cluster node i; E _i5the sleep energy consumption that represents leader cluster node i; E _i6the data processing energy consumption that represents leader cluster node i;

The energy consumption of i node is the total energy consumption of all nodes is

Sleep scheduling matrix T:

$T={\left[t_{\mathrm{ij}}\right]}_{N\&times;4}=\left[\begin{array}{cccc}{t}_{\&CenterDot;1}& t_{\&CenterDot;2}& t_{\&CenterDot;3}& t_{\&CenterDot;4}\end{array}\right]=\left[\begin{array}{cccc}{t}_{11}& t_{12}& t_{13}& t_{14}\\ t_{21}& t_{22}& t_{23}& t_{24}\\ .& .& .& .\\ .& .& .& .\\ .& .& .& .\\ t_{N1}& t_{N2}& t_{N3}& t_{N4}\end{array}\right]$

Wherein: i represents leader cluster node numbering, j represents the time attribute numbering of node, t _i1the time that wake up the next time of expression leader cluster node i; t _i2the time of data in receiving the next time that represents leader cluster node i bunch; t _i3the time of receiving son node number certificate next time that represents leader cluster node i; t _i4the time that sends data next time that represents leader cluster node i;

Data aggregate rate matrix AG:

The data aggregate rate of leader cluster node i is ag _i, i.e. AG=[ag ₁ag ₂ag _n];

Basic power consumption:

State is changed primary energy consumption into E _change, the energy consumption of carrying out a data aggregate is E _agg, sending data power consumption is P _send, receiving data power consumption is P _rec, idle listening power consumption is P _idle, sleep power consumption is P _sleep;

Feedback matrix FB:

The last feedback fb that sends data of leader cluster node i _irepresent, when needs resend data, fb _ibe recorded as the data volume that need to resend; Fb during initialization _ibe 0;

FB=[fb ₁fb ₂fb ₃fb _n], wherein

The packet loss that sends data to father node due to node i is therefore fb _i=S _{send, i}probability be s _{send, i}represent this data volume that need to resend of node i; When the data of node i send successfully, node i will be received feedback information ACK, fb _i=0; When the data of node i send unsuccessfully, node i will receive feedback information NAK or can not receive feedback information, fb _i=S _{send, i}, wherein, S _{send, i}represent this data volume that need to resend of node i;

Set other parameters:

T _{brc_i}represent that leader cluster node i can start to receive the moment of son node number certificate;

T _{bs_i}=t _i1+ t _i2+ t _i3represent that leader cluster node i can start to send to father node the moment of data;

S _sYNthe size that represents SYN bag;

S _aCKthe size that represents ACK bag;

S _fBthe size that represents FB bag;

2. build the cross-layer optimization model based on energy consumption

The calculating of each several part energy consumption:

For leader cluster node i, there is E _i=E _i1+ E _i2+ E _i3+ E _i4+ E _i5+ E _i6, represent the energy consumption of node i in one-period, be specifically calculated as follows;

(1) E _i1: state conversion energy consumption

If according to planning before, node i is waken up once in one-period, therefore E _i1=E _change;

(2) E _i2: the energy consumption of data in receiving bunch

Backbone node i just represents in network i bunch, in establishing bunch, has K _iindividual node, the data arrival rate of each node is λ _ik, in bunch, the network cycle of operation is T _{i_ch}, in bunch, be divided into G _iindividual group, every group of operating time is T _{i_ch}/ G _i, the work period T of leader cluster node i namely _i=T _{i_ch}/ G _i;

Node i within a work period need to bunch in the data volume that gathers be if the data transmission rate in bunch during data upload is μ _i, node i receive bunch in time of data be

Therefore: $E_{i2}=P_{\mathrm{rec}}\&CenterDot;t_{i2}=P_{\mathrm{rec}}\&CenterDot;\underset{k=1}{\overset{K_i}{\mathrm{\&Sigma;}}}{\mathrm{\&lambda;}}_{\mathrm{ik}}\&CenterDot;\frac{T_{i\_\mathrm{ch}}}{G_i{\mathrm{\&mu;}}_i};$

(3) E _i3: the energy consumption of receiving son node number certificate

If node i has M _iindividual child node, according to node periodic duty method, obtains the energy consumption that node i receives m child node and is: E _{i3, m}=E _{wait_SYN, m}+ E _{rec_SYN, m}+ E _{aCK, m}+ E _{receive, m}+ E _{fB, m};

Wherein, E _{wait_SYN, m}for waiting for synchronous energy consumption, E _{rec_SYN, m}for receiving the energy consumption of sync packet, E _{aCK, m}for sending the energy consumption of ACK, E _{receive, m}for receiving the energy consumption of data, E _fBfor sending the energy consumption of feedback; Specifically be calculated as follows:

