CN1988550A - Distributing realizing method for radio sensor network no-anchor point location - Google Patents

Abstract

The invention relates to a self-positioning technology in wireless sensor network, especially a distributed method without anchor position. Various network nodes measure the values of distance from each other to get the neighbor information within the scope of their respective perception, and exchange their estimated coordinates many times. The invention adopts the distributed calculation method, firstly estimates the approximate location of nodes, and then each node itself makes iterative calculation of the estimated coordinate location by using the particle spring optimization model. The invention provides a practical method for the wireless sensor network to get geographic information. In addition, the calculation process is implemented on each node without the large cost of communication required by the centralized control, and has a very low complexity of calculation, so that the consumption of positioning calculation is lower, which is suitable for the self-organizing wireless sensor network with nodes measuring distance.

Description

The distributed implementation method of a kind of radio sensor network no-anchor point location

Technical field

The present invention relates to self poisoning technical field in wireless sensor network, the distributed implementation method of particularly a kind of radio sensor network no-anchor point location.

Background technology

Wireless sensor network has significant values in actual applications.Because the node capacity of sensor network is very limited, these limited resources supplIes contents comprise limited power consumption, wireless bandwidth, internal memory and computing ability, thereby need between the node cooperation mutually to finish perception and communication task, wish that node calculates to minimize with the workload of communicating by letter.The coordinate information of each node physical location must be known in the application scenario of a lot of sensor networks.The method that obtains node coordinate by manual measurement or configuration is often infeasible, and then cost is too expensive for the scheme of node outfit global positioning system (GPS) in addition.Usually sensor network forms the coordinate of cover the whole network node by mutual range finding and information exchange between the internal nodes of network, and this is only economical and feasible targeting scheme.Content of the present invention relates to the orientation problem in the sensor network, promptly determines the geographical position of sensor node self under the situation that does not have outside plant (as base station and satellite).

Node locating is a fundamental sum key issue of sensor network configuration and operation.So-called location is meant the network node for one group of unknown position coordinate, to measures such as the distance of neighbor node or neighbours' numbers, utilizes the information that exchanges between node by estimating it, determines the mechanism and the method for each node location.Can not implement accurately control and artificial the setting to all nodes during configuration network under a lot of situations, dispense random dispersion at disposition region from aircraft as node.

Node locating is very big to the meaning of disposing wireless sensor network, and main cause is:

(1) data collected must know that the occurrence positions of perception data just has using value.For example during target following, transducer detects the target of being followed the tracks of, and need come calculated target positions by the position of transducer self; Fire alarm needs place on fire accurately; It is which subregion is found enemy's invasion that the battlefield monitoring then must be provided to the end.These all require to provide the node particular location that sends relevant perceptual signal.

(2) a lot of communication protocols of wireless sensor network are to move on the basis of known node position.As geographical route is exactly implementation data forwarding under the situation of no routing table, has good expandability, has node coordinate but precondition is requirement.In addition, the coordinate of node can be used to optimize the scheduling mechanism on duty during the network operation, and redundant node is had holidays by turns aperiodically with life-saving.

Usually a kind of good targeting scheme has following feature: (1) anchor node number is the least possible, and configuration network is easy; (2) Internet traffic minimizes, and the used power consumption of node is few; (3) numerical procedure is simple, and hardware cost is low; (4) strong robustness adapts to various geographical environments; (5) positioning accuracy height.

Traditional wireless sensor network node locating method is to adopt to lay anchor point in advance, anchor node is exactly a sensor node of knowing self geographical coordinate position, and the anchor point also requirement to some extent of residing position in the whole network to burying underground, can not be met in these requirements of some occasion.Therefore adopt the mode of no anchor point that node is carried out position estimation, be of great significance and value.

People supposed to the research of orientation problem that there were some anchor points in network in the past, and they have obtained geographical coordinate in advance, and there is the anchor point orientation problem technical scheme of finding the solution are not arranged seldom.In fact do not have anchor point and be positioned at and have higher-value in the sensor network, reason is: be not easy to or do not allow to be betrothed to and put anchor point in some occasion, can't arrive at as the region personnel; Secondly, too rely on the value of anchor point coordinate based on the targeting scheme of anchor point, if the coordinate value of anchor point is unreliable, then the position fixing process to whole network produces influence greatly; Bad based on the extensibility of the targeting scheme of anchor point in addition, if network size is very big, then need the anchor point number that disposes too many and be difficult to realize.

If adopt centralized control method to unifiedly calculate the coordinate of all nodes in network-wide basis, centralized calculating is characterised in that from the overall planning of the whole network angle amount of calculation and memory space then can obtain accurate relatively position estimation almost without limits.Its shortcoming mainly be with the nearer node in Centroid position can be because communication overhead be big, and run out of electric energy prematurely, cause the interruption of whole network and Centroid information interchange, cause and can't locate in real time.Centralized solution is not advocated in a lot of application of wireless sensor network.

Distributed Calculation is meant and relies on internodal information exchange and coordination that by the locate mode that node calculates voluntarily, it has overcome the defective of centralized calculating.

Be illustrated in figure 1 as Distributed localization process example, the distance numeral that the numeral among the figure on the line obtains by the range finding of RSSI (indication of received signal intensity) method based on range finding distance between the node.The present invention just is being based on distributed account form, and need not to lay anchor point.

Summary of the invention

The innovative idea of content of the present invention mainly contains 2 points:

(1) from the node of minimum ID numbering, sets up virtual anchor point and coordinate system,, obtain initial roughly estimated position, carry out the position adjustment again, satisfy no anchor point and distributed two kinds of requirements by initial virtual anchor point augmentation other node to the whole network gradually.

