CN103648164B - A kind of based on the difference time of advent and the wireless-sensor network distribution type localization method of Gossip algorithm - Google Patents

Abstract

A kind of based on the difference time of advent and the wireless-sensor network distribution type localization method of Gossip algorithm, the present invention relates to wireless-sensor network distribution type localization method.The present invention is to solve the difference low problem of the localization method positioning precision simple time of advent in wireless sensor network.One, anchor node obtains self position coordinates；Two, realize Distributed Time to synchronize；Three, anchor node wakes up monitoring unknown node at random up；Four, wake up anchor node up and preserve reception signal moment and local coordinate system；Five, whether all anchor nodes are fully completed signal monitoring and data preserve；Six, anchor node j receives the data of its all M adjacent anchor nodes；Seven, the anchor node j initial estimate for unknown node position is obtained；Eight, all anchor nodes obtain unknown node position initial estimate；Nine, run Gossip algorithm and randomly choose adjacent anchor node switching location data；Ten, algorithm terminates.The present invention is applied to wireless sensor network exemplary operation field.

Description

A kind of based on the difference time of advent and the wireless-sensor network distribution type localization method of Gossip algorithm

Technical field

The present invention relates to wireless-sensor network distribution type localization method.

Background technology

Wireless sensor network, as one of the core technology of Internet of Things, ten is propped up greatly for change human lives future by praise One of support technology.Since wireless sensor network is started by extensive concern, the focus of its location technology always theoretical research is asked Topic.In recent years, Chinese scholars proposes multiple wireless sensor network center type or distributed location method, but it is substantially It is that the situation for unknown node (node of its position to be measured in network) positive location designs.

Wireless sensor network is by the node institute group in a large number with functions such as data acquisition, data process and wireless data transceivings The communication network become.Typical wireless sensor network working method is to be arranged by substantial amounts of wireless sensor node or shed Region interested, wireless sensor node quickly forms a wireless network by the way of self-organizing.Each sensor saves Point is had oneself communication zone and is detected the information such as the temperature of surrounding, humidity or frequency spectrum by awareness apparatus, with Time can also be equipped with communication equipment and adjacent node and carry out short haul connection.

Own characteristic based on above-mentioned wireless sensor network, the usually such as military surveillance of the working region of sensor node, The region that the mankind such as road conditions detection are not easily accessible, and should for the typical case as the sensor network such as post-disaster reconstruction, climate monitoring With environment, sensor node generally the means such as is shed by aircraft and is placed in working region, thus their position is all random And the unknown.But in numerous applications, the data that node is gathered must be the most meaningful in conjunction with its residing coordinate position, Research thus for the location technology of wireless sensor network unknown node has great importance.

The method completing to measure by arriving the time difference of node based on two signals thus carry out positioning is called the time of advent Difference location (Time Difference of Arrival, TDOA).TDOA is a kind of localization method being widely adopted, at present It is defined as first-selected localization method by multiple location hardware platform.TDOA location technology is at cellular positioning system, WCDMA The wireless location systems such as technology, wireless distance finding radar, roadbed multipoint positioning have a very wide range of applications, and has ground Study carefully in the location application that personnel begin attempt to be expanded to wireless sensor network.But it is fixed at wireless sensor network at present Among the application of position, it is relatively low that TDOA algorithm also exists positioning precision, the problem that positioning performance is poor.And, current TDOA Algorithm is the most all used to realize self positive location of unknown node, not yet by its expanded application in similar military precision strike Or illegally access the distributed Passive Positioning field needed for the wireless sensor network typical operating environment such as frequency spectrum radio station monitoring.

Due to communication capacity, data-handling capacity and equipment dependability that wireless sensor node is limited, wireless sensor network In node the most all use the mode of intensive placement.Owing to node density is higher, the node institute therefore closed on geographical position The data gathered are generally of bigger redundancy.If can carry out these data merging, compressing in data transmission procedure, Just can be substantially reduced the number of packet during subsequent transmission and grouping dimension, thus improve network capacity and reduce node Energy expenditure.Therefore the research about the data compression in wireless sensor network and effective transmission means just becomes one and grinds Study carefully focus.Under this background, be born can be applicable to wireless-sensor network distribution type common recognition Gossip algorithm.Should Each node in algorithm randomly with its selected certain (or certain group) adjacent node exchange data, then the two (or This group) node is utilized respectively convex merging algorithm and merges these data, and replace oneself original data by new data.

Gossip algorithm was proposed by Tsitsiklis et al. first as far back as 1984, and this algorithm is merely with this locality of network node Information carries out data exchange with the information of its neighbor node, solves the average common recognition problem under distribution occasion.In wireless biography Sensor network is studied the earliest, achievement in research is the most ripe and to apply most common be paired Gossip algorithm.At present, base In the focus that the application in wireless sensor network of the average common recognition problem of Gossip algorithm is studied especially.With traditional road Different by algorithm, Gossip routing algorithm is emphasized by the way of the local information of node and neighbor node carry out data exchange Carrying out data renewal, we call an iteration process a data updating process.Owing in an iterative process, only relating to And to a hop neighbor nodal information and the exchange of local information, so network based on Gossip is not required to traditional routing and calculates The Route establishment process of method and maintenance process, do not have the repeating process of data yet, this be greatly lowered network energy expenditure, Extend the life cycle of network node.Gossip algorithm is a kind of stochastic route algorithm simultaneously, it is to avoid because of channel in network Compete the problems such as " packet loss " caused and network congestion.Further, in whole network, all of node status equity, does not has what is called " Centroid ", it also avoid the existence of bottle neck effect.So Gossip algorithm is to be distributed in wireless sensor network Formula is averagely known together the good solution of problem, thus it is envisioned that Gossip algorithm is the most necessarily equally applicable to wireless sensing Device network distribution type positions this typical distributed problem.

