CN102970744B - Wireless sensor network regional locating method based on node density - Google Patents

Abstract

The invention discloses a wireless sensor network regional locating method based on node density. The method includes: boundary detection, regional division and distance calculation. The boundary detection includes that nodes in a network are divided into boundary nodes and non-boundary nodes. The regional division includes a topology discovery stage, a non-boundary node region division stage and a boundary node region division stage, wherein in the topology discovery stage, each node obtains a neighbor information table and a route information table of the node through broadcasting data, and in the boundary node region division stage, the boundary nodes and nodes with the most neighbor nodes in the neighbor non-boundary nodes of the boundary nodes are divided into the same region. The method is applied to large-scale anisotropic networks with uneven distribution of nodes and irregular signal distribution under a complex environment, can be adapt to various topology structures through region division and single hop distance calculation of an independent region, and practicality of an algorithm in a real scene is improved. By means of the method, locating precision is improved, and a locating problem of unknown nodes in the anisotropic networks is solved effectively.

Description

A kind of wireless sensor network subregion localization method based on node density

The present invention relates to wireless sensor network (Wireless Sensor Network, WSN) location technology, be specifically related to a kind of wireless sensor network subregion localization method based on node density.

Background technology

As current state-of-the-art technology for information acquisition, WSN is often used to the state that environmental monitoring, target following etc. are used for reporting that monitored object is current, and state information needs to combine the state that accurately could reflect object under test with positional information usually.User not only needs to know what occurs usually, is also concerned about the position that event occurs simultaneously, therefore, determines that the positional information of node plays key effect to the validity that WSN applies.As in fire hazard monitoring, node detects the generation of fire by data acquisition, but only navigates to the position of fire generation, can arrange measure of effectively suing and labouring.Therefore, node self-localization technology is significant in WSN application.

In a WSN, positional information can also assist other WSN technology, and in Design of Routing Protocol, the Routing Protocol (GEM, GPSR etc.) based on geographical position is widely used; In the design of data anastomosing algorithm, positional information can improve the reliability of transfer of data greatly accurately; On node deployment strategy, positional information can affect the topographic morphologies of network and node to the coverage condition of network; In target following, in the position related application such as monitoring in real time, the importance of positional information is apparent especially.In WSN, location algorithm is divided into according to the different criteria for classifying locate based on range finding and the location of range-independence, Distribution and Centralization, active and Passive Positioning, close-coupled and loose couplings location etc.

The present invention measures the need of to internodal distance according to position fixing process, location algorithm is divided into the location algorithm based on range finding (Range-based) and range-independence (Range-free).Location algorithm based on range finding passes through the positional information of three limits or Maximum Likelihood Estimation Method computing node, this algorithm can provide effective positioning precision, for resource-constrained WSN, location cost is higher, need extra hardware facility or software cost, inapplicable in the scene of extensive location; The location algorithm of range-independence only needs network-in-dialing degree information just can realize distance estimations; determine the position of node; than being more suitable for extensive environment; due to without the need to range finding; its location cost is lower; positioning precision is low compared with the location algorithm based on range finding, faces the challenge of anisotropic network and sparse network simultaneously.

In range-independence location algorithm, DV-Hop algorithm is simple owing to realizing, and positioning precision is relatively high, and is widely used in extensive localizing environment, becomes the focus of current Position Research and concern.The distance of unknown node to beaconing nodes represents by average single-hop distance and the leapfrog number be multiplied by between the two by DV-Hop algorithm, when unknown node obtain with the distance of three or more beaconing nodes after, positioned by trilateration.Its process is divided into three phases:

First stage: locating information is propagated and calculated with minimum hop number, propagated by beaconing nodes log-on data, each beaconing nodes outwards broadcasts positioning starting packet, this packet comprises beaconing nodes own location information and distance leapfrog number, after neighbor node receives this packet, the leapfrog field value in bag is added the neighbours that the new packet of 1 formation is transmitted to it.The source of packet be labeled thus avoid data repeat transmit.Carry out data dissemination in this way, until all nodes of the whole network all obtain position and the minimum hop numerical value of all beaconing nodes, if a node have received multiple packet from same beaconing nodes, then only retain the bag of minimum hop count value, abandon other packets, ensure that this node is minimum to the jumping figure value of beaconing nodes.

Second stage: calculate single-hop distance, this stage carries out between beaconing nodes, each beaconing nodes calculates self average single-hop distance Hopsize according to the following formula _i,

${\mathrm{Hopsize}}_i=\underset{i\≠j}{\mathrm{\Σ}}\sqrt{{(x_i-x_j)}^2+{(y_i-y_j)}^2}/\underset{i\≠j}{\mathrm{\Σ}}{\mathrm{hops}}_{\mathrm{ij}}$

Wherein, (x _i, y _i), (x _j, y _j) be the coordinate of beaconing nodes i and j, hops _ijit is internodal leapfrog number.After calculating completes, the packet comprising average single-hop range information is outwards broadcasted, after neighbor node receives this packet, in this, as calculating the corrected value that oneself arrives this beaconing nodes spacing, if node receives the data of multiple beaconing nodes, then only need to retain distance first single-hop range information received, abandon other packet, ensure that receiving single-hop range information is that nearest beaconing nodes is sent.

Phase III: location, according to three limit range finding or Maximum Likelihood Estimation Method principles, the node utilizing the first two stage to calculate is to the distance of the beaconing nodes of more than 3 or 3, and the coordinate position of computing node self, reaches location object.

DV-Hop algorithm idea is simple, and amount of calculation is little, is easy to realize, but in the WSN application system of reality, it is low that DV-Hop algorithm but exists positioning precision, and location relies on the problems such as network topology.Also there is the improvement of some DV-Hop algorithms at present both at home and abroad, essence due to DV-Hop algorithm is that the distance of nodal distance beaconing nodes is calculated by the product of single-hop distance with leapfrog number, then three limit range finding or maximum likelihood estimate determination node coordinates are adopted, therefore, the improvement of DV-Hop algorithm is carried out from following three aspects.

