CN101282290A - Method for balancing and clustering load of wireless sensing network based on gradient - Google Patents

Abstract

The invention discloses a distributed clustering method of load balancing based on gradient of a wireless sensor network, for controlling distribution of cluster heads according to hop of node distance SINK (e.g., node gradient); the cluster head closer to SINK is more than that far away from SINK, so as to obtain better inter cluster load balance; the cluster with the same gradient is provided with the same cluster scale, so as to balance load in cluster; the method establishing gradient of the network through flooding broadcast of SINK, without node position information, with distributed selection of the cluster head, which can balance network traffic and energy well and prolongs network life time.

Description

Method for balancing and clustering load of wireless sensing network based on gradient

Technical field

The invention provides a kind of topological management method of sub-clustering of wireless sense network, particularly a kind of wireless sensor network method for balancing and clustering load based on gradient.

Background technology

(wireless sensor network WSN) is current emerging forward position research hot fields that receive much concern in the world, that intersected by multidisciplinary height to wireless sensor network.Wireless sensor network has very wide application prospect, in military and national defense, industrial or agricultural, city management, biologic medical, environmental monitoring, rescue and relief work, anti-probably anti-terrorism, many fields such as deathtrap Long-distance Control all have important scientific research to be worth and huge practical value, world many countries military circles have been caused, the great attention of academia and industrial quarters, and become the hot research field, emerging forward position of generally acknowledging that entered since 2000, be considered to one of the technology of power that will have an immense impact on to 21st century, it will exert far reaching influence to the human following life style.

In the research of wireless sensor network, the life span (lifetime) that how to prolong network is one of sixty-four dollar question.Assessment to the life span of wireless sensor network generally has: the ratio of node failure reaches three kinds of p, the non-connections of network etc. in the time of first node failure, the network.No matter be any assessment, all require researcher's network configuration reasonable in design and routing policy, the load of balancing network makes the energy consumption of each node reach unanimity, thus the life span of maximization network.

For the device radio sensing network that has great deal of nodes, the topological structure of sub-clustering has more advantage on network management and extensibility.In this network configuration, (data of collection are sent to the SINK node by the multi-hop routing mode between bunch head to bunch head for Cluster Head, CH) data collection and the work of treatment in bearing bunch.In the network configuration of this " many-one ", the node load of nearly SINK is higher than SINK node far away far away, may cause serious network congestion, simultaneously, high capacity means high energy consumption, promptly have so-called " hot-zone (hotspot) ", node will run out of energy too early and cause the network division, thereby reduce the life span of whole network.

Load balancing is one of focus of wireless sense network research.The researcher has proposed some solutions at this problem, by changing the load balancing that routing policy and network configuration reach network.

People such as Shah propose a kind of multipath routing policy of energy perception, the node route of selecting high dump energy when transfer of data is with balanced load, people such as Braginsky propose the routing algorithm based on Random Walk, randomization by route comes the balancing network load, and is comparatively effective under many SINK, query requests is many, network event is few situation.But, only be difficult to solve " hot-zone " problem of nearly SINK by routing algorithm.

On the other hand, the researcher sets about seeking load balance from the network configuration aspect that makes up stratification, its main thought is based on one or more network informations, try hard to make the scale that algorithm generates bunch more even, and come the balancing network load by bunch head of rotating at regular intervals, as LEACH, HEED, WCA etc., it mainly focuses on the equilibrium of bunch internal burden, to bunch between balanced less concern, and unresolved " hot-zone " problem.

In recent years, the researcher begins to seek the load balance of network by constructing clustering architecture heterogeneous.People such as Sora propose to adopt non-homogeneous sub-clustering to come balanced load first, the hypothesis network is two-layer donut in the literary composition, the energy consumption of communicating by letter in reducing bunch by the number of members that reduces internal layer bunch, with data forwarding task between more bearing bunch, their research be that heterogeneous network and leader cluster node position are default.It is non-homogeneous cluster structured that people such as Shu utilize power to control to make up, and it is more little to penetrate power the closer to bunch hair of SINK, comes a balance bunch energy consumption with this, and this algorithm is the centralized processing algorithm, requires node location known, and the network link that generates is asymmetric.

Summary of the invention

The purpose of this invention is to provide a kind of wireless sensor network method for balancing and clustering load based on gradient.

The problem of the present invention is directed to that existing network flow and energy balance are poor, network lifetime is short, the wireless sensor network method for balancing and clustering load based on gradient that has proposed comprises the steps:

Step 101: initiate gradient network by convergence node (SINK) and set up process.SINK radio network gradient is set up message, and diffuses to the whole network (as: by the mode of flooding);

Step 102: each node obtains the Grad of oneself in the network.Node is set up message by the gradient network that receives the neighbor node forwarding, can extrapolate the gradient of self more accurately.