In sum, although now widely used simple TDOA localization method performance ten in mobile radio networks Divide excellent, but the application in wireless sensor network still exists the problems such as positioning precision is relatively low, positioning performance is limited. And for how to use in the distributed Passive Positioning problem determined by the feature of wireless sensor network, be currently also In particular how one research blind spot, complete to carry out high accuracy for invasion unknown node in similar military affairs and hit, or for The illegal unknown radio set accessing frequency spectrum carries out special engineering application problems such as being accurately positioned, is also badly in need of research worker and proposes to cut Solid yardage case.

Summary of the invention

The present invention is to solve that in wireless sensor network, simple TDOA localization method positioning precision is low, positioning performance is limited And the problems such as this typical case's application scenarios of distributed Passive Positioning cannot be applicable to, and provide one and calculate poor for the time of advent Method and Gossip algorithm organically combine and are applicable to the brand-new localization method of wireless-sensor network distribution type Passive Positioning.

Wireless-sensor network distribution type localization method based on the difference time of advent and Gossip algorithm realizes according to the following steps:

Step one: the random or N number of anchor node of artificial layout in wireless sensor network coverage, each anchor node is borrowed GPS or the artificial mode arranged is helped to obtain self unknown coordinates (x_k,y_k,)；

Step 2: in whole wireless sensor network, all anchor nodes realize Distributed Time together by broadcast Gossip algorithm Step；

Step 3: each anchor node wakes up up at each slotted random, whether monitoring unknown node occurs；

Step 4: if unknown node occurs, then wake up anchor node k up and will receive the radio signal that unknown node is launched, anchor Node k record receives the moment t of signal_k, it is saved together (x together with local coordinate system_k,y_k,t_k)；

Step 5: each anchor node wakes up up successively, repeats step 4, until all anchor nodes all complete the signal of unknown node Monitoring and data preserve；

Step 6: the anchor node in wireless sensor network wakes up up at random, it is assumed that anchor node j is in wake-up states, this anchor saves The point all adjacent anchor nodes broadcast solicitations in the range of communication radius, it is desirable to it is returned node coordinate and receives the signal moment, This anchor node starts to receive data and whether real-time judge receives the data of whole adjacent anchor node, if not receiving the most adjacent Anchor node data, then continue to, if receiving whole adjacent anchor node data, then completes the reception of data and by anchor node J receives the data of its all M adjacent anchor nodes and is saved in this locality, obtains local data as follows

$\left[\begin{array}{ccc}{x}_0^j& y_0^j& t_0^j\end{array}\right]=\left[\begin{array}{ccc}{x}_1& y_1& t_1\\ x_1& y_2& t_2\\ \·& \·& \·\\ \·& \·& \·\\ \·& \·& \·\\ x_i& y_i& t_i\\ \·& \·& \·\\ \·& \·& \·\\ \·& \·& \·\\ x_M& y_M& t_M\end{array}\right];$

Step 7: the local data that anchor node j obtains according to step 6, according to the Taylor algorithm positioned based on TDOA Primary iteration location formula obtains unknown node position primary iteration value ((x_j(0),y_j(0)), run according to the thresholding pre-set The iterative process of Taylor algorithm, until obtaining anchor node j initial estimate ((x for unknown node position_j(0),y_j(0))；

Step 8: remaining anchor node in wireless sensor network wakes up up at random, repeats step according to the working method of anchor node j Rapid six and step 7 obtain this anchor node initial estimate for unknown node position, until all anchor nodes complete for not Knowing the initial estimation of node location, in wireless sensor network, the unknown node position initial estimate coordinate of all anchor nodes is

$\left[\begin{array}{cc}{x}^{\′}\left(0\right)& y^{\′}\left(0\right)\end{array}\right]=\left[\begin{array}{cc}{x}_1^{\′}\left(0\right)& y_1^{\′}\left(0\right)\\ x_2^{\′}\left(0\right)& y_2^{\′}\left(0\right)\\ \·& \·\\ \·& \·\\ \·& \·\\ x_j^{\′}\left(0\right)& y_j^{\′}\left(0\right)\\ \·& \·\\ \·& \·\\ \·& \·\\ x_N^{\′}\left(0\right)& y_N^{\′}\left(0\right)\end{array}\right];$

Step 9: all anchor nodes obtain this node initial estimation for unknown node position in wireless sensor network On the basis of value, all N number of anchor nodes of the whole network run paired Gossip algorithm, paired Gossip algorithm carrying out practically process As follows: to assume to be waken up at time slot t anchor node j, this anchor node randomly chooses an adjacent anchor node i exchange initial estimation Value Data

$x_j^{\′}(t+1)=x_i^{\′}(t+1)=\frac{1}{2}(x_i^{\′}\left(t\right)+x_j^{\′}\left(t\right))$

$y_j^{\′}(t+1)=y_i^{\′}(t+1)=\frac{1}{2}(y_i^{\′}\left(t\right)+y_j^{\′}\left(t\right))$

The whole iterative process of paired Gossip algorithm that the N number of anchor node of the whole network is run can be expressed as follows

$z^{\′}(t+1)=W\left(t\right)z^{\′}\left(t\right)=\underset{n=1}{\overset{t}{\mathrm{\Σ}}}W\left(n\right)z^{\′}\left(0\right)$

Wherein

$z^{\′}\left(t\right)=\left[\begin{array}{cc}{x}^{\′}\left(t\right)& y^{\′}\left(t\right)\end{array}\right]=\left[\begin{array}{cc}{x}_1^{\′}\left(t\right)& y_1^{\′}\left(t\right)\\ x_2^{\′}\left(t\right)& y_2^{\′}\left(t\right)\\ \·& \·\\ \·& \·\\ \·& \·\\ x_j^{\′}\left(t\right)& y_j^{\′}\left(t\right)\\ \·& \·\\ \·& \·\\ \·& \·\\ x_N^{\′}\left(t\right)& y_N^{\′}\left(t\right)\end{array}\right];$