1. single-hop distance

The people such as Peng Gang calculate for single-hop distance in DV-Hop algorithm, do on average as the single-hop distance of all nodes of the whole network by the single-hop distance calculated each beaconing nodes in network, the distance of unknown node to beaconing nodes is calculated with this, improve the accuracy of distance estimations, reduce the error of location Calculation; The people such as Lin Jinchao on the basis of the above, by analyzing the error of each beaconing nodes, revise average the whole network single-hop distance further; Zheng Jun has just waited people to utilize Cayley-Menger determinant, is optimized the distance estimations of unknown node, reduces range error, improve the positioning precision of node by the geometrical constraint of Euler space.

2. leapfrog number

The people such as Liu Lijun proposes the DV-Hop algorithm of the improvement based on Cluster management, adopts sub-clustering thought, and by network being divided into less bunch, restriction unknown node to the communications hop step number of anchor node, thus improves positioning precision, reduces and locates expense; The people such as Xi W. based on the nearer node of distance beaconing nodes usually than node far away have more before to continue node and this observation conclusion of less descendant node, propose the concept of virtual leapfrog number in position fixing process.By the skip count for each node maintenance two type, one is the real leapfrog based on route, another is the virtual skip count being called as hop by hop, the traditional hop number of virtual leapfrog correction is utilized to carry out, shield the comparatively large and error that causes of the harshness every hop distance difference caused due to wild environment in a certain sense, improve positioning precision.

The people such as Chen Kai propose a kind of BNL (Based on the number of NeighborsLocalization) location algorithm of improvement, network is divided into several monitoring subregions by this algorithm, and then the neighbours' number calculating each sub regions interior joint estimates internodal distance, reduce the impact that Node distribution calculates leapfrog number, improve the accuracy that node locating calculates to a certain extent.

3. displace analysis

Lin Jinchao etc., by weighted mass center algorithm being introduced in the process of node location calculating, revise the ratio raising positioning precision that different beacon participates in node locating, for the problem that the node coordinate error of three limit location Calculation is larger, construct the coordinate figure iterative numerical refining algorithms based on Taylor series expansion method, by choose reasonable iteration threshold value, suitably increase the amount of calculation of location node, improve positioning precision and position error stability.

In addition, also have some innovatory algorithm from the selection of beaconing nodes, the aspects such as topological structure are improved algorithm, conllinear degree concept is introduced in beaconing nodes selection course by the medium people of Liu Ke, by optimizing the topology location relation of position relationship between beaconing nodes and unknown node and beacon, the reliability of raising DV-Hop algorithm in the lower situation of beaconing nodes density and robustness.

Analyze the DV-Hop algorithm of above-mentioned improvement, we find:

Single-hop distance improved method is reached by improvement network average single-hop distance puies forward high-precision object.For single-hop distance, only have when the Node distribution of the whole network presents uniform state, just close to the most of single-hop distance in network.Therefore, this is improved one's methods and is more suitable for the uniform network of Node distribution, and non-homogeneous for Node distribution, irregular earthen ruins protection wireless sensor network is inapplicable, and error correction effect is limited; The angle innovatory algorithm of the leapfrog number of leapfrog number improved method from unknown node to beaconing nodes in essence, is consistent with method one, is equally applicable to the uniform regular network of Node distribution; Displace analysis improved method reduces error from the angle of position calculation, is a kind of error control technique, but does not process the basic reason (non-rectilinear error) that error produces.

Summary of the invention

The defect existed for above-mentioned prior art or deficiency, the object of the invention is to, a kind of wireless sensor network subregion localization method based on node density (A Node Density-basedSub-regional Localization Technology in WSN is provided, hereinafter referred to as NDL), the method is on a large scale, complex environment lower node skewness, the anisotropic network that signal distributions is irregular, calculated by the single-hop distance of Region dividing and isolated area, overcome the defect of the DV-Hop location algorithm of existing improvement, various topological structure can better be adapted to, improve the practicality of algorithm in real scene.Experimental result shows, and adopts the solution of the present invention to improve positioning precision, efficiently solves anisotropic network interior joint skewness, the orientation problem in the irregular situation of signal distributions.

In order to realize above-mentioned technical assignment, the present invention is achieved by the following technical solutions:

A kind of subregion localization method based on node density, the method positions for the unknown node in wireless sensor network, total n node in described wireless sensor network, wherein m beaconing nodes, n-m unknown node, node communication radius is R, and each node has RSSI range capability, and described localization method specifically comprises the steps:

Step 1, Boundary Detection: by Boundary Detection, the node in network is divided into boundary node and non-boundary node two parts;

Step 2, Region dividing: comprise Topology Discovery stage, non-boundary node Region dividing stage and boundary node Region dividing stage, wherein, in the described Topology Discovery stage, each node is obtaining self neighbor information table and route information table by broadcast data; In the described non-boundary node Region dividing stage, each node selects judgement to carry out density area division by the judgement of neighbours' RSSI value and neighbor node number; In the described boundary node Region dividing stage, be the same area by node division maximum for neighbor node number in boundary node and the non-boundary node of its neighbours; Through non-boundary node Region dividing and boundary node Region dividing, whole radio sensing network is divided into the relatively uniform multiple zonules of node density;

Step 3, distance calculate: the zonule do not covered for link between beaconing nodes, select the single-hop distance in the region at the beaconing nodes place nearest from this zonule as its single-hop distance, after obtaining the single-hop distance of each zonule in network, unknown node is according to the single-hop distance of the routing iinformation in its route information table with this zonule, unknown node place, calculate the distance of the beaconing nodes of this unknown node and more than three or three, then least square method is utilized to calculate self position, so far, in wireless network, unknown node is located successfully.