Step 103: each node is provided with the power of communications of competition cluster.Node is according to self gradient, the compute cluster radius, and corresponding power of communications is set;

Step 104: nodal information is mutual.Node is mutual Hello information under the power of communications that step 103 is provided with, and comprises bunch required information such as energy, the node degree etc. of head competition.

Step 105: node distributed contention cluster.Each node is competed bunch head and cluster under the information that step 104 is known.

The present invention controls the quantity of bunch head according to the jumping figure (that is: node gradient) of nodal distance SINK, bunch quantity height near SINK, bunch quantity away from SINK is low, load balance between obtaining preferably bunch with this, simultaneously, the cocooning tool that gradient is identical has identical bunch scale, comes balanced bunch internal burden with this; Algorithm is set up the gradient of network by the flooded broadcast of SINK, so need not node location information, and the selection of bunch head is distributed carrying out.This method has been done balance preferably to the flow and the energy of network, has prolonged network lifetime.

Description of drawings

Fig. 1 is a flow chart of the present invention.

Fig. 2 is the flow chart of the distributed contention cluster of the inventive method.

Fig. 3 is a schematic network structure of the present invention.

Specific implementation method

The present invention is further illustrated below in conjunction with drawings and Examples.

Performing step of the present invention is:

Step 101: initiate gradient network by convergence node (SINK) and set up process.SINK radio network gradient is set up message, and diffuses to the whole network.Provide a kind of among the present invention based on the information diffusion method that floods.The SINK node with bunch head bunch between power of communications radio network gradient set up message MSG_Grade_Construct, comprise gradient network territory Grade in this message, initial value is zero; Receive the node of MSG_Grade_Construct message, the value in Grade territory is added in the lump forward; Thus MSG_Grade_Construct message is diffused to the whole network.

Step 102: each node obtains the Grad of oneself in the network.In the process that MSG_Grade_Construct message spreads in network, node may receive the MSG_Grade_Construct message from different neighbor nodes, the value in its Grade territory is not quite similar, can adopt diverse ways to estimate self best Grad, get wherein minimum value in the present invention as the gradient network of node.

Step 103: each node is provided with the power of communications of competition cluster.Node is according to self gradient, and power of communications in corresponding bunch is set.Among the present invention, it is as follows to give that egress evenly distributes, each node data produces bunch radium computing method of seeking NLB under digit rate λ the same terms:

In the network model of example shown in the present, gradient is data forwarding two partially-workings of sending out in the data collection in the CH of i need be responsible for bunch with the CH of gradient i+1, does not consider data fusion, is that the shared ratio of CH is p in the node of i if suppose gradient _i, λ _{Total_i}For gradient is the flow summation of the CH burden of i, have

λ _{total_i}＝π(N ²-(i-1) ²)r ²λρ (1)

If make δ _i=1/p _iFor gradient be i bunch scale, bunch number of gradient i is n _{CH_i}=π (i ²-(i-1) ²) r ²ρ p _i, the data traffic that can get its equalization of incidence is

${\mathrm{\λ}}_i={\mathrm{\λ}}_{\mathrm{total}\_i}/n_{\mathrm{CH}\_i}$

$=\mathrm{\λ\ρ}\·\frac{N^2-{(i-1)}^2}{(2i-1)\mathrm{\ρ}{p}_i}\mathrm{\λ}\·{\mathrm{\δ}}_i\·\frac{N^2-{(i-1)}^2}{(2i-1)}---\left(2\right)$

The energy consumption of node can be divided into the data transmission and receive energy consumption two parts, and is general, and node receives and sends k data, has

ε _recv＝ε _elec·k (3)

ε _send＝ε _elec·k+ε _amp·r ²·k

If E _{CH_i}For gradient is the unit data transmitting-receiving energy consumption of the CH of i, then has

$E_{\mathrm{CH}\_i}=\frac{{\mathrm{\λ}}_i}{k}\·({\mathrm{\ϵ}}_{\mathrm{recv}}+{\mathrm{\ϵ}}_{\mathrm{send}})---\left(4\right)$

$=\mathrm{\λ}\·{\mathrm{\δ}}_i\·\frac{N^2-{(i-1)}^2}{(2i-1)}\·(2{\mathrm{\ϵ}}_{\mathrm{elec}}+{\mathrm{\ϵ}}_{\mathrm{amp}}\·r^2)$

