CN101695190B - Three-dimensional wireless sensor network node self-locating method based on neural network - Google Patents
Three-dimensional wireless sensor network node self-locating method based on neural network Download PDFInfo
- Publication number
- CN101695190B CN101695190B CN2009102363723A CN200910236372A CN101695190B CN 101695190 B CN101695190 B CN 101695190B CN 2009102363723 A CN2009102363723 A CN 2009102363723A CN 200910236372 A CN200910236372 A CN 200910236372A CN 101695190 B CN101695190 B CN 101695190B
- Authority
- CN
- China
- Prior art keywords
- node
- self
- wireless sensor
- sensor network
- neural net
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Expired - Fee Related
Links
Images
Landscapes
- Mobile Radio Communication Systems (AREA)
Abstract
The invention discloses a three-dimensional wireless sensor network node self-locating method based on a neural network. The three-dimensional wireless sensor network node self-locating method comprises the following steps that firstly initializing a wireless sensor network, secondly establishing a neural network, thirdly extracting training examples for the neural network, fourthly training the neural network through the obtained training examples, fifthly obtaining the actual distance from each unknown node to an anchor node which is not adjacent to the unknown node according to the well-trained neural network, and sixthly obtaining the three-dimensional coordinates of the known node. The obtaining method of the training examples is simple and has strong representative, and the well-trained neural network can better represent the main properties of a geometric structure of the three-dimensional wireless sensor network. The three-dimensional wireless sensor network node self-locating method solves the problem of overlarge estimation error of distance brought by the accumulation system of shortest jumping distances in the locating process of the three-dimensional wireless sensor, and effectively improves the locating precision.
Description
Technical field
The node that the present invention relates to wireless sensor network in the wireless sensor network node self-align field, particularly three dimensions is self-align, a kind of 3-D wireless sensor network node method for self-locating based on neural net.
Background technology
Along with the development of technology such as wireless telecommunications, transducer, MEMS (micro electro mechanical system), digital and electronic, data-centered wireless sensor network (WSN) has become one of research focus of present IT field.Wireless sensor node has data acquisition, processing and function such as communicate by letter, and the various environmental informations in monitoring in real time, perception and the processing wireless sensor network distributed areas send the terminal use of required information then to.Wireless sensor network all has purposes very widely in many-sides such as military security, environmental monitoring and forecast, Industry Control, communications and transportation, Smart Home, logistics management, reading intelligent agriculture and medical treatment and nursing.
The sensor node self align technology is one of critical support technology of wireless sensor network.In actual applications, positional information is the important component part of sensor node monitoring information.For example, the wireless sensor network of monitoring risk of forest fire not only needs to notify at once fire condition when monitoring the condition of a fire, also need to report the geographical position of fire generation, the spot so that fire department can arrive in the very nick of time; When leaking appearred in natural gas line, the sensor node that is deployed on the pipe network also needed to provide the position of concrete leakage except reporting the leakage information; Deployment wireless sensor network afield only provides enemy's particular location, could implement strike accurately etc. to it, and the realization of these functions all requires sensor node to know the position of self in advance.Yet, in the wireless sensor network practical application, sensor node is disposed (dispensing as aircraft) usually at random and carry out various monitoring tasks in different environment, and the position of self can't be determined in advance, so node at first needs to carry out in real time self-align after deployment.Location technology is also significant for the research of the location-based procotol of wireless sensor network (as network management, geographical route etc.).
The self-align node (anchor node) that is meant the node (unknown node) of self-position the unknown according to the minority known location of sensor node is determined self position according to certain location mechanism.According to the distance (or angle) between the needs measured node whether in the position fixing process, localization method can be divided into based on the localization method of distance (range-based) and the localization method of range-independence (range-free).In the most methods, position fixing process mainly comprises two stages: the measurement of (1) euclidean distance between node pair (or angle): the technology of employing mainly contains received signal intensity indication (RSSI), time of arrival (toa)/time difference (TOA/TDOA), signal arrives angle (AOA) etc.; When node does not have distance measurement function, can use internodal jumping figure to replace actual range.(2) obtain the unknown node coordinate: when unknown node acquire some anchor nodes apart from the time, can use three (four) limit methods, three (four) horn cuppings, the maximum likelihood estimation technique etc. to obtain self coordinate; In addition, can also obtain the coordinate of whole unknown node by global approach such as convex programming, MDS-MAP.