The present invention also comprises following other technologies feature:

The Topology Discovery stage in described step 2 specifically comprises the steps:

Step S1-1: all beaconing nodes (BN in network ₁, BN ₂, BN ₃bN _m) broadcast Preq packet;

Step S1-2: node judges whether to receive Preq packet, if it is performs step S1-3, otherwise performs step S1-7;

Step S1-3: node judges whether it is first time receive beaconing nodes BN _ithe Preq packet sent, if it is performs S1-4, otherwise performs step S1-8;

Step S1-4: node updates self is to the leapfrog number of beaconing nodes, and by count+1, the Preq packet outwards after this renewal of broadcast, performs step S1-5;

Step S1-5: the node reverts back Pack packet receiving Preq packet, to the node sending Preq packet, acknowledges receipt of message;

Step S1-6: after the node sending Preq packet receives Pack packet, upgrade oneself neighbor information table and route information table;

Step S1-7: this node is considered to bad node, does not participate in position fixing process;

Step S1-8: node judges beaconing nodes BN in current routing information table _icount value whether be less than the count value of new data packets, if it is perform step S1-9, otherwise perform step S1-4;

Step S1-9: abandon this packet.

The structure of described neighbor information table: comprise the RSSI value Neighbor_RSSI between the id Neighbor_id of the neighbor node of node and this node and neighbor node.

The structure of described route information table: comprise the id Beacon_id of beaconing nodes, the coordinate coordinate of beaconing nodes, the sequence node route [N] of process on path between the leapfrog information count of this beaconing nodes of nodal distance and node to beaconing nodes.

The described non-boundary node Region dividing stage specifically comprises the following steps:

Step S2-1: all beaconing nodes (BN in network ₁, BN ₂, BN ₃bN _m), with own node numbering formation zone numbering area_id, and in its neighbor information table all non-boundary node sending zones partition request Pdiv packet, self zone division symbolizing flag is set to 1 simultaneously;

Step S2-2: node i receives (i < n, j < n, i ≠ j) after the Pdiv request data package that node j sends, and carries out neighbours' RSSI value and selects to judge: if r by formula (1) _ij=0, then node i performs step S2-3, otherwise performs step S2-4;

$g_i=\frac{r_{\mathrm{ij}}-E\left(R_i\right)}{E\left(R_i\right)}---\left(1\right)$

Wherein, $r_{\mathrm{ij}}=\left\{\begin{array}{cc}1& g_i\&le;t_r\\ 0& g_i>t_r\end{array}\right.,$ T _rthe threshold value of node i neighbours RSSI value, t _rget (0.02-0.04), E (R _i) represent the average of node i all neighbours' RSSI value,

Step S2-3: deleted from the neighbor node set of node i by node j, meanwhile, by the neighbours of a node i numerical value n _isubtract 1 as the new neighbours' numerical value of node i, obtain a node i neighbours number vector N upgrade after vectorial N '=[n ₁', n ' ₂..., n ' _n], perform step S2-5;

Step S2-4: keep the neighbor node set of node i constant, perform step S2-5;

Step S2-5: node i is carried out neighbor node number by formula (2) and selected to judge, if formula (2) is set up, then performs step S2-6, otherwise performs step S2-7;

|n′ _i-n′ _j|≤t _n(2)，

Wherein, t _nfor threshold value in Pdiv packet, t _nget (2-5);

Step S2-6: node i judges oneself to belong to the region of the zone number in Pdiv request data package as area_id, and node i and node j are divided in same density area, and node i is by the flag flag set 1 of oneself;

Step S2-7: node i judges self not belong to the region of the zone number in Pdiv request data package as area_id, using the node_id of oneself as new area_id, in self neighbor table, non-boundary node sends Pdiv packet, thus the same with beaconing nodes, initiate the Region dividing centered by oneself.

The concrete steps that the distance of described step 3 calculates are as follows:

Note zonule A _iin single-hop distance be if the link in network between beaconing nodes is m bar, actual distance corresponding on every bar link is the zonule A of link process between statistics beaconing nodes _iwith the leapfrog number hops in each zonule _i, set up formula:

$\left\{\begin{array}{c}{\tilde{d}}_1\&times;{\mathrm{hops}}_1+{\tilde{d}}_2\&times;{\mathrm{hops}}_2+...+{\tilde{d}}_i\&times;{\mathrm{hops}}_i=d_l^1\\ .\\ .\\ .\\ {\tilde{d}}_1\&times;{\mathrm{hops}}_1+{\tilde{d}}_2\&times;{\mathrm{hops}}_2+...+{\tilde{d}}_j\&times;{\mathrm{hops}}_j=d_l^m\end{array}\right.$

Optimal method is utilized to carry out each zonule d _isolve, for the zonule that link between beaconing nodes does not cover, select the single-hop distance in the region at the beaconing nodes place nearest from this zonule as its single-hop distance, after obtaining the single-hop distance of each zonule in network, unknown node, according to the single-hop distance of the routing iinformation in its route information table with this zonule, unknown node place, calculates this unknown node and beaconing nodes BN by following formula _idistance dist _i:

${\mathrm{dist}}_i=\underset{t=1}{\overset{k}{\mathrm{\&Sigma;}}}({\tilde{d}}_t\&times;{\mathrm{hops}}_t)$

Unknown node utilizes least square method to calculate self position after calculating the beaconing nodes distance of itself and more than three or three.

Compared with prior art, the present invention has the following advantages:

1) improve location algorithm precision

Invention introduces location, subregion thought, single-hop distance is calculated by independent in regional, improve DV-Hop location algorithm in anisotropic network environment, due to the position error that fixing single-hop distance is brought, effectively improve the positioning performance of existing DV-Hop algorithm in anisotropic network.Solve anisotropic network interior joint skewness, the orientation problem in the irregular situation of signal distributions.