Step 10: judge whether Gossip algorithm reaches iterations set in advance, if reaching iteration set in advance The most whole algorithm of number of times completes, and in wireless sensor network, all anchor nodes complete distributed common recognition location, and each anchor node obtains Obtaining identical unknown node location estimation value, the final unknown node location estimation value that the method is obtained is

$z^{\′}=\left[\begin{array}{cc}{x}^{\′}& y^{\′}\end{array}\right]=\left[\begin{array}{cc}{x}_1^{\′}& y_1^{\′}\\ x_2^{\′}& y_2^{\′}\\ \·& \·\\ \·& \·\\ \·& \·\\ x_j^{\′}& y_j^{\′}\\ \·& \·\\ \·& \·\\ \·& \·\\ x_N^{\′}& y_N^{\′}\end{array}\right]=\frac{1}{N}\stackrel{\→}{1}{\stackrel{\→}{1}}^T\left[\begin{array}{cc}{x}_1^{\′}\left(0\right)& y_1^{\′}\left(0\right)\\ x_2^{\′}\left(0\right)& y_2^{\′}\left(0\right)\\ \·& \·\\ \·& \·\\ \·& \·\\ x_j^{\′}\left(0\right)& y_j^{\′}\left(0\right)\\ \·& \·\\ \·& \·\\ \·& \·\\ x_N^{\′}\left(0\right)& y_N^{\′}\left(0\right)\end{array}\right]$

I.e. complete a kind of based on the difference time of advent and the wireless-sensor network distribution type localization method of Gossip algorithm.

Invention effect: by for the poor and wireless sensor network of Gossip algorithm based on the time of advent proposed by the invention The simulation analysis of network Distributed localization method, can obtain as drawn a conclusion: (1) anchor node number in simulation process sets respectively When being set to 5,10,20,30,40,50, positioning precision improves 23.37%, 58.63%, 75.88%, 77.84%, 83.53% respectively, 84.81%, thus confirm that algorithm proposed by the invention can actually be effectively improved the positioning precision of former algorithm, and according to Simulation analysis data are it can be seen that increasing along with anchor node number, and positioning precision increases the most therewith, but increasing degree with Reduction；(2) in simulation process, measure time error standard deviation and be respectively set to 3.3ns, 6.7ns, 10.0ns, 13.3ns, 16.7ns, the range error standard deviation being namely converted into correspondence is respectively 1m, when 2m, 3m, 4m, 5m, positioning accurate Degree improves 78.38%, 77.61%, 76.80%, 76.06%, 76.68% respectively, equally confirms proposed by the invention Algorithm really can be effectively improved positioning precision, and positioning precision substantially meets the rule increased and reduce along with time determination error Rule；(3) algorithm proposed by the invention can complete distributed common recognition location, say, that in network each anchor node by Help in paired Gossip algorithm obtains the identical estimated value for unknown node position, thus is military accurate Hit the wireless sensor network exemplary operation such as invasion unknown node or the real-time illegal unknown radio station monitoring access frequency spectrum to lead Good basis is established in territory.

The present invention prepares unknown node distributed Passive Positioning field sight transferred to wireless sensor network as background. Invention uses TDOA location algorithm as basic fixed position technology, can randomly choose adjacent by means of Gossip algorithm simultaneously Node carries out information exchange and is finally reached the feature of distributed average common recognition, and brand-new design goes out based on Gossip mechanism The wireless-sensor network distribution type difference positioning algorithm time of advent.Compared with traditional location algorithm, this algorithm not only possesses essence The really advantage such as positioning precision, excellent positioning performance, and go for the rare environment of such as communication spectrum completely for illegally The radio set using frequency spectrum such as is quickly accurately positioned at the special engineering application background, can be made respectively by the algorithm of this brand-new design Individual anchor node obtains the location estimation value coordinate of identical unknown node by Gossip algorithm, thus completes distributed Common recognition location.

Present invention is generally directed in the engineer applied with wireless sensor network as background.Such as, its typical case's application scenarios can To be the region for certain communication spectrum scarcity of resources, the random or a number of monitoring anchor node of artificial layout, non-to prevent Method the unknown radio set accesses frequency spectrum.When there being unknown radio set illegally to access frequency spectrum, we monitors anchor node and receives the unknown The radio signal that radio set is launched, and record the time receiving signal.We monitors anchor node and uses Gossip afterwards Algorithm exchanges coordinate and time data at random, runs TDOA location algorithm, it is thus achieved that one right on each monitoring anchor node Estimation position in unknown radio set.Hereafter our all monitoring anchor nodes participate in a Gossip process, are running After Gossip iteration updates, each monitoring anchor node reaches distributed average common recognition, say, that each monitoring anchor node Obtain a most accurate and identical unknown station location estimated coordinates, thus it is accurate to complete distributed common recognition Location, lays a good foundation for implementing the illegal unknown radio set arresting access frequency spectrum further.

Additionally want especially it is emphasized that the positioning precision important indicator of interest that is location algorithm at this, general weigh fixed The conventional metric of position precision is mean square error (MSE) or the root mean square error (RMSE) of location solution, imitating of this algorithm Very launch for root mean square error.Mean square error and root mean square error are given below in two-dimensional space location estimation Computational methods:

$\mathrm{MSE}=\frac{1}{\mathrm{MC}}\underset{h=1}{\overset{\mathrm{MC}}{\mathrm{\Σ}}}{(x-x_h)}^2+{(y-y_h)}^2---\left(1\right)$

$\mathrm{RMSE}=\sqrt{\frac{1}{\mathrm{MC}}\underset{h=1}{\overset{\mathrm{MC}}{\mathrm{\Σ}}}{(x-x_h)}^2+{(y-y_h)}^2}---\left(2\right)$

Wherein, MC is the Monte Carlo simulation number of times carried out, and (x y) is unknown node actual position, (x_h,y_h) it is to imitate for the h time Really test the estimated value for unknown node position obtained.