The problem of the load balance of CH can be described as seeks { δ _i, i=1 ..., N}, feasible { E _{CH_i}, i=1,2 ..., the variances sigma minimum of N}, promptly

${\mathrm{\σ}}^2={(\mathrm{\λ}\·({2\mathrm{\ϵ}}_{\mathrm{elec}}+{\mathrm{\ϵ}}_{\mathrm{amp}}\·r^2))}^2$

$\·\underset{i=1}{\overset{N}{\mathrm{\Σ}}}{(\frac{N^2-{(i-1)}^2}{2i-1}\·{\mathrm{\δ}}_i-\frac{1}{N}\underset{j=1}{\overset{N}{\mathrm{\Σ}}}\frac{N^2-{(j-1)}^2}{2j-1}\·{\mathrm{\δ}}_j)}^2---\left(5\right)$

In the algorithm, according to { the δ that obtains _i, i=1 .., N} can calculate corresponding bunch of radius.

Make r _iFor gradient is bunch radius of i, and make node density ρ=1/ (π r ₀ ²), then have

$r_i=\sqrt{{\mathrm{\δ}}_i/\left(\mathrm{\π\ρ}\right)}---\left(6\right)$

Each node is provided with corresponding communication power and gets final product according to the covering radius of formula (6) compute cluster.

Step 104: nodal information is mutual.Node is periodic broadcast Hello message under the power of communications that step 103 is provided with, and comprises node ID, current ENERGY E _i, weights W _i(initial value is 0), present node state information such as (being initially UN), node is according to bunch head competition weights W of the information updating oneself that receives _i

Step 105: node distributed contention cluster.A kind of competition of distributed bunch of head and clustering method based on weights is provided among the present invention, and it is cluster structured preferably with acquisition only need can to change weighting parameter according to different application by the information interaction of neighbor node, and expense is less and adaptability preferably arranged.

Its step as shown in Figure 2, specifically details are as follows:

Step 201: node initializing.There are four kinds of states in node in the cluster stage: uncertain (UN), bunch head (CH), accurate bunch member (PreCM) and bunch member (CM), cluster is finished then only CH and CM two states.In a bunch head competition starting stage, node is uncertain (UN) state.

Step 202: the node weight value information is mutual.This step is intended to seek the minimum node of the local weights of network, and competition becomes leader cluster node, can carry out simultaneously with step 104.

Step 203: election of cluster head.If the weights Wi of node is minimum in its neighbours, puts oneself and be CH, beginning periodic broadcast Beacon frame; If have the CH of weights among the node neighbours, put the own PreCM of being less than self weights; After this step is finished, only there are two kinds of node states of CH and PreCM in the network.

Step 204: cluster.

A) the PreCM node is intercepted channel, receives the Beacon frame of CH node, sets up the CH table, selects the best CH in the CH table, and adds application MSG_JOIN_REQ message to its transmission, and the WaitACK timer is set, and waits for MSG_JOIN_ACK message;

B) CH receives MSG_JOIN_REQ message, if bunch less than, send MSG_JOIN_ACK message and allow node to add, otherwise, send MSG_JOIN_DENY message refusal and add application;

C) the PreCM node is then put oneself for CM if receive the MSG_JOIN_ACK message that CH sends, and finishes clustering process, if receive MSG_JOIN_DENY message or WaitACK timer expiry, then changes over to a), selects other CH;

D) as if no available CH in the CH table of PreCM node, then put oneself and be CH, the Beacon frame of going on the air is finished clustering process.Finally, CH node and CM node are only arranged in the network.

Described above only is a most preferred embodiment of the present invention, and protection scope of the present invention is not limited to above-mentioned specific embodiment, and the conspicuous change that all those skilled in the art are made according to the present invention all drops in protection scope of the present invention.

Claims (1)

1,, it is characterized in that it comprises the steps: based on the wireless sensor network method for balancing and clustering load of gradient

Step 101: initiate gradient network by convergence node (SINK) and set up process, SINK radio network gradient is set up message, and diffuses to the whole network;

Step 102: each node obtains the Grad of oneself in the network.Node is set up message by the gradient network that receives the neighbor node forwarding, can extrapolate the gradient of self more accurately;

Step 103: each node is provided with the power of communications of competition cluster.Node is according to self gradient, the compute cluster radius, and corresponding power of communications is set;

Step 104: nodal information is mutual.Node is mutual Hello information under the power of communications that step 103 is provided with, and comprises bunch required information such as energy, the node degree etc. of head competition;

Step 105: node distributed contention cluster.Each node is competed bunch head and cluster under the information that step 104 is known.

Images