2) there is good extensibility

The present invention passes through Region dividing, the anisotropy of the whole network is converted into the isotropism network in regional, solve DV-Hop algorithm (anisotropic network) immalleable problem in complex network environment, location algorithm is reduced the dependence of network topology, the application in various true environment can be applicable to.

Accompanying drawing explanation

Fig. 1 is the flow chart of NDL algorithm of the present invention.

Fig. 2 is NDL algorithm Topology Discovery phase flow figure of the present invention.

Fig. 3 is NDL algorithm of the present invention non-boundary node Region dividing phase flow figure.

Fig. 4-Figure 13 is NDL algorithm of the present invention and the performance comparison result of DV-hop algorithm under different application embodiment.Wherein:

Fig. 4 is linear network network for location.

Fig. 5 is rectangular net network for location.

Fig. 6 is linear network position error average figure.

Fig. 7 is rectangular net position error average figure.

Fig. 8 is emulation rectangular net topological diagram.

Fig. 9 is emulation C-network topological diagram.

Figure 10 is communication range variation diagram.

Figure 11 is beaconing nodes number variation diagram.

Figure 12 is range error figure.

Figure 13 is position error figure.

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

Embodiment

In wireless sensor network, location refers to the sensor node utilized in network, by mutually cooperating between part Given information and node, obtains node self physical location in a network with certain algorithm.In network node network pockety, traditional location algorithm can not be suitable for.Such as, in earthen ruins protection application, in order to resolved collection data are defect information of which position, ruins, need to position deployment node.Meanwhile, because ruins degree of disease is different, needs the situation for cultural sight to have the deployment node of emphasis, so just cause the non-uniform phenomenon that WSN network node distributes, thus form anisotropy sensor network.Traditional DV-Hop algorithm, because single-hop is apart from fixing, thus increases in the position error for the anisotropic network in earthen ruins, can not be suitable for.

In order to address this problem, the present invention introduces node density concept, by neighbours' number of defined node as Region dividing standard, the whole network anisotropic network is divided into the isotropism network of regional, calculate the single-hop distance of regional respectively, thus reduce the error accumulation of DV-Hop algorithm under the whole network single-hop distance is fixing, effectively improve positioning precision.Meanwhile, algorithm, by Region dividing, reduces network size, has fine autgmentability.

We introduce the term and noun that relate in some the present invention below:

Beaconing nodes (Beacon Nodes): at the node of netinit stage clear and definite self-position, other nodes auxiliary as a reference in location.Beaconing nodes occupies little ratio usually within network nodes, can help node determination self-position, become beaconing nodes by artificial deployment or GPS positioning equipment.

Unknown node (Unknown Nodes): the node except beacon.Mostly by random placement in a network, the position of oneself cannot be obtained, need to utilize beaconing nodes to calculate the positional information of oneself by position fixing process.Node self-localization is exactly the process that unknown node carries out positional information calculation.

Neighbor node (Neighbor Nodes): the set that within the scope of node communication, all nodes are formed except oneself.

Leapfrog number (Hop Count): the leapfrog number summation that two node communications experience.

Single-hop distance (One hop size): two communications do not need a jumping of other node relayings can reach the spacing of node.

Average single-hop distance (Average One Hop size): the average of the whole network node single-hop distance, the single-hop distance sum of all nodes is divided by leapfrog number.In WSN, network average single-hop distance is unique.

Source node (Source Node): refer to serve as the network node that information source sends raw data packets.

Destination node (Destination Node): refer to serve as the network node that the stay of two nights accepts packet.

Intermediate node (Intermediate Node): refer to the network node participating in packet forwarding except source node, destination node.

As shown in Figure 1; the present invention is based on the subregion localization method of node density; study for wireless sensor network orientation problem in earthen ruins protection scene; total n node in assumed wireless sensor network; wherein m beaconing nodes; n-m unknown node, node communication radius is R, and each node has RSSI range capability.Method of the present invention specifically comprises the steps:

1, Boundary Detection:

Boundary Detection is very perfect in WSN research, and the border detection algorithm of current main flow is divided into the Boundary Detection of Boundary Detection based on geometry topology and Corpus--based Method thought.The boundary detection method that the present invention adopts the people such as WangY. to propose is to determine the boundary node of anisotropic network, this algorithm does not carry out any hypothesis to network itself, only utilize node topology information, on the basis that " irregular leapfrog distance can be caused in network cavity " observes, carried out the detection of inner boundary and external boundary by tectonic network shortest path tree.After having detected network boundary, node density model has been adopted to carry out network division.After Boundary Detection, the node in network is divided into boundary node and non-boundary node two parts, respectively with set with represent, wherein, v _erepresent boundary node, v _rrepresent non-boundary node, t is boundary node number, and n-t is non-boundary node number.

2, Region dividing: comprise the Topology Discovery stage, non-boundary node Region dividing stage, boundary node Region dividing stage.