(2) initial network layout possible error is very big, if carry out follow-up optimization step simply, the node that possible some position error is very big can't obtain revising all the time.The neighbor information of node can be used for assist location, thus to node definition location confidence level.If the estimated position deviation of node is very big, thereby the location confidence level is lower, then carry out an operation strategy that is called the position variation, this helps these nodes and jumps out the locally optimal solution that may be absorbed in the iteration optimization process, makes the bigger node of position error have an opportunity to obtain position correction.

The object of the present invention is to provide and a kind ofly carry out the node self poisoning according to range finding numerical value between the node neighbours and information exchange, the positioning accuracy height, required cost on network communication is little, thereby used low in energy consumption of node.The no anchor point orientation problem that is based on range finding that solves, this problem model is as follows: the node of given one group of unknown coordinates position, node can calculate the distance to neighbours, determines self-position by the communication between the adjacent node.The coordinate position of the whole network node is called network topology (Layout).

For the location implementation method that the clear the present invention of elaboration proposes, some relevant definition and notions of definition earlier:

Definition 1 (neighbor node): if radio communication mutually between two nodes, then they are called neighbor node mutually.

Definition 2 (range findings): range finding (ranging) is meant that two mutual nodes in communication measurements estimate distance each other.

Definition 3 (anchor points): anchor point (anchor) is meant the node that obtains position coordinates by alternate manner in advance, is also referred to as beacon (beacon) node sometimes.Corresponding all the other nodes are called non-anchor point in the network.The node that hereinafter described reaches generally is meant non-anchor point.

Definition 4 (connection degree): node connectivity is meant the neighbor node number of the detectable discovery of node.Network connection degree is the mean value of neighbours' number of all nodes, and it has reflected the dense degree of sensor configuration.

The invention is characterized in: it is to obtain neighbor information in the sensing range separately by the mutual measuring distance value of each network node, repeatedly exchange estimated position coordinate each other, adopt distributed computational methods, at first estimate the approximate location of node, utilize particle spring Optimization Model then, in the estimated coordinate position of node self difference iterative computation oneself.

Method content of the present invention can be verified that a plurality of sensor nodes of TOSSIM platform emulation simultaneously move same program, Debugging message and configuring condition when operation is provided, real time monitoring network situation in the TOSSIM network platform.

This location algorithm needs following condition when realizing: location Calculation requires sensor node to have the ability that communicates and find range with neighbors, and the hardware module with distance measurement function is a prerequisite; Need during on-premise network node is distributed the unique ID numbering of the overall situation.

At present available distance-finding method comprises received signal intensity indication (RSSI) and the time of advent/time of advent poor (ToA/TDoA).To be receiver determine and the distance of transmitter by the energy of measuring radio frequency rf signal the RSSI method.The ToA/TDoA method is estimated distance between two nodes by measuring the transmission time.

Location Calculation need sensor node to have and neighbor node between carry out radio communication and range finding ability, wireless communication ability is the radio function module that common sensor node all has, and the hardware module with distance measurement function is the prerequisite of the inventive method operation.Distance-finding method and strengths and weaknesses analysis thereof commonly used at present are as follows:

(1) received signal intensity indication.To be receiver determine and the distance of transmitter by the energy of measuring radio frequency rf signal the RSSI method.Owing to realize simply being widely adopted, weak point is that covering or refraction effect can cause that receiving terminal produces measure error, and precision is lower sometimes.

(2) time of advent/time of advent is poor.The ToA/TDoA method is estimated distance between two nodes by measuring the transmission time, and precision is better.Shortcoming is that the very little error in the time measurement can cause very big error amount when the transmission speed of wireless signal is very fast.These two kinds of time-based distance-finding methods are applicable to multiple signal, as radio frequency, acoustics, infrared and ultrasonic signal etc.

Effect of the present invention is: compared with prior art, adopt method of the present invention, have the characteristics and the high orientation precision of Distributed Calculation.More approaching with it technical scheme comprises following two kinds:

(1) N.B.Priyantha, H.Balakrishnan, E.Demaine and S.Teller are at No. 892 technical report of MIT Laboratory for Computer Science in 2003: in " Anchor-FreeDistributed Localization in Sensor Networks ", proposed also that a kind of to approach distributed no anchor point method for calculating and locating be AFL (Anchor-Free Localization) algorithm.Though the AFL algorithm is distributed when the coordinate position of the iteration optimization node of second stage, but, it is non-distributed to see that on the whole the AFL location algorithm comes down to, coordinate time is based on the whole network statistical path jumping figure because it produces initially roughly in the phase I of location, belongs to towards centralized calculation mode; In addition, be absorbed in local solution easily in the iterative process of AFL location, do not take measures to guarantee that node can jump out the local optimum that may be absorbed in and separate.

(2) C.Savarese, J.M.Rabaey and J.Beutel are at the IEEE of calendar year 2001 Int ' l Conf.Acoustics, paper in the Speech and Signal Processing meeting: " Locationing inDistributed Ad-hoc Wireless Sensor Networks ", propose the location algorithm of a kind of called after ABC (Assumption Based Coordinates), its process is based on the coordinate Calculation mode of hypothesis and comes augmentation to diffuse to other interior node of the whole network.Account form distributed, no anchor point that this localization method belongs to, position fixing process is very simple, but designer oneself thinks that also its positioning accuracy is too low, can't directly use, can only replenishing and ancillary method as other localization method.

The invention provides a kind of no anchor point distributed panel-point positioning method for wireless sensor network; Be applicable to the sensor network system with self-organizing feature; Output rusults is the estimation coordinate position of each node; Obtain neighbor information in the sensing range separately by the mutual measuring distance numerical value of each network node; Repeatedly exchange estimated coordinates each other; Adopt distributed computational methods; At first estimate the approximate location of node; Utilize then the mass spring Optimized model; Estimated coordinate position in node self difference iterative computation oneself

By setting up virtual anchor point and coordinate system, by the distance measurement value between the node, proposition realizes a kind of scheme of Distributed localization at each node place, owing to taken the adjacency information of node into consideration, the result carries out creditability measurement and processing to the estimation of node position intermediate, avoids the position fixing process of the whole network to be absorbed in local optimum.