In wireless sensor network, because the restriction of factors such as cost, volume, the quantity of anchor node is very limited, and the number of anchor node does not often reach and obtains the required minimum anchor node quantity of self coordinate in the unknown node communication context.Multi-hop positioning method generally uses unknown node to replace actual distance to the shortest jumping segment distance of anchor node, can solve the problem that location coverage rate that the anchor node lazy weight causes descends like this, but the distance estimations error that the mechanism of adding up of the shortest jumping segment distance is brought makes positioning accuracy descend greatly, and this influence shows particularly outstanding in three-dimensional fix.
Neural net is the most widely used a kind of artificial intelligence technology in present each field, it is a kind of intelligent network system that is interconnected to form by certain mode by a plurality of very simple processing units (neuron), it can reflect some basic functions of human brain, is the non-linear dynamic model of a high complexity.In recent years, some scholars were used for the wireless sensor network node location with neural net, had obtained comparatively desirable effect.The functional relation that Ali Shareef etc. uses three kinds of different neural nets (MLP, RBF and RNN) to approach euclidean distance between node pair and position respectively, and quantitative comparison performance separately; Probability of use neural nets (PNN) such as S.Rajaee are estimated the coordinate of unknown node, and utilize independent component analysis (ICA) to reduce required amount of calculation and energy consumption in the position fixing process; Yurong Xu etc. obtain self jumping figure coordinate (Hop-Coordinates) by structure BP neural net on each node, combine with localization method based on jumping figure then, have effectively improved the positioning accuracy of node; Yun Sukhyun etc. utilizes neural net to set up approximation relation between signal strength signal intensity and node location, realizes the node locating etc. of the comparatively simple wireless sensor network of topological structure with this.
Yet the shortest jumping segment distance was not to the influence of positioning performance when above method considered that unknown node and anchor node are non-conterminous; Just obtain the required sample of neural metwork training roughly, representativeness of sample is not strong, and the neural net after the training can not embody the main character of wireless sensor network geometry preferably, can only be applicable to the situation that topological structure is comparatively fixing mostly; In addition, above method at be the node locating problem of two-dimensional space, and wireless sensor network often is distributed in the three dimensions in the practical application, therefore the research to the self-align problem of three dimensions node has more realistic meaning.
Summary of the invention
The objective of the invention is in order to address the above problem, a kind of 3-D wireless sensor network node method for self-locating based on neural net is provided, by being configured to calculate the neural net of actual range between non-conterminous node, and extracting training sample according to the relation of the position between anchor node neural net is trained; The neural net that the unknown node utilization trains obtains self actual range to non-conterminous anchor node, and then obtains the three-dimensional coordinate of self, realizes self-align.
A kind of 3-D wireless sensor network node method for self-locating of the present invention based on neural net, realize by following steps:
Step 1: intiating radio sensor network;
Step 2: set up the neural net that is used to calculate actual range between non-conterminous node;
Step 3: according to the shortest jumping distance, d of the anchor node in the wireless sensor network to other anchor node
Min, beeline jumping figure H
Cou, the D of beeline local density
SumAnd anchor node three-dimensional coordinate separately, the training sample of extraction neural net;
Step 4: utilize the training sample that obtains in the step 3 that neural net is trained;
Step 5: according to the neural net that trains, each unknown node obtains self actual range to non-conterminous anchor node;
Step 6: unknown node obtains self three-dimensional coordinate according to the three-dimensional coordinate of anchor node and self to the actual range of anchor node.
The invention has the advantages that:
(1) the present invention has set up the neural net that is used to calculate actual range between non-conterminous node, has reduced the distance estimations error that the shortest jumping segment distance mechanism of adding up is brought, and has effectively improved 3-D wireless sensor network positions precision;
(2) according to the training sample of the relation extraction of the position between anchor node neural net, the sample obtain manner is simple and representativeness of sample is stronger, and the neural net that trains can embody the principal character of wireless sensor network geometry preferably;
(3) situation about positioning with respect to the anchor node that only uses in the unknown node communication context, unknown node is with the information substitution neural net of self record in the method, obtain self actual range to non-conterminous anchor node, when improving the location coverage rate, positioning accuracy but is not subjected to too big influence.
Description of drawings
Fig. 1 is the flow chart of a kind of 3-D wireless sensor network node method for self-locating based on neural net of the present invention;
Fig. 2 is for upgrading the flow chart of anchor node location information data frame among the present invention;
Fig. 3 is the structural representation of neural net among the present invention;
Fig. 4 is for using the flow chart that obtains the unknown node three-dimensional coordinate based on the least square method of Taylor series expansion among the present invention;
Fig. 5 is a 3-D wireless sensor network design structural representation in the embodiment of the invention;
Fig. 6 is node positioning method of the present invention and each node locating error comparison diagram of distance vector method;
Fig. 7 is average position error curve chart under the heterogeneous networks degree of communication for node positioning method of the present invention and distance vector method.