Wait for synchronous energy consumption: E _{wait_SYN, m}=P _idlet _{wait_SYN, m}

Wherein, $t_{\mathrm{wait}\_\mathrm{SYN},m}=\left\{\begin{array}{cc}{t}_{\mathrm{bs}\_{\mathrm{cn}}_{\mathrm{im}}}-t_{\mathrm{brc}\_i,m}& t_{\mathrm{bs}\_{\mathrm{cn}}_{\mathrm{im}}}>t_{\mathrm{brc}\_i,m}\\ t_{\mathrm{bs}\_{\mathrm{cn}}_{\mathrm{im}}}+S_m(t_{\mathrm{SYN}}\_t_{\mathrm{inter}\_\mathrm{SYN}})-t_{\mathrm{brc}\_i,m}& t_{\mathrm{bs}\_{\mathrm{cn}}_{\mathrm{im}}}<t_{\mathrm{brc}\_i,m}\end{array}\right.$ For waiting for the synchronous time;

Wherein, represent that m child node of node i is ready to send to father node the moment of data, the 1st, 2,3 time attribute that represent respectively m child node of node i, t _sYNthe time that represents node sending/receiving synchronization request bag, t _{inter_SYN}represent to send request for twice the time interval of bag;

represent that node i is ready to receive the moment of m sub-node data, wherein, t _aCKthe time that represents node sending/receiving confirmation bag, t _{rec, m-1}represent that m-1 child node of node i receives the time of data, t _fBthe time that represents node sending/receiving feedback packet;

m the child node that represents node i carried out the quantity of replying the sync packet of wrapping of not receiving of transmission when synchronous;

Receive the energy consumption of sync packet: wherein, m the child node that represents leader cluster node i is to the data transmission rate on the link of leader cluster node i;

Send the energy consumption of confirming bag: wherein, represent that leader cluster node i is to the data transmission rate on the link of m the child node of leader cluster node i;

Receive the energy consumption of data:

With $S_{\mathrm{receive}}=S_{\mathrm{send},{\mathrm{cn}}_{\mathrm{im}}}$ Represent the data package size receiving, $E_{\mathrm{receive},m}=P_{\mathrm{rec}}\&CenterDot;\frac{S_{\mathrm{send},{\mathrm{cn}}_{\mathrm{im}}}}{f_{{\mathrm{cn}}_{\mathrm{im}},i}};$

Send the energy consumption of feedback: $E_{\mathrm{FB}}=P_{\mathrm{send}}\&CenterDot;\frac{S_{\mathrm{FB}}}{f_{i,{\mathrm{cn}}_{\mathrm{im}}}};$

(4) E _i4: the energy consumption that sends data

According to node periodic duty method, have:

E _i4＝E _SYN+E _{inter_SYN}+E _{rec_ACK}+E _send+E _{rec_FB}

Wherein, E _sYNfor sending the energy consumption of synchronization request bag, E _{inter_SYN}for synchronization request interval energy consumption, E _{rec_ACK}for the energy consumption of confirmation of receipt bag, E _sendfor sending the energy consumption of data, E _{rec_FB}for receiving the energy consumption of feedback packet; Specifically be calculated as follows:

The father node of node i is fn _i, i is fn _ic child node, the moment that node i is ready for sending data is t _{bs_i}, the moment that father node is prepared its data of reception is have:

Send the energy consumption of synchronization request bag: E _sYN=P _send(s _i+ 1) t _sYN

Wherein, represent to send the time of synchronization request bag, the data transmission rate of the link of expression from leader cluster node i to father node, the quantity of the sync packet that does not obtain reply sending in expression node i request synchronizing process;

Synchronization request interval energy consumption: E _{inter_SYN}=P _idles _it _{inter_SYN};

The energy consumption of confirmation of receipt bag: E _aCK=P _rect _aCK, wherein

Send the energy consumption of data:

The data that send comprise two parts, the data that gather in bunch and the data of collecting from child node; If it is unsuccessful that the last data of node send, need the last data of preferential transmission, then send the data of this collection;

The data that gather in bunch and the data of collecting from child node need to send after data aggregate again, establish after data aggregate, and the data volume that node i need to send to father node is $S_{\mathrm{send},i}={\mathrm{fb}}_i+{\mathrm{ag}}_i(\underset{k=1}{\overset{K_i}{\mathrm{\&Sigma;}}}{\mathrm{\&lambda;}}_{\mathrm{ik}}\&CenterDot;T_i+\underset{m=1}{\overset{M_i}{\mathrm{\&Sigma;}}}{S}_{\mathrm{send},{\mathrm{cn}}_{\mathrm{im}}});$