1) in the Topology Discovery stage: netinit, realizes Topology Discovery by broadcast data, node obtains the first step of neighbor node number point and neighbours RSSI sequence using Topology Discovery as oneself.Specifically comprise following steps:

Step S1-1: all beaconing nodes (BN in network ₁, BN ₂, BN ₃bN _m) broadcast Preq packet;

Step S1-2: node judges whether to receive Preq packet, if it is performs step S1-3, otherwise performs step S1-7;

Step S1-3: node judges whether it is first time receive beaconing nodes BN _ithe Preq packet sent, if it is performs S1-4, otherwise performs step S1-8;

Step S1-4: node updates self is to the leapfrog number of beaconing nodes, and by count+1, the Preq packet outwards after this renewal of broadcast, performs step S1-5;

Step S1-5: the node reverts back Pack packet receiving Preq packet, to the node sending Preq packet, acknowledges receipt of message;

Step S1-6: after the node sending Preq packet receives Pack packet, upgrade oneself neighbor information table and route information table;

Step S1-7: this node is considered to bad node, does not participate in position fixing process;

Step S1-8: node judges beaconing nodes BN in current routing information table _icount value whether be less than the count value of new data packets, if it is perform step S1-9, otherwise perform step S1-4;

Step S1-9: abandon this packet;

As shown in table 1, described Location Request Preq packet structure: node_id represents No. id of beaconing nodes; Coordinates represents the coordinate (x, y) of beaconing nodes; Count represents that beaconing nodes arrives the leapfrog number of unknown node, and initial count value is 0; Route [N] represents the sequence node of process on shortest path between beaconing nodes to unknown node, and wherein N value adjusts according to nodes scale, and in the experiment of this patent, N value gets 20.

Table 1 beaconing nodes broadcast Location Request Preq data packet format

node_id

coordinates

count

route[N]

As shown in table 2, the structure of described Pack packet: node_id represents No. id of the beaconing nodes receiving Preq packet; RSSI represents the RSSI value between sending node and receiving node.

Table 2 Pack data packet format

node_id

RSSI

As shown in table 3, the structure of described neighbor information table: Neighbor_id represents No. id of the neighbor node of node, and Neighbor_RSSI represents the RSSI value between this node and neighbor node.

Table 3 neighbor information table

Neighbor_id	Neighbor_RSSI
		v 1	RSSI 1
……	……
		v n	RSSI n

As shown in table 4, the structure of described route information table: Beacon_id represents No. id of beaconing nodes, coordinate represents the coordinate (x of beaconing nodes, y), count represents the leapfrog information of this beaconing nodes of nodal distance, and route [N] represents the sequence node of process on path between node to beaconing nodes.

Table 4 route information table

Beacon_id	coordinate	count	route[N]
				B 1	(x 1,y 1)	3	v 1,v 5
……	……	……	……
				B n	(x n,y n)	5	v 2,v 4,v 8,v 13

2) in the non-boundary node Region dividing stage: consider neighbours' number of boundary node and the difference of non-boundary node, divide only at non-boundary node in carry out.Below, we provide some definition:

First we describe the neighbor node number of nodes: the given WSN network S with n node, the vectorial N=[n of neighbours' number of all node i in network ₁, n ₂..., n _n] ^trepresent, wherein, n _irepresent neighbours' number of node i; The vectorial R of neighbours RSSI sequence of node i _i=[r _i1, r _i2..., r _in] ^trepresent, wherein, r _ijrepresent the RSSI value between node i and node j, neighbours' RSSI value of all node i of whole network can use matrix R=[R ₁, R ₂..., R _n] represent, the neighbours RSSI sequence of node i is shown in the i-th list of matrix.

All nodes are provided with Region dividing mark flag, and initial value is that 0,0 expression does not divide, if this node has completed Region dividing to a certain region, then this is masked as 1.

The non-boundary node Region dividing stage specifically comprises the following steps:

Step S2-1: all beaconing nodes (BN in network ₁, BN ₂, BN ₃bN _m), with own node numbering formation zone numbering area_id, and in its neighbor information table all non-boundary node sending zones partition request Pdiv packet, self zone division symbolizing flag is set to 1 simultaneously;

Step S2-2: node i receives (i < n, j < n, i ≠ j) after the Pdiv request data package that node j sends, and carries out neighbours' RSSI value and selects to judge: if r by formula (1) _ij=0, then node i performs step S2-3, otherwise performs step S2-4;

$g_i=\frac{r_{\mathrm{ij}}-E\left(R_i\right)}{E\left(R_i\right)}---\left(1\right)$

Wherein, $r_{\mathrm{ij}}=\left\{\begin{array}{cc}1& g_i\&le;t_r\\ 0& g_i>t_r\end{array}\right.,$ T _rthe threshold value of node i neighbours RSSI value, t _rget (0.02-0.04), E (R _i) represent the average of node i all neighbours' RSSI value,

Step S2-3: deleted from the neighbor node set of node i by node j, meanwhile, by the neighbours of a node i numerical value n _isubtract 1 as the new neighbours' numerical value of node i, obtain a node i neighbours number vector N upgrade after vectorial N '=[n ₁', n ' ₂..., n ' _n], perform step S2-5;

Step S2-4: keep the neighbor node set of node i constant, perform step S2-5;

Step S2-5: node i is carried out neighbor node number by formula (2) and selected to judge, if formula (2) is set up, then performs step S2-6, otherwise performs step S2-7;

|n′ _i-n′ _j|≤t _n(2)，

Wherein, t _nfor threshold value in Pdiv packet, t _nget (2-5);

Step S2-6: node i judges oneself to belong to the region of the zone number in Pdiv request data package as area_id, and node i and node j are divided in same density area, and node i is by the flag flag set 1 of oneself;

Step S2-7: node i judges self not belong to the region of the zone number in Pdiv request data package as area_id, using the node_id of oneself as new area_id, in self neighbor table, non-boundary node sends Pdiv packet, thus the same with beaconing nodes, initiate the Region dividing centered by oneself.

As shown in table 5, the structure of Region dividing request Pdiv packet: node_id represents No. id of node; t _nneighbor node number threshold value when representing non-boundary node Region dividing; t _rneighbours' RSSI value threshold value when representing non-boundary node Region dividing; Area_id represents zone number, uniquely.

Table 5 Region dividing request Pdiv data packet format

node_id

t n

t r

area_id

Selected by RSSI and the selection of neighbor node number, we construct the Node distribution density model of whole network, and this model depends on the logical topology state of network and topological density.