The simulation result done by the present invention and analysis, it can be seen that the Taylor based on Gossip that invention is proposed is fixed Position algorithm (PGA-Taylor) can obtain the positioning precision of optimum.Thus Taylor based on Gossip proposed by the invention Location algorithm strictly one can obtain less positioning precision, and is capable of the distributed common recognition of wireless sensor network The outstanding location algorithm of one of location.Its effectiveness and superiority is it turned out, it may be said that this algorithm is one by simulating, verifying Plant the location algorithm being worthy to be popularized in wireless sensor network positioning field.

Accompanying drawing explanation

Fig. 1 is the flow chart of the PGA-Taylor algorithm that the present invention proposes；

Fig. 2 is change curve when increasing with anchor node number of the root mean square error in detailed description of the invention one；WhereinRepresent Taylor,Represent PGA-Taylor；

Fig. 3 is root mean square error change curve when increasing with surveyed time error standard deviation；WhereinRepresent Taylor,Represent PGA-Taylor.

Detailed description of the invention

Detailed description of the invention one: the wireless sensor network based on the difference time of advent and Gossip algorithm of present embodiment divides Cloth localization method realizes according to the following steps:

Step one: the random or N number of anchor node of artificial layout in wireless sensor network coverage, each anchor node is borrowed GPS or the artificial mode arranged is helped to obtain self unknown coordinates (x_k,y_k,)；

Step 2: in whole wireless sensor network, all anchor nodes realize Distributed Time together by broadcast Gossip algorithm Step；

Step 3: each anchor node wakes up up at each slotted random, whether monitoring unknown node occurs；

Step 4: if unknown node occurs, then wake up anchor node k up and will receive the radio signal that unknown node is launched, anchor Node k record receives the moment t of signal_k, it is saved together (x together with local coordinate system_k,y_k,t_k)；

Step 5: each anchor node wakes up up successively, repeats step 4, until all anchor nodes all complete the signal of unknown node Monitoring and data preserve；

Step 6: the anchor node in wireless sensor network wakes up up at random, it is assumed that anchor node j is in wake-up states, this anchor saves The point all adjacent anchor nodes broadcast solicitations in the range of communication radius, it is desirable to it is returned node coordinate and receives the signal moment, This anchor node starts to receive data and whether real-time judge receives the data of whole adjacent anchor node, if not receiving the most adjacent Anchor node data, then continue to, if receiving whole adjacent anchor node data, then completes the reception of data and by anchor node J receives the data of its all M adjacent anchor nodes and is saved in this locality, obtains local data as follows

$\left[\begin{array}{ccc}{x}_0^j& y_0^j& t_0^j\end{array}\right]=\left[\begin{array}{ccc}{x}_1& y_1& t_1\\ x_1& y_2& t_2\\ \·& \·& \·\\ \·& \·& \·\\ \·& \·& \·\\ x_i& y_i& t_i\\ \·& \·& \·\\ \·& \·& \·\\ \·& \·& \·\\ x_M& y_M& t_M\end{array}\right];$

Step 7: the local data that anchor node j obtains according to step 6, according to the Taylor algorithm positioned based on TDOA Primary iteration location formula obtains unknown node position primary iteration value ((x_j(0),y_j(0)), run according to the thresholding pre-set The iterative process of Taylor algorithm, until obtaining anchor node j initial estimate ((x for unknown node position_j(0),y_j(0))；

Step 8: remaining anchor node in wireless sensor network wakes up up at random, repeats step according to the working method of anchor node j Rapid six and step 7 obtain this anchor node initial estimate for unknown node position, until all anchor nodes complete for not Knowing the initial estimation of node location, in wireless sensor network, the unknown node position initial estimate coordinate of all anchor nodes is

$\left[\begin{array}{cc}{x}^{\′}\left(0\right)& y^{\′}\left(0\right)\end{array}\right]=\left[\begin{array}{cc}{x}_1^{\′}\left(0\right)& y_1^{\′}\left(0\right)\\ x_2^{\′}\left(0\right)& y_2^{\′}\left(0\right)\\ \·& \·\\ \·& \·\\ \·& \·\\ x_j^{\′}\left(0\right)& y_j^{\′}\left(0\right)\\ \·& \·\\ \·& \·\\ \·& \·\\ x_N^{\′}\left(0\right)& y_N^{\′}\left(0\right)\end{array}\right];$

Step 9: all anchor nodes obtain this node initial estimation for unknown node position in wireless sensor network On the basis of value, all N number of anchor nodes of the whole network run paired Gossip algorithm, paired Gossip algorithm carrying out practically process As follows: to assume to be waken up at time slot t anchor node j, this anchor node randomly chooses an adjacent anchor node i exchange initial estimation Value Data

$x_j^{\′}(t+1)=x_i^{\′}(t+1)=\frac{1}{2}(x_i^{\′}\left(t\right)+x_j^{\′}\left(t\right))$

$y_j^{\′}(t+1)=y_i^{\′}(t+1)=\frac{1}{2}(y_i^{\′}\left(t\right)+y_j^{\′}\left(t\right))$

The whole iterative process of paired Gossip algorithm that the N number of anchor node of the whole network is run can be expressed as follows

$z^{\′}(t+1)=W\left(t\right)z^{\′}\left(t\right)=\underset{n=1}{\overset{t}{\mathrm{\Σ}}}W\left(n\right)z^{\′}\left(0\right)$

Wherein

$z^{\′}\left(t\right)=\left[\begin{array}{cc}{x}^{\′}\left(t\right)& y^{\′}\left(t\right)\end{array}\right]=\left[\begin{array}{cc}{x}_1^{\′}\left(t\right)& y_1^{\′}\left(t\right)\\ x_2^{\′}\left(t\right)& y_2^{\′}\left(t\right)\\ \·& \·\\ \·& \·\\ \·& \·\\ x_j^{\′}\left(t\right)& y_j^{\′}\left(t\right)\\ \·& \·\\ \·& \·\\ \·& \·\\ x_N^{\′}\left(t\right)& y_N^{\′}\left(t\right)\end{array}\right];$