The described no anchor point distributed panel-point positioning method that is used for wireless sensor network, no anchor point feature comprises the virtual coordinates system that sets up network, does not utilize global positioning system (GPS) or other manual location measure to obtain the geographical coordinate positional information of some sensor network nodes in the network design stage.

The described no anchor point distributed panel-point positioning method that is used for wireless sensor network, distributed nature comprises the coordinate position that does not adopt centralized approach to calculate each node, the calculating of node coordinate be need not the participation of Surveillance center or information destination node, only be responsible for self with the neighbor node exchange message and in the computational process of this node enforcement by each node, Surveillance center is as the part of sensor network, other function such as the data that can participate in implementing network are handled, but to implement location process irrelevant with this method.

The described no anchor point distributed panel-point positioning method that is used for wireless sensor network in the network design posterior nodal point self-starting process, is cooperated mutually with neighbor node, realizes and finish the position fixing process of this node, comprises the steps:

Step 1: set up the virtual coordinate system initial reference node of unifying, with node with minimum node numbering as the initial point of virtual coordinates system and as first reference node, the node that has the highest connection degree among the neighbours that second reference node is initial point, more than two reference nodes be equivalent to two virtual anchor points, because they have had accurate coordinate position in the virtual coordinates system, the 3rd reference node is the common neighbours of first and second reference node, when a plurality of such nodes exist, then get node with the highest connection degree;

Step 2: judge the common neighbor node set of three initial reference nodes, calculate the estimated coordinates of this set interior nodes to determine the formula method;

Step 3: obtaining the node set of estimated position, augmentation travels through other node to the network, adopts bikini Min-max method to calculate the node coordinate of unknown position;

Step 4: if the situation of interrupting with three point mode augmentation, then with the traversal of augmentation any 2 as initial augmentation node, determine their neighbor node estimated coordinates with two point form Min-max method, as long as the augmentation process is normal, then should continue to adopt three point mode augmentation other node to the whole network;

Step 5: each node calculates own location confidence level, and it is illustrated in the real neighbours' number of node, and is current by estimating that coordinate position thinks the ratio of the interstitial content in its neighbours' scope;

Step 6: the condition threshold value of node location variation is set, if the location confidence level of node, then starts the scheme that the coordinate position variation is adjusted less than this threshold value;

Step 7: if node enforcing location variation is then got location confidence level among its neighbours and occupied three nodes of high front three, serve as that the Min-max method is carried out on the basis, with the rectangle barycenter determined estimated position as this node with these three nodes;

Step 8: node adopts particle spring Optimization Model, repeatedly the coordinate position of iterative computation self according to the estimated position of its neighbor node.

The described no anchor point distributed panel-point positioning method that is used for wireless sensor network, rectangle barycenter method is got in the phase I utilization, begin other all nodes in the diffusion traversal network-wide basis from reference node, thereby obtain the coordinate position roughly similar to the primitive network topological structure; Second stage is to revise and optimize the coordinate figure of node, makes it to approach the result of real network node location.

The described no anchor point distributed panel-point positioning method that is used for wireless sensor network, the location confidence level of defined node, if the estimated position deviation of node is very big, then locating confidence level shows as very low, XM position mutation operation helps jumping out the locally optimal solution that may be absorbed in the iteration optimization process.

Localization method provided by the invention is based on self coordinate that two dimensional surface is determined sensor node, also can be generalized to three-dimensional situation.Relate in particular to when the placement sensor network and need not under the condition of any anchor point, realize the method for Distributed Calculation coordinate position at sensor node self, when not being equipped with global positioning system or manual configuration node, still can obtain a cover virtual coordinates system of sensor network system.

Further describe specific embodiments of the invention below in conjunction with accompanying drawing.

Description of drawings

Fig. 1 is the information of finding range mutually according between the neighbor node, carries out the schematic diagram of distributed node location Calculation;

Fig. 2 is the operational flowchart of one exemplary embodiment of the present invention;

Fig. 3 gets the Min-max method positioning principle enforcement illustration that three rectangles occur simultaneously;

Fig. 4 is an estimated position coordinate diagram of calculating other node according to the position augmentation of reference node;

Fig. 5 is the processing method exemplary plot when the augmentation interruption occurring.

Embodiment

Be described in greater detail with reference to the attached drawings some embodiment of the present invention.Content of the present invention provides a kind of distributed no anchor point location algorithm of practicality based on the computational geometry theory, as long as adjacent node can be measured distance each other.Position fixing process adopts distributed operation to set up a cover virtual coordinates system, at first estimates the approximate location of node, utilizes particle spring Optimization Model then and in conjunction with the adjacency information of node, iterative coordinate position.This scheme overcomes the defective that existing targeting scheme needs anchor point, and computational process is in the distributed enforcement of each node, avoids centralized operation to produce the mass communication expense.

Figure 2 shows that the positioning step of the inventive method and the flow chart of details.Particularly, wireless sensor network locating method provided by the invention comprises following the implementation phase and step successively:

Phase I: obtain the roughly estimated coordinates of node, finish the initialization layout of network.