3) the boundary node Region dividing stage: on the basis that above-mentioned non-boundary node divides, divide the boundary node of network, avoid unstable due to border neighbours' number and areal that is that cause increases, Region dividing is inaccurate.If boundary node area identification be A ' _i, make A ' _i=A _j, wherein with neighbor node each other }, by boundary node the node maximum with neighbor node number in the non-boundary node of its neighbours be divided into the same area; In theory, according to the regularity of distribution, neighbor node number is more, illustrate this node from zone boundary more away from, therefore, boundary node self is selected to be divided in the node that in the non-boundary node of its neighbours, neighbor node number is maximum region in the most reasonable, thus according to the above-mentioned density criteria for classifying divide more accurate.Therefore, the method efficiently solves the problem that boundary node cannot correctly divide because of neighbor node number difference.

Through non-boundary node Region dividing and boundary node Region dividing, whole radio sensing network is divided into the relatively uniform multiple zonules of node density.

3, distance calculates

By the single-hop distance in segmentation network, in each zonule, calculate different single-hop distances respectively, effectively improve positioning precision.Note zonule A _iin single-hop distance be if the link in network between beaconing nodes is m bar, actual distance corresponding on every bar link is the zonule A of link process between statistics beaconing nodes _iwith the leapfrog number hops in each zonule _i, set up formula:

$\left\{\begin{array}{c}{\tilde{d}}_1\&times;{\mathrm{hops}}_1+{\tilde{d}}_2\&times;{\mathrm{hops}}_2+...+{\tilde{d}}_i\&times;{\mathrm{hops}}_i=d_l^1\\ .\\ .\\ .\\ {\tilde{d}}_1\&times;{\mathrm{hops}}_1+{\tilde{d}}_2\&times;{\mathrm{hops}}_2+...+{\tilde{d}}_j\&times;{\mathrm{hops}}_j=d_l^m\end{array}\right.$

Optimal method is utilized to carry out each zonule d _isolve, for the zonule that link between beaconing nodes does not cover, select the single-hop distance in the region at the beaconing nodes place nearest from this zonule as its single-hop distance, after obtaining the single-hop distance of each zonule in network, unknown node, according to the single-hop distance of the routing iinformation in its route information table with this zonule, unknown node place, calculates this unknown node and beaconing nodes BN by following formula _idistance dist _i:

${\mathrm{dist}}_i=\underset{t=1}{\overset{k}{\mathrm{\&Sigma;}}}({\tilde{d}}_t\&times;{\mathrm{hops}}_t)$

Unknown node utilizes least square method to calculate self position after obtaining the beaconing nodes distance of itself and more than three or three.Adopt the subregion localization method based on node density of the present invention, the division in region is not by the impact of beaconing nodes, node locating calculates also not by the restriction in region simultaneously, even if system (is no less than three) when beaconing nodes number is little, also normally can work, there is good extensibility.

In order to verify the validity of method of the present invention, inventor provides following embodiment and further illustrates:

Embodiment 1:

This embodiment is carried out in the linear network of Node distribution piecewise uniform, and network is divided into 3 sections, and the distance between each section of interior joint is equal, is respectively 0.8m, 1.6m, 2.4m.Dispose 16 Micaz nodes altogether, nodal distance ground 80cm, with the frequency of 30s, the outside broadcast data packet of power of-10dBm.Whole network maximum distance is 56.5m, and distance two node leapfrog numbers are farthest 4.In experiment, the communication radius of node is about 13m, and in network, the nodes neighbors number of zones of different is respectively 7.4,9,4.8 (nodes).

As shown in Figure 4, using the boundary node at network two ends and intermediate node as 3 beaconing nodes, as can be seen from the figure, NDL algorithm has obvious reduction in position error, and particularly when nodal pitch increases, the performance boost of algorithm is more obvious.Meanwhile, for the node that indivedual position error is larger, as No. 4 nodes, NDL algorithm effectively improves its positioning precision.

Embodiment 2:

This embodiment is carried out in the uniform rectangular net in Node distribution subregion, and network is divided into 3 regions, and each region area is equal, is 30 × 10=300m ², the node number in region is respectively 21,10,4.35 Micaz node deployments are adopted to test on square, each nodal distance ground 80cm.Node with the frequency of 30s, the outside broadcast data packet of power of-10dBm.The whole network is made up of 35 nodes, and maximum leapfrog number is 3 jumpings.Wherein one jump the node of communication to being in the great majority, totally 60 is right, double bounce communication be 54 right, three to jump having of communication 10 right.Jump communication for one, scope from 4.2m to 15.7m not etc.

As shown in Figure 3, Stochastic choice 3 beaconing nodes position, and in figure, position error is according to the order arrangement from big to small of DV-Hop algorithm, can find out, NDL algorithm positioning performance rises to some extent, and particularly in the region that node deployment is intensive, position error obviously reduces.Compared to DV-Hop algorithm, in NDL, there is good treatment effect for larger position error, improve the positioning performance of whole network.

As can be seen from Figures 6 and 7, to maximum, on average, the analysis of minimal error, in conjunction with the anisotropic degree of two kinds of topological lower network, NDL algorithm of the present invention comparatively DV-Hop algorithm, in two kinds of network topology situations, performance is significantly improved; Under rectangular net topology situation, the positioning precision of NDL algorithm improves more obvious; Relative to the improvement of minimal error, NDL algorithm is better than DV-Hop algorithm to the tolerance of larger position error.By actual experiment, we demonstrate the improvement to traditional DV-Hop algorithm on the basis of subregion single-hop distance estimations of NDL algorithm, the valid certificates adaptability of NDL algorithm to anisotropic network.