Step 10: judge whether Gossip algorithm reaches iterations set in advance, if reaching iteration set in advance The most whole algorithm of number of times completes, and in wireless sensor network, all anchor nodes complete distributed common recognition location, and each anchor node obtains Obtaining identical unknown node location estimation value, the final unknown node location estimation value that the method is obtained is

$z^{\′}=\left[\begin{array}{cc}{x}^{\′}& y^{\′}\end{array}\right]=\left[\begin{array}{cc}{x}_1^{\′}& y_1^{\′}\\ x_2^{\′}& y_2^{\′}\\ \·& \·\\ \·& \·\\ \·& \·\\ x_j^{\′}& y_j^{\′}\\ \·& \·\\ \·& \·\\ \·& \·\\ x_N^{\′}& y_N^{\′}\end{array}\right]=\frac{1}{N}\stackrel{\→}{1}{\stackrel{\→}{1}}^T\left[\begin{array}{cc}{x}_1^{\′}\left(0\right)& y_1^{\′}\left(0\right)\\ x_2^{\′}\left(0\right)& y_2^{\′}\left(0\right)\\ \·& \·\\ \·& \·\\ \·& \·\\ x_j^{\′}\left(0\right)& y_j^{\′}\left(0\right)\\ \·& \·\\ \·& \·\\ \·& \·\\ x_N^{\′}\left(0\right)& y_N^{\′}\left(0\right)\end{array}\right]$

I.e. complete a kind of based on the difference time of advent and the wireless-sensor network distribution type localization method of Gossip algorithm.

Present embodiment, in step 2, broadcast Gossip algorithm is the one in Gossip algorithm, is different from paired Gossip Algorithm, it is impossible to ensure finally to converge on the average of all initial values, its typical case's application completes distributed clock exactly and synchronizes；

In step 10, selecting according to the accuracy selection Gossip algorithm of concrete network topology structure and demand of iterations Iterations, it is ensured that algorithm can converge on initial mean value.

Detailed description of the invention two: present embodiment is unlike detailed description of the invention one: the broadcast described in step 2 Gossip algorithm particularly as follows:

Node in network wakes up up at random and broadcasts its state value, and this state value is connect by all nodes in its communication radius Receiving, all receiving nodes are according to its local state value of algorithm design update, and other node state values keep constant.Other step And parameter is identical with detailed description of the invention one.

Detailed description of the invention three: present embodiment is unlike detailed description of the invention one or two:

One, TDOA location

Assume initially that all anchor nodes reach Distributed Time by broadcast Gossip algorithm and synchronize, the communication of anchor node half Footpath is limited by TDOA location hardware module in network, and topological structure G (N, R) of N point wireless sensor network is by adjoining Matrix Φ_ijRepresent, as i ≠ j, if anchor node i and anchor node j is adjacent, then Φ_ij=1；Otherwise, Φ_ij=0；For Anchor node j, definition M={i ∈ 1,2 ..., N}: Φ_ij≠0}；

TDOA location algorithm is called again hyperbola location algorithm, and its mathematics mechanism is to ask for hyp intersection point thus obtains The estimation position of unknown node；For waking up anchor node j up, its Hyperbolic Equation is

$R_{i,j}=R_i-R_j=\sqrt{{(x_i-x)}^2+{(y_i-y)}^2}-\sqrt{{(x_j-x)}^2+{(y_j-y)}^2}---\left(3\right)$

Wherein R_iFor the distance of unknown node to i-th anchor node, R_j(definition anchor node j is to wake up anchor node up for signal source arrival Service wakes up anchor node up) distance, R_i,jFor unknown node to i-th anchor node and the range difference waking up anchor node j up, R_i,j=c(t_i-t_j), wherein c is the light velocity, t_iAnd t_jIt is respectively node i and node j receives unknown node and launches signal Moment；Make i=1 wherein, 2 ..., M, (x y) is the coordinate of unknown node, (x_i,y_i) it is the position coordinates of i-th anchor node, (x_j,y_j) for waking up anchor node position coordinates up；

Two, Taylor algorithm

Hyperbolic Equation (3) carrying out linearisation, uses Taylor series expansion algorithms, Taylor series expansion algorithms is one Plant the recursive algorithm needing signal source initial estimated location, by solving the office of TDOA measurement error in recurrence each time Portion's least square solution improves the estimation position to signal source.In actual applications, Taylor Series Expansion Method is generally of relatively Good positioning performance.

Taylor sequence expansion needs initial estimated location, by solving TDOA measurement error in recurrence each time Local least square method solution comes improved estimator position.For one group of TDOA measured value, first this algorithm chooses primary iteration position Put, choose the most as follows

$p={\left(G_a^T{Q}^{-1}{G}_a\right)}^{-1}{G}_a^T{Q}^{-1}h---\left(4\right)$

Wherein, Q is the covariance matrix of TDOA

$h=\frac{1}{2}\left[\begin{array}{c}{R}_{1,j}^2-x_1^2-y_1^2+x_j^2+y_j^2\\ R_{2,j}^2-x_2^2-y_2^2+x_j^2+y_j^2\\ \·\\ \·\\ \·\\ R_{M,j}^2-x_M^2-y_M^2+x_j^2+y_j^2\end{array}\right],G_a=-\left[\begin{array}{ccc}{x}_{1,j}& y_{1,j}& R_{1,j}\\ x_{2,j}& y_{2,j}& R_{2,j}\\ \·& \·& \·\\ \·& \·& \·\\ \·& \·& \·\\ x_{M,j}& y_{M,j}& R_{M,j}\end{array}\right]$

(x in formula_i,j,y_i,j) it is i-th anchor node and the coordinate difference waken up up between anchor node j.