According to shown in Figure 2, this stage comprises following process:

201: on-premise network, all nodes are given Global ID's numbering; Placement sensor node in wireless sensor network, all nodes have node serial number, node are not implemented manual its coordinate of location or are taked other measure to obtain real geographical coordinate;

202: between the starting period behind the network design, node is initiated self-align process, obtains the neighbor list of each node and to each neighbour's measuring distance value by radio communication and distance-finding method;

203: node judges whether oneself is the node of lowest number, if then oneself is defined as the initial point (n of virtual coordinates system ₀), coordinate is got work (0,0);

204: by checking the neighbor list of arbitrary node in the network, node judges oneself whether to belong to the neighbours 203 that the lowest number node is an initial point, if then circulate a notice of oneself connection number of degrees value to initial point.Initial point is judged the neighbours with maximum connection degree value according to the numerical value that receives, and it is got make initial reference node (n ₁), according to the distance of the range finding between them (r ₀₁), determine (n ₀) coordinate get work (r ₀₁, 0), set up the virtual coordinates system of network thus, thereby had n ₀And n ₁Two virtual anchor points;

205:n ₀And n ₁Two virtual anchor points are circulated a notice of the location situation of oneself to neighbours separately, if network node learns that oneself is simultaneously for the public neighbours of these two virtual anchor points, and judge node and in these public neighbours, have maximum connection degree value, then this node is got as with reference to node (n ₂), if the minimum connection degree of each node is not less than 3 in the network, then certainly exist n ₀, n ₁And n ₂These reference nodes,

Suppose r ₀₂Be n ₀To n ₂Between distance, r ₁₂Be n _lTo n ₂Between distance, then can draw n by geometric knowledge ₂The coordinate value following two kinds of results (403,404) can be arranged:

${\mathrm{<figure-callout\; id="2"\; label="n"\; filenames="A20051013068700231.png"\; state="\{\{state\}\}">n</figure-callout>}}_2+(\frac{r_{01}^2+r_{02}^2-r_{12}^2}{2r_{01}},\sqrt{r_{02}^2-x_2^2})$ Perhaps $n_2-(\frac{r_{01}^2+r_{02}^2-r_{12}^2}{2r_{01}},-\sqrt{r_{02}^2-x_2^2})$

Whether there is n simultaneously in the neighbor list of network node judgement oneself ₀, n ₁And n ₂, if, then showing the public neighbours of this (a bit) node for these three reference nodes, these nodes are formed set $N_3=\{n_{3_1},n_{3_2},n_{3_3},...,n_{3_m}\},$

Suppose r ₀₃Be n _oTo n ₃Between distance, r ₁₃Be n _lTo n ₃Between distance, r ₂₃Be n ₂To n ₃Between distance, determine set N respectively according to geometrical relationship ₃All interior node coordinates are as follows:

$n_3(\frac{r_{01}^2+r_{03}^2-r_{13}^2}{2r_{01}},\frac{r_{03}^2-r_{23}^2+x_2^2+{\mathrm{<figure-callout\; id="2"\; label="y"\; filenames="A20051013068700231.png"\; state="\{\{state\}\}">y</figure-callout>}}_2^2-2x_2{x}_3}{2r_{01}})$

Because to n ₃Coordinate Calculation in, n no matter ₂The coordinate value be n ₂+ or n ₂-, n ₃The X and Y coordinates value fix, only with the range finding distance dependent of reality;

If node has 3 or 3 above neighbours to obtain initial coordinate, then this node is carried out Min-max method location, produces the initial position of self, and such neighbors is equivalent to anchor point, although there is certain error in this step location, amount of calculation is very little;

206: if node has 3 neighbours to obtain initial estimated coordinates value in the network, then carry out Min-max method (A.Savvides, H.Park, and M.B.Srivastava, " The Bits andFlops of the N-hop Multilateration Primitive For Node Localization Problems; " Networked and Embedded Systems Lab, Electrical Engineering Department, University of Califomia, Los Angeles NESL techinical report TM-UCLA-NESL-2002-03-07, March 2002), determine the initial estimation coordinate of self with this.Here 3 neighbours are equivalent to obtain the anchor node of estimated position, thereby can form an intersection rectangle, and the central point of this rectangle is the estimated position of the node of asking.Because the method is calculated simple,, can be used as the coordinate estimation measure in elementary step though there is bigger error.

Obtained estimated coordinates if node has more than 3 neighbours, then appointed and get 3 such neighbor nodes and finish above-mentioned coordinate estimation procedure.

If adopt monolateral mensuration (A.Savvides based on least mean-square estimate, C.Han, and M.B.Strivastava, " Dynamic Fine-Grained Localization in AdHoc Networks ofSensors; " presented at International Conference On Mobile Computing andNetworking (MobiCom) 2001, Rome, Italy, July 2001), also can set up the solving equation group, determine the coordinate of the node of asking with 3 known coordinate nodes.But the amount of calculation of obvious this solving equation group is very big, and the application scenario of estimated coordinates does not have essential meaning for finding the solution roughly, thereby should not adopt.

According to the node set of above-mentioned steps by known estimated coordinate position, adopt the Min-max method gradually augmentation travel through other node to the whole network.If node only has the neighbours of two known coordinates, serve as that basis execution Min-max method is determined the position also then with them, though this moment, estimation precision was lower, can guarantee that the augmentation process of initial layout is able to normal continuation; Figure 4 shows that according to node n ₀, n ₁And n ₃Coordinate position, adopt the Min-max method to calculate and produce node n ₄Instantiation procedure.

If surpass the regular hour limit, neighbours' number that certain node has known estimation coordinate still has only 2, then based on these 2 neighbours, carries out the Min-max method and determines its coordinate.Though this moment, estimation precision was lower, can guarantee that normal initial alignment augmentation process is continued.

Second stage: implement positioning and optimizing, obtain the accurate estimated coordinates of node.