Embodiment 3:

This embodiment is emulation experiment, and we carry out the algorithm simulating in two-dimensional space by Matlab, emulates and select rectangle topology and C type topology in network topology, as shown in Figure 8 and Figure 9.From node communication radius, beaconing nodes number, network topology form three aspects, analyze NDL algorithm to the adaptability of anisotropic network in heterogeneous networks situation, the positioning performance of simulation evaluation algorithm.

As shown in Figure 10, the communication radius of node is adjusted to 125m from 80cm every 5m by us, and positioning result shows, and NDL algorithm is DV-Hop algorithm comparatively, and when node communication radius increases, positioning performance is better, and specific experiment analysis result is as follows.When node communication radius is identical, NDL algorithm demonstrates good positioning performance, comparatively DV-Hop algorithm, and it is in rectangle and C-network, and algorithm table reveals the positioning precision raising of maximum 9% and 25% respectively; Along with node communication radius increases, the positioning precision of two kinds of algorithms increases all to some extent, and the increase degree of NDL algorithm is slightly larger than DV-Hop algorithm.

The beaconing nodes number participating in location is increased to 14 by 3 every one by us.As shown in figure 11, positioning result shows, and NDL algorithm is DV-Hop algorithm comparatively, and when beaconing nodes number increases, positioning precision is higher, and specific experiment analysis result is as follows.When beaconing nodes number is identical, NDL algorithm is higher than DV-Hop algorithm positioning precision, and in rectangular net and C-network, the precision showing maximum 10% and 40% respectively improves; Along with the increase of beaconing nodes number, positioning precision is all significantly improved concerning two kinds of algorithms.But the raising degree of NDL algorithm is larger; Along with beacon number increases, position error slowly reduces gradually, and when beaconing nodes number is increased to 9, curve is smooth-out, particularly in rectangular net.

We discuss the impact of network topology form for positioning result, and two kinds of network sizes are all 500 × 500m ².Node distribution is 3 uneven density areas, and positioning result shows, and NDL algorithm is DV-Hop algorithm comparatively, and in anisotropic network, positioning performance is more excellent, has more tolerance.Specific experiment analysis result is as follows.DV-Hop and NDL two kinds of algorithms, the positioning precision under rectangular net topology is all higher than C-network, and this is that location is relatively more difficult, and therefore, position error is larger because the anisotropic degree of C-network is larger; From the degree that positioning precision improves, no matter be when node radius increases or increases in beaconing nodes number, C-network is all large than the precision increase rate of rectangular net, reach the low precision of maximum 16% and 30% respectively, reflect the adaptability that NDL is stronger to anisotropic network.

We are from the angle analysis of node error, and the positioning performance of NDL algorithm is discussed.Figure 12 and 13 reaction be the range finding of whole network under above-mentioned emulation rectangle topology form and position error CDF cumulative chart.From figure, we obviously can draw to draw a conclusion.In range error, what NDL mainly improved is error component larger in DV-Hop, and range error when being 15m-25m for distance in figure is improved comparatively obvious.This is mainly due to when euclidean distance between node pair is nearer, corresponding leapfrog number is less, and the deviation accumulation caused with the single-hop of this DV-Hop algorithm distance is not obvious, and when euclidean distance between node pair is far away, the error of range finding just obviously increases, and the advantage of NDL algorithm is represented; By the impact of range error, NDL algorithm is also in large position error part performance advantage, by reducing position error larger in network, thus improves the positioning precision of algorithm.In emulation, NDL algorithm has the node locating error of 80% within 20m, and DV-Hop algorithm only has about 65%.

To sum up, relative to DV-Hop algorithm conventional at present, the advantage of NDL algorithm of the present invention is as follows:

(1) algorithm positioning precision:

Calculated by the single-hop distance of Region dividing and isolated area, NDL algorithm can obtain range measurement comparatively accurately in anisotropic network environment, thus improves the deficiency of DV-Hop algorithm, effectively improves positioning precision.Solve anisotropic network interior joint skewness, the orientation problem in the irregular situation of signal distributions.

(2) extensibility of algorithm:

By Region dividing, solve DV-Hop algorithm (anisotropy) immalleable problem in complex network environment, location algorithm is reduced the dependence of network topology, the application in various true environment can be applicable to.

Claims (6)

1. the subregion localization method based on node density, it is characterized in that, the method positions for the unknown node in wireless sensor network, total n node in described wireless sensor network, wherein m beaconing nodes, n-m unknown node, node communication radius is R, each node has RSSI range capability, and described localization method specifically comprises the steps:

Step 1, Boundary Detection: by Boundary Detection, the node in network is divided into boundary node and non-boundary node two parts;

Step 2, Region dividing: comprise Topology Discovery stage, non-boundary node Region dividing stage and boundary node Region dividing stage, wherein, in the described Topology Discovery stage, each node is obtaining self neighbor information table and route information table by broadcast data; In the described non-boundary node Region dividing stage, each node selects judgement to carry out density area division by the judgement of neighbours' RSSI value and neighbor node number; In the described boundary node Region dividing stage, be the same area by node division maximum for neighbor node number in boundary node and the non-boundary node of its neighbours; Through non-boundary node Region dividing and boundary node Region dividing, whole radio sensing network is divided into the relatively uniform multiple zonules of node density;

Step 3, distance calculate: the zonule do not covered for link between beaconing nodes, select the single-hop distance in the region at the beaconing nodes place nearest from this zonule as its single-hop distance, after obtaining the single-hop distance of each zonule in network, unknown node is according to the single-hop distance of the routing iinformation in its route information table with this zonule, unknown node place, calculate the distance of the beaconing nodes of this unknown node and more than three or three, then least square method is utilized to calculate self position, so far, in wireless network, unknown node is located successfully.