Taylor series expansion algorithms is chosen as follows for the primary iteration value of unknown node position

$\left\{\begin{array}{c}{x}_j\left(0\right)=p\left(1\right)\\ y_j\left(0\right)=p\left(2\right)\end{array}\right.---\left(5\right)$

Recurrence makes next time:

x′_j(0)=x_j(0)+△x,y′_j(0)=y_j(0)+△y (6)

△ x, △ y choose as follows

$\mathrm{\δ}=\left[\begin{array}{c}\mathrm{\Δx}\\ \mathrm{\Δy}\end{array}\right]={\left(G_t^T{Q}^{-1}{G}_t\right)}^{-1}{G}_t^T{Q}^{-1}{h}_t---\left(7\right)$

Wherein

$h_t=\left[\begin{array}{c}{R}_{1,j}-R_1+R_j\\ R_{2,j}-R_2+R_j\\ \·\\ \·\\ \·\\ R_{M,j}-R_M+R_j\end{array}\right],G_t=\left[\begin{array}{cc}(x_j-x)/R_j-(x_1-x)/R_1& (y_j-y)/R_j-(y_1-y)/R_1\\ (x_j-x)/R_j-(x_2-x)/R_2& (y_j-y)/R_j-(y_2-y)/R_2\\ \·& \·\\ \·& \·\\ \·& \·\\ (x_j-x)/R_j-(x_M-x)/R_M& (y_j-y)/R_j-(y_M-y)/R_M\end{array}\right]$

Repeat above procedure, until △ x, △ y be sufficiently small, meet certain thresholding:

Wherein ε is a positive number the least, ((x ' now_j(0),y′_j(0) anchor node j initially estimating for unknown node position) it is Evaluation.

Other step and parameter are identical with detailed description of the invention one or two.

Detailed description of the invention four: present embodiment is unlike one of detailed description of the invention one to three:

Assume that wireless sensor network has N number of sensor node, each sensor node j to have a state value a in t_j(t), t=1,2,…；

Need to obtain the state value of oneself at Gossip algorithm incipient stage i.e. t=0, each sensor node j according to task a_j(0), this value can be the value of information of any pre-monitoring；Subsequently, in wireless sensor network, any sensor node j is permissible Moment t by Random Activation and and it select at random possess state value a_iThe adjacent sensors node i swap status value of (t), Exemplary formula is

$a_j(t+1)=a_i(t+1)=\frac{1}{2}(a_i\left(t\right)+a_j\left(t\right))---\left(9\right)$

The state of all the sensors node in wireless sensor network is represented by N-dimensional vector a (t), for the ease of grinding Study carefully, (9) formula be rewritten into vector form:

a(t+1)=W(t)a(t) (10) Wherein W (t) is a random matrix, and the sensor node that it is waken up in depending primarily on time slot t has thus obtained warp The general mathematics model of the paired Gossip algorithm of allusion quotation.From algorithm model it can be seen that classical clean culture Gossip algorithm in pairs Realize very simple, and key issue is its convergence have also been obtained theoretical proof., the purpose of Gossip algorithm It is through the fewest iterations, makes the state value of each node j in network all converge on all the sensors node initial The average of state value $(1/N){\mathrm{\Σ}}_{m=1}^N{a}_m\left(0\right).$

Other step and parameter are identical with one of detailed description of the invention one to three.

Emulation experiment:

This emulation carries out location algorithm emulation in the two dimensional surface of 100m × 100m, it is assumed that unknown node position coordinates is (20m,20m).TDOA time determination error meets the Gauss distribution of zero-mean, and Taylor algorithm iteration thresholding is set to ε=10^-8。 Figure of description 2 is respectively to anchor node number N=5, and the situation of 10,20,30,40,50 emulates, this emulation anchor node position Put random arrangement in the range of 100m × 100m.The present invention draws original Taylor algorithm and based on Gossip respectively Taylor location algorithm (PGA-Taylor) is along with root-mean-square error curve when anchor node number increases, the number in whole emulation According to being by the result of acquisition after 1000 independent experiments are averaged.This emulation experiment is anchor node number in simulation process When being respectively set to 5,10,20,30,40,50, positioning precision improves 23.37%, 58.63%, 75.88%, 77.84% respectively, 83.53%, 84.81%, thus confirm that PGA-Taylor algorithm proposed by the invention can actually be effectively improved former algorithm Positioning precision, and according to simulation analysis data it can be seen that increasing along with anchor node number, positioning precision increases the most therewith Greatly, but increasing degree reduces therewith.Figure of description 3 is fixed anchor nodes number N=20 and anchor node position also exists In the range of 100m × 100m during random distribution, root-mean-square error was carried out with time determination error standard deviation in the case of of change Emulation.Still respectively Taylor algorithm and PGA-Taylor algorithm is along with root-mean-square error curve when time determination error becomes big, Whole emulation experiment be also by 1000 independent experiments average after result.In this simulation process, measure the time by mistake Difference standard deviation is respectively set to 3.3ns, 6.7ns, 10.0ns, 13.3ns, 16.7ns, is namely converted into the range finding of correspondence by mistake Difference standard deviation is respectively 1m, and when 2m, 3m, 4m, 5m, positioning precision improves 78.38%, 77.61%, 76.80% respectively, 76.06%, 76.68%, equally confirm that algorithm proposed by the invention can be effectively improved positioning precision really, and Positioning precision substantially meets the rule increased and reduce along with time determination error.