The energy consumption that sends data is: $E_{\mathrm{send}}=P_{\mathrm{send}}\&CenterDot;t_{\mathrm{send},i}=P_{\mathrm{send}}\&CenterDot;\frac{S_{\mathrm{send},i}}{f_{i,f{n}_i}};$

Receive the energy consumption of feedback: E _fB=P _rect _fB, wherein

(5) E _i5: sleep energy consumption

Each node is on carrying out once during sleep scheduling method, and current sleep scheduling matrix is T, and sleep scheduling matrix is next time T ', completes before sleep scheduling method enters sleep state, makes T=T ', and T ' empties;

Node i is the length of one's sleep in next cycle: t _sleep=t _i1'-(t _i1+ t _i2+ t _i3+ t _i4);

The sleep energy consumption of node i in next cycle is: E _i5=P _sleept _sleep;

(6) E _i6: data processing energy consumption

Data processing energy consumption refers to that node carries out the energy loss in data aggregate process, and each node carries out data aggregate one time in one-period, therefore E _i6=E _agg;

To sum up, obtaining the total energy consumption of network in one-period is:

$\begin{array}{c}{E}_{\mathrm{total}}=\underset{i=1}{\overset{N}{\mathrm{\&Sigma;}}}\underset{j=1}{\overset{6}{\mathrm{\&Sigma;}}}{E}_{\mathrm{ij}}\\ =\underset{i=1}{\overset{N}{\mathrm{\&Sigma;}}}(E_{i1}+E_{i2}+E_{i3}+E_{i4}+E_{i5}+E_{i6})\\ =\underset{i=1}{\overset{N}{\mathrm{\&Sigma;}}}[E_{\mathrm{change}}+E_{\mathrm{agg}}+P_{\mathrm{rec}}\&CenterDot;\underset{k=1}{\overset{K_i}{\mathrm{\&Sigma;}}}{\mathrm{\&lambda;}}_{\mathrm{ik}}\&CenterDot;\frac{T_{i\_\mathrm{ch}}}{G_i{\mathrm{\&mu;}}_i}+\underset{m=1}{\overset{M_i}{\mathrm{\&Sigma;}}}(P_{\mathrm{idle}}\&CenterDot;t_{\mathrm{wait}\_\mathrm{SYN},m}+P_{\mathrm{rec}}\&CenterDot;\frac{S_{\mathrm{SYN}}}{f_{{\mathrm{cn}}_{\mathrm{im}},i}}+P_{\mathrm{send}}\&CenterDot;\frac{S_{\mathrm{ACK}}}{f_{i,c{n}_{\mathrm{im}}}}\\ +P_{\mathrm{rec}}\&CenterDot;\frac{S_{\mathrm{send},{\mathrm{cn}}_{\mathrm{im}}}}{f_{c{n}_{\mathrm{im}},i}}+P_{\mathrm{send}}\&CenterDot;\frac{S_{\mathrm{FB}}}{f_{i,{\mathrm{cn}}_{\mathrm{im}}}})+(P_{\mathrm{send}}\&CenterDot;(s_i+1)t_{\mathrm{SYN}}+P_{\mathrm{idle}}\&CenterDot;s_i\&CenterDot;t_{\mathrm{inter}\_\mathrm{SYN}}+P_{\mathrm{rec}}\&CenterDot;\frac{S_{\mathrm{ACK}}}{f_{{\mathrm{fn}}_i,i}})\\ +P_{\mathrm{send}}\&CenterDot;\frac{{\mathrm{fb}}_i+{\mathrm{ag}}_i(\underset{k=1}{\overset{K_i}{\mathrm{\&Sigma;}}}{\mathrm{\&lambda;}}_{\mathrm{ik}}\&CenterDot;T_i+\underset{m=1}{\overset{M_i}{\mathrm{\&Sigma;}}}{S}_{\mathrm{send},{\mathrm{cn}}_{\mathrm{im}}})}{f_{i,{\mathrm{fn}}_i}}+P_{\mathrm{rec}}\&CenterDot;\frac{S_{\mathrm{FB}}}{f_{{\mathrm{fn}}_i,i}})+P_{\mathrm{sleep}}\&CenterDot;t_{\mathrm{sleep}}]\\ =\underset{i=1}{\overset{N}{\mathrm{\&Sigma;}}}[E_{\mathrm{change}}+E_{\mathrm{agg}}+P_{\mathrm{rec}}(\underset{k=1}{\overset{K_i}{\mathrm{\&Sigma;}}}{\mathrm{\&lambda;}}_{\mathrm{ik}}\&CenterDot;\frac{T_{i\_\mathrm{ch}}}{G_i{\mathrm{\&mu;}}_i}+\underset{m=1}{\overset{M_i}{\mathrm{\&Sigma;}}}(\frac{S_{\mathrm{SYN}}}{f_{{\mathrm{cn}}_{\mathrm{im}},i}}+\frac{S_{\mathrm{send},{\mathrm{cn}}_{\mathrm{im}}}}{f_{{\mathrm{cn}}_{\mathrm{im}},i}})+\frac{S_{\mathrm{ACK}}}{f_{{\mathrm{fn}}_i,i}}+\frac{S_{\mathrm{FB}}}{f_{{\mathrm{fn}}_i,i}})\\ +P_{\mathrm{send}}(\underset{m=1}{\overset{M_i}{\mathrm{\&Sigma;}}}(\frac{S_{\mathrm{ACK}}}{f_{i,c{n}_{\mathrm{im}}}}+\frac{S_{\mathrm{FB}}}{f_{i,{\mathrm{cn}}_{\mathrm{im}}}})+(s_i+1)\frac{S_{\mathrm{SYN}}}{f_{i,{\mathrm{fn}}_i}}+\frac{{\mathrm{fb}}_i+{\mathrm{ag}}_i(\underset{k=1}{\overset{K_i}{\mathrm{\&Sigma;}}}{\mathrm{\&lambda;}}_{\mathrm{ik}}\&CenterDot;T_i+\underset{m=1}{\overset{M_i}{\mathrm{\&Sigma;}}}{S}_{\mathrm{send},{\mathrm{cn}}_{\mathrm{im}}})}{f_{i,{\mathrm{fn}}_i}})\\ +P_{\mathrm{idle}}(\underset{m=1}{\overset{M_i}{\mathrm{\&Sigma;}}}{t}_{\mathrm{wait}\_\mathrm{SYN},m}+s_i\&CenterDot;t_{\mathrm{inter}\_\mathrm{SYN}})+P_{\mathrm{sleep}}\&CenterDot;t_{\mathrm{sleep}}]\end{array}$