According to shown in Figure 2, this stage comprises the steps:

207: the compute location confidence level.Each node calculates the location confidence level of oneself in the network.Suppose that real neighbours' number is n in the neighbor list of node _a, find the solution in its estimated coordinates process in optimization, according to these real neighbours' estimated position coordinate, judge to their distance whether still in neighbours' scope, n is made in the estimated position still real neighbours' number note in neighbours' scope _T, the location confidence level parameter θ of node is defined as follows: $\mathrm{\&theta;}=\frac{n_T}{n_a},$

The location confidence level is 100% and do not mean that all of its neighbor relation of node this moment is all correct, because have in neighbours' scope that the estimated coordinates of other node drops on this node;

208: whether the location confidence level of decision node less than defined threshold (as getting 90%), if less than, then strategy is adjusted in the variation of XM position; Otherwise carry out the optimizing process of particle spring model;

209: the optimizing process of carrying out the particle spring model.

Physical " particle-spring " model is as follows: fluid or beformable body resolve into some particles, spring by different modes between the particle connects, and be subjected to connecting the thrust or the pulling force of spring, the synthetic result of power makes particle to moving in the other direction, adjust the suffered more and more littler and state that tends to balance of size of making a concerted effort by continuous position.One group of particle that system model adopts spring to connect is described integrality, by minimizing of system capacity found the solution each particle position.Particle spring Optimization Model is existing more application study in computer graphics, as the simulation of clothing cloth, the autoplacement of network topology etc.

Every limit in the network diagram is mapped as two springs between the particle, and the natural length of poised state is two internodal distance measurement values.For the orientation problem of sensor network, if the distance value between the current estimated position greater than distance measurement value, then two nodes will be subjected to attraction pulling force that they are drawn close; Otherwise the repulsive force that will be separated.To make a concerted effort be 0 when node is suffered, and then the optimizing process of position adjustment stops; For the whole network node,, then stop global optimization procedure if comprehensively to make a concerted effort be 0.

The details of carrying out particle spring optimizing process is as follows:

Node n _iHas a current estimated position

Node regularly sends the estimated position information of this node to neighbours, thereby node has all been grasped the estimated position of oneself and the estimated coordinates of all neighbor nodes.By estimated coordinates found the solution node n _iCan calculate to its certain neighbor node n _j(current estimated position is Estimated distance Also grasp to neighbor node n by methods such as radio distance-measurings simultaneously _jMeasuring distance r _{I, j}

The expression direction

Arrive

On unit vector, direction

Promptly powerful (force) component of directing force be: ${\stackrel{\&RightArrow;}{F}}_{i,j}={\hat{v}}_{i,j}({\hat{d}}_{i,j}-r_{i,j}).$ Node n _iWhole brute force add up to: ${\stackrel{\&RightArrow;}{F}}_i=\underset{j}{\mathrm{\&Sigma;}}{\stackrel{\&RightArrow;}{F}}_{i,j}.$

There are difference in measuring distance and estimated distance, and this difference can adopt node n _iWith node n _jBetween ENERGY E _{I, j}Measure expression, promptly by Square determining of size.Node n _iCan measure do as follows:

$E_i=\underset{j}{\&Sum;}{E}_{i,j}=\underset{j}{\&Sum;}{({\hat{d}}_{i,j}-r_{i,j})}^2$

When carrying out particle spring Optimization Model, as node n _iDirection to whole brute force

Move one when indivisible, its ENERGY E _iCan reduce.The amount of movement of determining the node estimated coordinates is extremely important, because it will guarantee that the energy of reposition is littler than the energy of origin-location, and guarantee to move and can not produce locally optimal solution.Rule of thumb can select the amount of movement of each node to be

m _iBe n _iReal neighbor node number.

In addition, also can adopt the directing force amendment scheme, promptly not to all powerful vector additions, and according to the distance values size of directing force, only getting its maximum first three items directing force carries out vector addition, gets its reverse vector then and carries out the coordinate position correction.Because all the other directing forces are less to the influence of making a concerted effort, and can reduce unnecessary amount of calculation.

210: the process of XM position variation is as follows: get confidence level the highest preceding 3 nodes in location among the neighbours of node, if the minimum connection degree of node is not less than 3, then certainly exist these nodes, adopt the Min-max method to determine the intersection rectangle to these 3 nodes, get the estimated position of rectangle barycenter as this node;

211: whether decision node satisfies the iteration termination condition, if satisfy then obtain the final accurate coordinates of estimating of node, otherwise the whole process of second stage is carried out in circulation, promptly restarts from the location confidence level of computing node;

The judgment criterion of iteration termination and condition can adopt any one in following three kinds: the gross energy E of (1) node _iApproach 0 or, then can end particle spring optimizing process less than a certain scope; (2) the node iterations reaches certain limit, can obtain accurate estimated coordinates; (3) the estimated coordinates numerical value of node no longer changes in nearest several times iteration, can declare also that then the position fixing process of this node finishes;

212: the final estimated coordinates that obtains node.

When the characteristics of above-mentioned steps are the meticulous adjustment of coordinate, if node is absorbed in local optimum, the locally optimal solution problem that can avoid other method to cause by the position variation then, thus help improving positioning performance.

Above-mentioned method, its characteristics are, node is in the iteration of second stage is adjusted, owing to take node locating confidence level and threshold decision criterion, if node is absorbed in local optimum and separates, then can jump out local optimum and separate by the node location strategy that makes a variation, thus the local solution problem of avoiding the AFL algorithm to cause.

Above-mentioned method, its characteristics are that also the whole process of node locating is distributed fully, need not the centralized control of the whole network, and centralized approach is can't be practicable in a lot of application scenarios of wireless sensor network.In the method that the present invention proposes, node only need be carried out the correlation computations step at this node self place, and in conjunction with the positional information of neighbours exchange, just can finish whole position fixing process, thereby be convenient in the software function module of node communication protocol stack, realize when the real sensor network design, to have very big using value.

In the following description, the ranging system of sensor node is realized by received signal intensity indication (RSSI), but can certainly be realized by the technology of other any wireless measurement distance.For example, can adopt the method for the time of advent/time of advent poor (ToA/TDoA), estimate distance between two nodes by measuring the transmission time.