Claims (3)

1., based on the difference time of advent and a wireless-sensor network distribution type localization method for Gossip algorithm, its feature exists Realize according to the following steps in wireless-sensor network distribution type localization method based on the difference time of advent and Gossip algorithm:

Step one: random or artificial arrange N number of anchor node in wireless sensor network coverage, each anchor node by GPS or the artificial mode arranged obtains self unknown coordinates (x_k,y_k,)；

Step 2: in whole wireless sensor network, all anchor nodes realize Distributed Time synchronization by broadcast Gossip algorithm；

Step 3: each anchor node wakes up up at each slotted random, whether monitoring unknown node occurs；

Step 4: if unknown node occurs, then wake up anchor node k up and will receive the radio signal that unknown node is launched, anchor Node k record receives the moment t of signal_k, it is saved together (x together with local coordinate system_k,y_k,t_k)；

Step 5: each anchor node wakes up up successively, repeats step 4, until all anchor nodes all complete the signal of unknown node Monitoring and data preserve；

Step 6: the anchor node in wireless sensor network wakes up up at random, it is assumed that anchor node j is in wake-up states, this anchor saves The point all adjacent anchor nodes broadcast solicitations in the range of communication radius, it is desirable to it is returned node coordinate and receives the signal moment, This anchor node starts to receive data and whether real-time judge receives the data of whole adjacent anchor node, if not receiving the most adjacent Anchor node data, then continue to, if receiving whole adjacent anchor node data, then completes the reception of data and by anchor node J receives the data of its all M adjacent anchor nodes and is saved in this locality, obtains local data as follows

$\[\begin{array}{ccc}{x}_0^j& y_0^j& t_0^j\end{array}\]=\left[\begin{array}{ccc}{x}_1& y_1& t_1\\ x_1& y_2& t_2\\ .& .& .\\ .& .& .\\ .& .& .\\ x_i& y_i& t_i\\ .& .& .\\ .& .& .\\ .& .& .\\ x_M& y_M& t_M\end{array}\right];$

Step 7: the local data that anchor node j obtains according to step 6, at the beginning of the Taylor algorithm positioned based on TDOA Beginning iterative position formula obtains unknown node position primary iteration value (x_j(0),y_j(0)), run according to the thresholding pre-set The iterative process of Taylor algorithm, until obtaining the anchor node j initial estimate (x for unknown node position_j(0),y_j(0))；

Described Taylor algorithm based on TDOA location particularly as follows:

One, TDOA location

Assume initially that all anchor nodes reach Distributed Time by broadcast Gossip algorithm and synchronize, the communication radius of anchor node Being limited by TDOA location hardware module in network, topological structure G (N, R) of N point wireless sensor network is by adjacency matrix Φ_ijRepresent, as i ≠ j, if anchor node i and anchor node j is adjacent, then Φ_ij=1；Otherwise, Φ_ij=0；Anchor is saved Point j, definition M={i ∈ 1,2 ..., N}: Φ_ij≠0}；

For waking up anchor node j up, its Hyperbolic Equation is

$R_{i,j}=R_i-R_j=\sqrt{{(x_i-x)}^2+{(y_i-y)}^2}-\sqrt{{(x_j-x)}^2+{(y_j-y)}^2}---\left(1\right)$

Wherein R_iFor the distance of unknown node to i-th anchor node, R_jThe distance waking up anchor node up, R is arrived for signal source_i,jFor the unknown Node is to i-th anchor node and the range difference waking up anchor node j up, R_i,j=c (t_i-t_j), wherein c is the light velocity, t_iAnd t_jRespectively Receive unknown node for node i and node j and launch the moment of signal；Make i=1 wherein, 2 ..., M, (x y) is unknown joint The coordinate of point, (x_i,y_i) it is the position coordinates of i-th anchor node, (x_j,y_j) for waking up anchor node position coordinates up；

Two, Taylor algorithm

Using Taylor algorithm that Hyperbolic Equation (1) carries out linearisation, Taylor sequence expansion needs initial estimated location, In recurrence each time, the local least square method solution by solving TDOA measurement error comes improved estimator position, for one group TDOA measured value, first this algorithm is chosen primary iteration position, is chosen the most as follows

$p={\left(G_a^T{Q}^{-1}{G}_a\right)}^{-1}{G}_a^T{Q}^{-1}h---\left(2\right)$

Wherein, Q is the covariance matrix of TDOA

$h=\frac{1}{2}\left[\begin{array}{c}{R}_{1,j}^2-x_1^2-y_1^2+x_j^2+y_j^2\\ R_{2,j}^2-x_2^2-y_2^2+x_j^2+y_j^2\\ .\\ .\\ .\\ R_{M,j}^2-x_M^2-y_M^2+x_j^2+y_j^2\end{array}\right],G_a=-\left[\begin{array}{ccc}{x}_{1,j}& y_{1,j}& R_{1,j}\\ x_{2,j}& y_{2,j}& R_{2,j}\\ .& .& .\\ .& .& .\\ .& .& .\\ x_{M,j}& y_{M,j}& R_{M,j}\end{array}\right]$

(x in formula_i,j,y_i,j) it is i-th anchor node and the coordinate difference waken up up between anchor node j；

Taylor algorithm is chosen as follows for the primary iteration value of unknown node position

$\left\{\begin{array}{c}{x}_j\left(0\right)=p\left(1\right)\\ y_j\left(0\right)=p\left(2\right)\end{array}\right.---\left(3\right)$

Recurrence makes next time:

x′_j(0)=x_j(0)+Δx,y′_j(0)=y_j(0)+Δy (4)

Δ x, Δ y choose as follows

$\mathrm{\δ}=\left[\begin{array}{c}\mathrm{\Δ}x\\ \mathrm{\Δ}y\end{array}\right]={\left(G_t^T{Q}^{-1}{G}_t\right)}^{-1}{G}_t^T{Q}^{-1}{h}_t---\left(5\right)$

Wherein

$h_t=\left[\begin{array}{c}{R}_{1,j}-R_1+R_j\\ R_{2,j}-R_2+R_j\\ .\\ .\\ .\\ R_{M,j}-R_M+R_j\end{array}\right],G_t=\left[\begin{array}{cc}(x_j-x)/R_j-(x_1-x)/R_1& (y_j-y)/R_j-(y_1-y)/R_1\\ (x_j-x)/R_j-(x_2-x)/R_2& (y_j-y)/R_j-(y_2-y)/R_2\\ .& .\\ .& .\\ .& .\\ (x_j-x)/R_j-(x_M-x)/R_M& (y_j-y)/R_j-(y_M-y)/R_M\end{array}\right]$