2., as claimed in claim 1 based on the subregion localization method of node density, it is characterized in that, the Topology Discovery stage in described step 2 specifically comprises the steps:

Step S1-1: all beaconing nodes BN in network ₁, BN ₂, BN ₃bN _mbroadcast Preq packet;

Step S1-2: node judges whether to receive Preq packet, if it is performs step S1-3, otherwise performs step S1-7;

Step S1-3: node judges whether it is Preq packet first time receiving beaconing nodes BNi transmission, if it is performs S1-4, otherwise performs step S1-8;

Step S1-4: node updates self is to the leapfrog number of beaconing nodes, and by count+1, the Preq packet outwards after this renewal of broadcast, performs step S1-5, count and represent that beaconing nodes arrives the leapfrog number of unknown node;

Step S1-5: the node reverts back Pack packet receiving Preq packet, to the node sending Preq packet, acknowledges receipt of message;

Step S1-6: after the node sending Preq packet receives Pack packet, upgrade oneself neighbor information table and route information table;

Step S1-7: this node is considered to bad node, does not participate in position fixing process;

Step S1-8: node judges whether the count value of beaconing nodes BNi in current routing information table is less than the count value of new data packets, if it is performs step S1-9, otherwise performs step S1-4;

Step S1-9: abandon this packet.

3. as claimed in claim 1 based on the subregion localization method of node density, it is characterized in that, the structure of described neighbor information table: comprise the RSSI value Neighbor_RSSI between the id Neighbor_id of the neighbor node of node and this node and neighbor node.

4. as claimed in claim 1 based on the subregion localization method of node density, it is characterized in that, the structure of described route information table: comprise the id Beacon_id of beaconing nodes, the coordinate coordinate of beaconing nodes, the sequence node route [N] of process on path between the leapfrog information count of this beaconing nodes of nodal distance and node to beaconing nodes.

5., as claimed in claim 1 based on the subregion localization method of node density, it is characterized in that, the described non-boundary node Region dividing stage specifically comprises the following steps:

Step S2-1: all beaconing nodes BN in network ₁, BN ₂, BN ₃bN _m, with own node numbering formation zone numbering area_id, and in its neighbor information table all non-boundary node sending zones partition request Pdiv packet, self zone division symbolizing flag is set to 1 simultaneously;

Step S2-2: node i receives (i < n, j < n, i ≠ j) after the Pdiv request data package that node j sends, and carries out neighbours' RSSI value and selects to judge: if r by formula (1) _ij=0, then node i performs step S2-3, otherwise performs step S2-4;

$g_i=\frac{r_{\mathrm{ij}}-E\left(R_i\right)}{E\left(R_i\right)}---\left(1\right)$

Wherein, $r_{\mathrm{ij}}=\left\{\begin{array}{cc}1& g_i\&le;t_r\\ 0& g_i>t_r\end{array}\right.,$ T _rthe threshold value of node i neighbours RSSI value, t _rget (0.02-0.04), E (R _i) represent the average of node i all neighbours' RSSI value,

Step S2-3: deleted from the neighbor node set of node i by node j, meanwhile, by the neighbours of a node i numerical value n _isubtract 1 as the new neighbours' numerical value of node i, obtain a node i neighbours number vector N upgrade after vectorial N '=[n ' ₁, n ' ₂..., n ' _n], perform step S2-5;

Step S2-4: keep the neighbor node set of node i constant, perform step S2-5;

Step S2-5: node i is carried out neighbor node number by formula (2) and selected to judge, if formula (2) is set up, then performs step S2-6, otherwise performs step S2-7;

|n′ _i-n′ _j|≤t _n(2)，

Wherein, t _nfor threshold value in Pdiv packet, t _nget (2-5);

Step S2-6: node i judges oneself to belong to the region of the zone number in Pdiv request data package as area_id, and node i and node j are divided in same density area, and node i is by the flag flag set 1 of oneself;

Step S2-7: node i judges self not belong to the region of the zone number in Pdiv request data package as area_id, using the node_id of oneself as new area_id, in self neighbor table, non-boundary node sends Pdiv packet, thus the same with beaconing nodes, initiate the Region dividing centered by oneself.

6. as claimed in claim 1 based on the subregion localization method of node density, it is characterized in that, the concrete steps that the distance of described step 3 calculates are as follows:

Note zonule A _iin single-hop distance be if the link in network between beaconing nodes is m bar, actual distance corresponding on every bar link is the zonule A of link process between statistics beaconing nodes _iwith the leapfrog number hops in each zonule _i, set up formula:

$\left\{\begin{array}{c}{\tilde{d}}_1\&times;{\mathrm{hops}}_1+{\tilde{d}}_2\&times;{\mathrm{hops}}_2+...+{\tilde{d}}_i\&times;{\mathrm{hops}}_i=d_l^1\\ .\\ .\\ .\\ {\tilde{d}}_1\&times;{\mathrm{hops}}_1+{\tilde{d}}_2\&times;{\mathrm{hops}}_2+...+{\tilde{d}}_j\&times;{\mathrm{hops}}_j=d_j^m\end{array}\right.$

Optimal method is utilized to carry out each zonule d _isolve, for the zonule that link between beaconing nodes does not cover, select the single-hop distance in the region at the beaconing nodes place nearest from this zonule as its single-hop distance, after obtaining the single-hop distance of each zonule in network, unknown node, according to the single-hop distance of the routing iinformation in its route information table with this zonule, unknown node place, calculates this unknown node and beaconing nodes BN by following formula _idistance dist _i:

${\mathrm{dist}}_i=\underset{t=1}{\overset{k}{\mathrm{\&Sigma;}}}({\tilde{d}}_t\&times;{\mathrm{hops}}_t)$

In formula, k is unknown node and beaconing nodes BN _ibetween the number of zonule of link process, t is variable, represents the sequence number of these zonules;

Unknown node utilizes least square method to calculate self position after calculating the beaconing nodes distance of itself and more than three or three.