Now two Main Stage of distributed no anchor point position fixing process are carried out detailed example and describe, this programme can be embedded into the communication protocol stack of sensor node.

(1) the whole network node is set up initial roughly coordinate

This stage implementation process is as follows:

1. the true coordinate of initialization node guarantees the minimum connection number of degrees value of the connection degree of each point greater than regulation, and it is more suitable that for example minimum connection degree value is got work 3, both satisfied connectivity platform, and the density that network is laid is excessive.Certainly also may there be isolated node cluster, forms the isolated island subgraph of cutting apart, if the node cluster of this class isolated island class is cancelled that then these nodes are positioned operation, and do not influenced the positioning action to most of nodes in the network with minimum probability.

For example, the number of sensor node gets 100 in the network, and the wireless distance finding scope between the node is got 30 meters, i.e. the neighbor node each other of node in this scope, and network configuration is in the landform zone of 100 meters of 100 meters *.

2. obtain the neighbor list of each node by distance-finding method and to neighbours' measuring distance value.

3. node judges whether oneself is the node of lowest number, if then oneself is defined as the initial point 401 (n of virtual coordinates system ₀), coordinate is got work (0,0).

4. node is checked neighbor list, judges oneself whether to belong to the neighbours that the lowest number node is an initial point, if, then to the connection number of degrees value of initial point circular oneself.Initial point is judged the neighbours with maximum connection degree value according to the numerical value that receives, and it is got make initial reference node 402 (n ₁), according to the distance of the range finding between them (r ₀₁), determine (n ₀) coordinate get work (r ₀₁, 0), node 402 is positioned on the X reference axis.Set up the virtual coordinates system of network thus, and have n ₀And n ₁Two virtual anchor points.

5. two virtual anchor points are circulated a notice of the location situation of oneself to neighbours separately, if it oneself is the public neighbours of two virtual anchor points simultaneously that network node is learnt, and judge node and in these public neighbours, have maximum connection degree value, then this node is got as with reference to node (n ₂).

6. whether there is n simultaneously in the neighbor list of network node judgement oneself ₀, n ₁And n ₂, if, then showing the public neighbours of this (a bit) node for these three reference nodes, these nodes are formed set $N_3=\{n_{3_1},n_{3_2},n_{3_3},...,n_{3_m}\}$ 。According to the method for above introducing, determine set N ₃The estimated coordinates of interior nodes.

Because to n ₃Coordinate Calculation in, n no matter ₂The coordinate value be n ₂+ (403) are n still ₂-(404), n ₃The X and Y coordinates value be 405 positions of fixing.

7., then carry out the Min-max method and determine the initial estimation coordinate of self if node has three neighbours to obtain initial estimated coordinates value in the network.

Figure 3 shows that the Min-max method that adopts three anchor points location, promptly with the coordinate position (x of certain anchor point i (i=1,2,3) _i, y _i) be the basis, add or deduct distance measurement value d _i, obtain the bounding box of anchor point i: [x _i-d _i, y _i-d _i] * [x _i+ d _i, y _i+ d _i].At all location point [x _i+ d _i, y _i+ d _i] in get minimum value, all [x _i-d _i, y _i-d _i] in get maximum, the rectangle that then occurs simultaneously is:

[max(x _i-d _i)，max(y _i-d _i)]×[min(x _i+d _i)，min(y _i+d _i)]

Anchor point 301, anchor point 303 and the anchor point 304 common intersection rectangle frames 302 that form.The central point 305 of this rectangle is the estimated position of the node of asking, because the method is calculated simply, though there is bigger error, can be used as the coordinate estimation in elementary step.

8. according to the node set of above-mentioned steps 7 by known estimated coordinate position, adopt the Min-max method gradually augmentation travel through other node to the whole network.

Figure 4 shows that according to node n ₀, n ₁And n ₃Coordinate position, adopt the Min-max method to calculate and produce node n ₄Instantiation procedure.Three rectangles (0,4), (1,4) (3,4) produce new intersection rectangle 409,406 rectangles (1,4) by node n ₁And n ₄Range finding distance be the rectangle radius, with n ₁Be the rectangle frame that rectangular centre forms, the formation of other rectangle is similar with it.The barycenter 410 of new intersection rectangle is got makes node n ₄Estimated coordinates.

9. if surpass the regular hour limit, this time dimension can be provided with in the time synchronized mechanism of network system, find that neighbours' number that certain node has a known estimation coordinate still has only two,, carry out the Min-max method and determine its coordinate then based on these two neighbours.Though this moment, positioning accuracy was lower, can guarantee that normal initial alignment augmentation process is continued.

Figure 5 shows that the processing method example when location augmentation interruption appears in network.Under normal circumstances, node is according to forming the rectangle barycenter, promptly producing estimated position 508 based on three neighbours (505,506 and 507) of known location.Analyze from the whole network angle,, calculate the coordinate of new node if the augmentation interrupt situation is a foundation with two initial augmentation points (502 and 503) that dotted line among Fig. 5 is drawn then.

(2) process of iterative accurate coordinates

Neighborhood in the second stage iteration is constant, and this stage details is as follows:

1. each node calculates the location confidence level parameter θ of oneself in the network.Calculate distance with neighbours according to the estimated position, might exceed the maximum wireless finding range.

2. whether the location confidence level of decision node less than defined threshold, if less than, then XM position variation, otherwise carry out particle spring Optimization Model.

3. the node location mutation process is as follows: get first three the highest node of location confidence level among the neighbours of node, adopt the Min-max method to determine the intersection rectangle, get the estimated position of rectangle barycenter as this node.