The model that can be optimized is:

min?E _total

$s.t.|t_{i1}-(t_{c{n}_{i1}}+\mathrm{\&Delta;T})|\&le;\mathrm{\&delta;}$

$|t_{i1}-(t_{{\mathrm{bn}}_{i1}}+\mathrm{\&Delta;t})|\&le;\mathrm{\&delta;};$

0＜ag _i＜1(0＜i≤N)

$M_i\stackrel{\&OverBar;}{{\mathrm{\&lambda;}}_i}<f_{i,{\mathrm{fn}}_i}$

In constraints, represent waking up the time difference constantly of father node and its first child node, the time difference of waking the moment up that represents leader cluster node and its first brotgher of node, δ < < 1,0 < ag _i< 1 (0 < i≤N) represents the scope of the data aggregate rate of leader cluster node i, the data arrival rate that represents leader cluster node i is less than this node to the data transmission rate of father node transmission data, wherein, and M _ibunch interior nodes quantity that represents leader cluster node i place bunch, bunch interior nodes average data arrival rate that represents leader cluster node i place bunch, represent that leader cluster node i is to the data transmission rate of the link of father node;

According to constraints, solve E _totalminimum value, obtains sleep scheduling matrix T, by sleep scheduling matrix T determine receive time of waking up the next time of each leader cluster node, next time bunch in time, time of receiving son node number certificate next time of data, send the time of data next time;

(3) obtain each leader cluster node sleep scheduling method first;

By the cross-layer optimization model based on energy consumption obtaining in step (2), determined the time in the periodic duty method of leader cluster node in step (1), obtained the sleep scheduling method first of each node, each leader cluster node sleep scheduling method is first completed by SINK node, and informs successively other leader cluster nodes in wireless sensor network;

Step 4, wireless sensor network are according to node sleep dispatching method, and each node brings into operation according to the sleep scheduling method of this node, to father node uploading data;

Step 5, when each node is after father node sends data, before entering sleep state, according to current wireless sensor network state, obtain the input of the cross-layer optimization model based on energy consumption, adopt the cross-layer optimization model based on energy consumption that in step 3, (2) build, obtain this node node sleep dispatching method next time, and the method is informed to father node and the brotgher of node of this node;

Step 6, wireless sensor network judge whether the energy of leader cluster node exhausts, and if so, proceeds to step 7, otherwise, return to step 4;

If there is leader cluster node depleted of energy in step 7 backbone network, adopt Routing Algorithm, whether the backbone network node that judges current residual can rebuild data aggregate tree, if can rebuild data aggregate tree, return to step 2, otherwise backbone network depleted of energy, wireless sensor network stops operation.