Repeat above procedure, until Δ x, Δ y are sufficiently small, meet and set thresholding:

Δ x+ Δ y ＜ ε (6)

Wherein ε is a positive number the least, (x ' now_j(0),y′_j(0) the anchor node j initial estimation for unknown node position) it is Value；

Step 8: remaining anchor node in wireless sensor network wakes up up at random, repeats step according to the working method of anchor node j Rapid six and step 7 obtain this anchor node initial estimate for unknown node position, until all anchor nodes complete for not Knowing the initial estimation of node location, in wireless sensor network, the unknown node position initial estimate coordinate of all anchor nodes is

$\[x^{\′}\left(0\right)y^{\′}\left(0\right)\]=\left[\begin{array}{cc}{x}_1^{\′}\left(0\right)& y_1^{\′}\left(0\right)\\ x_2^{\′}\left(0\right)& y_2^{\′}\left(0\right)\\ .& .\\ .& .\\ .& .\\ x_j^{\′}\left(0\right)& y_j^{\′}\left(0\right)\\ .& .\\ .& .\\ .& .\\ x_N^{\′}\left(0\right)& y_N^{\′}\left(0\right)\end{array}\right];$

Step 9: all anchor nodes obtain this node initial estimate for unknown node position in wireless sensor network On the basis of, all N number of anchor nodes of the whole network run paired Gossip algorithm, and paired Gossip algorithm carrying out practically process is such as Under: assuming to be waken up at time slot t anchor node j, this anchor node randomly chooses an adjacent anchor node i exchange initial estimate number According to

$x_j^{\′}(t+1)=x_i^{\′}(t+1)=\frac{1}{2}\left(x_i^{\′}\right(t)+x_j^{\′}(t\left)\right)$

$y_j^{\′}(t+1)=y_i^{\′}(t+1)=\frac{1}{2}\left(y_i^{\′}\right(t)+y_j^{\′}(t\left)\right)$

The whole iterative process of paired Gossip algorithm that the N number of anchor node of the whole network is run can be expressed as follows

$z^{\′}(t+1)=W\left(t\right)z^{\′}\left(t\right)=\underset{n=1}{\overset{t}{\Σ}}W\left(n\right)z^{\′}\left(0\right)$

Wherein

$z^{\′}\left(t\right)=\[x^{\′}\left(t\right)y^{\′}\left(t\right)\]=\left[\begin{array}{cc}{x}_1^{\′}\left(t\right)& y_1^{\′}\left(t\right)\\ x_2^{\′}\left(t\right)& y_2^{\′}\left(t\right)\\ .& .\\ .& .\\ .& .\\ x_j^{\′}\left(t\right)& y_j^{\′}\left(t\right)\\ .& .\\ .& .\\ .& .\\ x_N^{\′}\left(t\right)& y_N^{\′}\left(t\right)\end{array}\right];$

Step 10: judge whether Gossip algorithm reaches iterations set in advance, if reaching iteration set in advance time Several, whole algorithm completes, and in wireless sensor network, all anchor nodes complete distributed common recognition location, and each anchor node obtains Identical unknown node location estimation value, the final unknown node location estimation value that the method is obtained is

$z^{\′}=\[x^{\′}{y}^{\′}\]=\left[\begin{array}{cc}{x}_1^{\′}& y_1^{\′}\\ x_2^{\′}& y_2^{\′}\\ .& .\\ .& .\\ .& .\\ x_j^{\′}& y_j^{\′}\\ .& .\\ .& .\\ .& .\\ x_N^{\′}& y_N^{\′}\end{array}\right]=\frac{1}{N}\stackrel{\→}{1}{\stackrel{\→}{1}}^T\left[\begin{array}{cc}{x}_1^{\′}\left(0\right)& y_1^{\′}\left(0\right)\\ x_2^{\′}\left(0\right)& y_2^{\′}\left(0\right)\\ .& .\\ .& .\\ .& .\\ x_j^{\′}\left(0\right)& y_j^{\′}\left(0\right)\\ .& .\\ .& .\\ .& .\\ x_N^{\′}\left(0\right)& y_N^{\′}\left(0\right)\end{array}\right]$

I.e. complete a kind of based on the difference time of advent and the wireless-sensor network distribution type localization method of Gossip algorithm.

A kind of wireless sensor network distribution based on the difference time of advent and Gossip algorithm the most according to claim 1 Formula localization method, it is characterised in that the broadcast Gossip algorithm described in step 2 particularly as follows:

Node in network wakes up up at random and broadcasts its state value, and this state value is received by all nodes in its communication radius, All receiving nodes are according to its local state value of algorithm design update, and other node state values keep constant.

The most according to claim 1 and 2 a kind of based on the difference time of advent and the wireless sensor network of Gossip algorithm Distributed localization method, it is characterised in that paired Gossip algorithm described in step 9 particularly as follows:

Assume that wireless sensor network has N number of sensor node, each sensor node j to have a state value a in t_j(t), T=1,2 ...；

Need to obtain the state value a of oneself at Gossip algorithm incipient stage i.e. t=0, each sensor node j according to task_j(0)； Subsequently, in wireless sensor network, any sensor node j can be at moment t by Random Activation and its most selected possessing State value a_iT the adjacent sensors node i swap status value of (), exemplary formula is

$a_j(t+1)=a_i(t+1)=\frac{1}{2}\left(a_i\right(t)+a_j(t\left)\right)---\left(7\right)$

The state of all the sensors node in wireless sensor network is represented by N-dimensional vector a (t), (7) formula is rewritten The form of one-tenth vector:

A (t+1)=W (t) a (t) (8)

Wherein W (t) is a random matrix, the sensor node that it is waken up in depending primarily on time slot t, obtains classical paired The general mathematics model of Gossip algorithm, Gossip algorithm, by the fewest iterations, makes the shape of each node in network State value all converges on the average of all the sensors node initial state value