4. the optimizing process of particle spring model is as follows:

If node n _iHas a current estimated position Node regularly sends this positional information to neighbor node, thereby each node is all known the estimated position of oneself and the estimated position of all neighbor nodes.By calculating to estimated coordinates, node n _iCan calculate to its certain neighbor node n _j(current estimated position is Estimated distance

Also know to neighbor node n by wireless distance finding simultaneously _jMeasuring distance r _{I, j}By making a concerted effort of all directing forces of computing node, perhaps the first three items directing force to maximum carries out vector addition, gets its reverse resultant vector and in addition the scale convergent-divergent is (promptly As correction direction and size to this estimated coordinates.

There are difference in measuring distance and estimated distance, and this difference adopts node n _iWith node n _jBetween ENERGY E _{I, j}Measure expression.Whole network system can measure work: $E=\underset{j}{\mathrm{\&Sigma;}}{E}_i.$ In the node Distributed localization was calculated, the notion of the gross energy of network system did not adopt in fact, only has reference value, can be used for the whole network positioning accuracy is assessed when the whole network is monitored.

5. whether decision node satisfies iteration optimization and ends condition, if satisfy then obtain the final accurate coordinates of estimating of node, otherwise repeats the whole process of second stage.

Node constantly carries out the position to be upgraded, up to the iterations that satisfies regulation.If iterations is very few, then positioning accuracy is lower.According in computer graphics, particle spring Research on optimal model result being shown,, be enough to satisfy solving precision as long as iterations reaches 50 times.

The above only is the preferable main implementation process of the present invention, is not to be used for limiting practical range of the present invention; Every according to equivalence variation and modification that the present invention did, be all claim of the present invention and contain.

Claims (7)

1, a kind of no anchor point distributed panel-point positioning method for wireless sensor network; Be applicable to the sensor network system with self-organizing feature; Output rusults is the estimation coordinate position of each node; It is characterized in that: obtain neighbor information in the sensing range separately by the mutual measuring distance numerical value of each network node; Repeatedly exchange estimated coordinates each other; Adopt distributed computational methods; At first estimate the approximate location of node; Utilize then the mass spring Optimized model; Estimated coordinate position in node self difference iterative computation oneself

By setting up virtual anchor point and coordinate system, by the distance measurement value between the node, proposition realizes a kind of scheme of Distributed localization at each node place, owing to taken the adjacency information of node into consideration, the result carries out creditability measurement and processing to the estimation of node position intermediate, avoids the position fixing process of the whole network to be absorbed in local optimum.

2, the no anchor point distributed panel-point positioning method that is used for wireless sensor network as claimed in claim 1, it is characterized in that, no anchor point feature comprises the virtual coordinates system that sets up network, does not utilize global positioning system or other manual location measure to obtain the geographical coordinate positional information of some sensor network nodes in the network design stage.

3, the no anchor point distributed panel-point positioning method that is used for wireless sensor network as claimed in claim 1, it is characterized in that, distributed nature comprises the coordinate position that does not adopt centralized approach to calculate each node, the calculating of node coordinate be need not the participation of Surveillance center or information destination node, only be responsible for self with the neighbor node exchange message and in the computational process of this node enforcement by each node, Surveillance center is as the part of sensor network, other function such as the data that can participate in implementing network are handled, but to implement location process irrelevant with this method.

4, according to claim 1 or the 2 or 3 described no anchor point distributed panel-point positioning methods that are used for wireless sensor network, it is characterized in that, in the network design posterior nodal point self-starting process, cooperate mutually with neighbor node, realize and finish the position fixing process of this node, comprise the steps:

Step 1: set up the virtual coordinate system initial reference node of unifying, with node with minimum node numbering as the initial point of virtual coordinates system and as first reference node, the node that has the highest connection degree among the neighbours that second reference node is initial point, more than two reference nodes be equivalent to two virtual anchor points, because they have had accurate coordinate position in the virtual coordinates system, the 3rd reference node is the common neighbours of first and second reference node, when a plurality of such nodes exist, get node with the highest connection degree;

Step 2: judge the common neighbor node set of three initial reference nodes, calculate the estimated coordinates of this set interior nodes to determine the formula method;

Step 3: obtaining the node set of estimated position, augmentation travels through other node to the network, adopts bikini Min-max method to calculate the node coordinate of unknown position;

Step 4: if the situation of interrupting with three point mode augmentation, then with the traversal of augmentation any 2 as initial augmentation node, determine their neighbor node estimated coordinates with two point form Min-max method, as long as the augmentation process is normal, then should continue to adopt three point mode augmentation other node to the whole network;

Step 5: each node calculates own location confidence level, and it is illustrated in the real neighbours' number of node, and is current by estimating that coordinate position thinks the ratio of the interstitial content in its neighbours' scope;

Step 6: the condition threshold value of node location variation is set, if the location confidence level of node, then starts the scheme that the coordinate position variation is adjusted less than this threshold value;

Step 7: if node enforcing location variation is then got location confidence level among its neighbours and occupied three nodes of high front three, serve as that the Min-max method is carried out on the basis, with the rectangle barycenter determined estimated position as this node with these three nodes;

Step 8: node adopts particle spring Optimization Model, repeatedly the coordinate position of iterative computation self according to the estimated position of its neighbor node.

5, the no anchor point distributed panel-point positioning method that is used for wireless sensor network according to claim 4, it is characterized in that, rectangle barycenter method is got in the phase I utilization, begin other all nodes in the diffusion traversal network-wide basis from reference node, thereby obtain the coordinate position roughly similar to the primitive network topological structure; Second stage is to revise and optimize the coordinate figure of node, makes it to approach the result of real network node location.

6, the no anchor point distributed panel-point positioning method that is used for wireless sensor network according to claim 4, it is characterized in that, the location confidence level of defined node, if the estimated position deviation of node is very big, then locating confidence level shows as very low, XM position mutation operation helps jumping out the locally optimal solution that may be absorbed in the iteration optimization process.

7, the node positioning method that is used for sensor network according to claim 5 is characterized in that, the phase I utilization is got rectangle barycenter method and come initialization network topology, obtains the roughly coordinate of node; Second stage adopts particle spring Optimization Model, and estimated coordinates is optimized adjustment,

The phase I process is as follows:

201: on-premise network, all nodes are given Global ID's numbering;

202: between the starting period behind the network design, node is initiated self-align process, obtains the neighbor list of each node and to each neighbour's measuring distance value by radio communication and distance-finding method;

203: node judges whether to be minimum ID, if then be defined as the initial point n of virtual coordinates system ₀(0,0);

204: by checking the neighbor list of arbitrary node in the network, node judges oneself whether to belong to the neighbours 203 that the lowest number node is an initial point, if, then circulate a notice of the connection number of degrees value of oneself to initial point, initial point is according to the numerical value that receives, judge neighbours, it is got make initial reference node (n with maximum connection degree value ₁), according to the distance of the range finding between them (r ₀₁), determine (n ₀) coordinate get work (r ₀₁, 0), set up the virtual coordinates system of network thus, thereby had n ₀And n ₁Two virtual anchor points;

205:n ₀And n ₁Two virtual anchor points are circulated a notice of the location situation of oneself to neighbours separately, if network node learns that oneself is simultaneously for the public neighbours of these two virtual anchor points, and judge node and in these public neighbours, have maximum connection degree value, then this node is got as with reference to node (n ₂), if the minimum connection degree of each node is not less than 3 in the network, then certainly exist n ₀, n ₁And n ₂These reference nodes,

Suppose r ₀₂Be n ₀To n ₂Between distance, r ₁₂Be n ₁To n ₂Between distance, then can draw n by geometric knowledge ₂The coordinate value following two kinds of results (403,404) can be arranged:

$n_2+(\frac{r_{01}^2+r_{02}^2-r_{12}^2}{2r_{01}},\sqrt{r_{02}^2-x_2^2})$ Perhaps $n_2-(\frac{r_{01}^2+r_{02}^2-r_{12}^2}{2r_{01}},-\sqrt{r_{02}^2-x_2^2})$

Whether there is n simultaneously in the neighbor list of network node judgement oneself ₀, n ₁And n ₂, if, then showing the public neighbours of this node for these three reference nodes, these nodes are formed set $N_3=\{n_{3_1},n_{3_2},n_{3_3},...,n_{3_m}\},$

Suppose r ₀₃Be n ₀To n ₃Between distance, r ₁₃Be n ₁To n ₃Between distance, r ₂₃Be n ₂To n ₃Between distance, determine set N respectively according to geometrical relationship ₃All interior node coordinates are as follows:

$n_3(\frac{r_{01}^2+r_{03}^2-r_{13}^2}{2r_{01}},\frac{r_{03}^2-r_{23}^2+x_2^2+y_2^2-2x_2{x}_3}{2r_{01}})$

Because to n ₃Coordinate Calculation in, n no matter ₂The coordinate value be n ₂+ or n ₂-, n ₃X and Y coordinate value fix, only with the range finding distance dependent of reality;

If node has 3 or 3 above neighbours to obtain initial coordinate, then this node is carried out Min-max method location, produces the initial position of self, and such neighbors is equivalent to anchor point, although there is certain error in this step location, amount of calculation is very little;

206: by the set of node of known estimated position, employing Min-max method augmentation gradually travels through to other node, if node only has the neighbours of two known coordinates, serve as that basis execution Min-max method is determined the position also then with them, though this moment, estimation precision was lower, can guarantee that the augmentation process of initial layout is able to normal continuation;

The second stage process is as follows:

207: the compute location confidence level, each node calculates the location confidence level of oneself in the network.Suppose that real neighbours' number is n in the neighbor list of node _a, find the solution in its estimated coordinates process in optimization, according to these real neighbours' estimated position coordinate, judge to their distance whether still in neighbours' scope, n is made in the estimated position still real neighbours' number note in neighbours' scope _T, the location confidence level parameter θ of node is defined as follows: $\mathrm{\&theta;}=\frac{n_T}{n_a},$

The location confidence level is 100% and do not mean that all of its neighbor relation of node this moment is all correct, because have in neighbours' scope that the estimated coordinates of other node drops on this node;

208: whether the location confidence level of decision node less than defined threshold, if less than, then carry out some position position variation strategy; Otherwise carry out the optimization of particle spring model;

209: when carrying out particle spring Optimization Model, as node n _iDirection to whole brute force Move one when indivisible, its ENERGY E _iCan reduce, determine that the amount of movement of node estimated coordinates is extremely important,, and guarantee to move and can not produce locally optimal solution, rule of thumb can select the amount of movement of each node to be because it will guarantee that the energy of reposition is littler than the energy of origin-location m _iBe n _iReal neighbor node number;

210: the process of XM position variation is as follows: get confidence level the highest preceding 3 nodes in location among the neighbours of node, if the minimum connection degree of node is not less than 3, then certainly exist these nodes, adopt the Min-max method to determine the intersection rectangle to these 3 nodes, get the estimated position of rectangle barycenter as this node;

211: whether decision node satisfies the iteration termination condition, if satisfy then obtain the final accurate coordinates of estimating of node, otherwise the whole process of second stage is carried out in circulation, promptly restarts from the location confidence level of computing node;

The judgment criterion of iteration termination and condition can adopt any one in following three kinds: the gross energy E of (1) node _iApproach 0 or, then can end particle spring optimizing process less than a certain scope; (2) the node iterations reaches certain limit, can obtain accurate estimated coordinates; (3) the estimated coordinates numerical value of node no longer changes in nearest several times iteration, can declare also that then the position fixing process of this node finishes;

212: the final estimated coordinates that obtains node.

Images