Embodiment
The present invention is described in further detail below in conjunction with accompanying drawing.
A kind of 3-D wireless sensor network node method for self-locating of the present invention based on neural net, flow process realizes by following steps as shown in Figure 1:
Step 1: intiating radio sensor network;
Each node obtains node self local density by carrying out information interaction with self neighbor node, and measures self distance to neighbor node; Location information data frame by all anchor nodes is propagated in wireless sensor network, and all nodes obtain self the shortest jumping segment distance, beeline jumping figure and beeline local density to all anchor nodes; Described node self local density is meant the number of neighbor node in the node communication context; The shortest described jumping segment distance is meant the length of the shortest path that can communicate with one another between two nodes; Described beeline jumping figure is meant the jumping hop count order that the path at the shortest jumping segment distance place between two nodes comprises; Described beeline local density is meant self local density's sum of all nodes that the path at the shortest jumping segment distance place comprises;
Concrete steps are:
(1) all node broadcasts comprise the request-reply Frame of self ID, transmission reply data frame was replied after neighbor node was received the request-reply Frame, the number of the reply data frame that each node statistics self is received obtains self local density, and measures self distance to neighbor node;
(2) all anchor nodes are broadcasted the location information data frame Frame that comprises self ID and self three-dimensional coordinate in wireless sensor network
m, m is a natural number; Frame
mForm as follows:
Frame
m={ID
m,x
m,y
m,z
m,H,d,D} (1)
Wherein, ID
mBe the ID of m anchor node, (x
m, y
m, z
m) be the three-dimensional coordinate of m anchor node; H is Frame
mThe jumping hop count order of process, d is Frame
mThe jumping segment distance sum of process, D is Frame
mLocal density's sum of all nodes of process; H, d are initialized as 0, and D is initialized as the local density of anchor node;
(3) node receives the anchor node location information data frame Frame that self neighbor node sends or transmits in wireless sensor network
mThe time, whether decision node self received the location information data frame Frame of this anchor node
m, flow process as shown in Figure 2;
1. work as node and do not receive this anchor node location information data frame Frame
mThe time, then need upgrade Frame
m, the anchor node location information data frame Frame ' after the renewal
mFor:
Frame′
m={ID
m,x
m,y
m,z
m,(H+1),(d+d′),(D+D′)} (2)
Wherein, d ' arrives to self sending or transmit Frame for node
mThe distance of neighbor node, D ' is the local density of node self, node is preserved the anchor node location information data frame Frame ' after upgrading
m, and with Frame '
mBe broadcast to whole wireless sensor network;
2. work as node and received this anchor node location information data frame Frame
mThe time, judge Frame
mIn d add the jumping segment distance sum whether d ' has preserved less than node self to this anchor node;
If less than, then according to formula (2) this anchor node location information data frame Frame of preserving of new node self more
mBe Frame '
m, and with Frame '
mBe broadcast to whole wireless sensor network, otherwise abandon this anchor node location information data frame Frame
m
When no longer including information interaction in the wireless sensor network, the anchor node location information data frame Frame of all nodes records
mIn d, H, D be self the shortest jumping distance, d to the respective anchors node
Min, beeline jumping figure H
CouWith the D of beeline local density
Sum
Step 2: set up the neural net that is used to calculate actual range between non-conterminous node;
Foundation comprises the three-layer neural network of input layer I, output layer O and a hidden layer H, and as shown in Figure 3, wherein, input layer I comprises three neurons, is respectively neuron I
1, neuron I
2With neuron I
3, neuron I
1Be input as the shortest internodal jumping distance, d
Min, neuron I
2Be input as beeline jumping figure H
Cou, neuron I
3Be input as the D of beeline local density
SumOutput layer O comprises a neuron O
1, be output as internodal actual range d
rHidden layer H comprises G
HIndividual neuron is respectively H
1, H
2...,
G
HNumber for hidden layer neuron; Described hidden layer neuron number G
HUsually the 2-10 that gets the input layer number doubly.
If neural net input layer I
pWith hidden layer neuron H
qThe connection weights be w
Pq, hidden layer neuron H
qWith output layer neuron O
1The connection weights be v
Q1, hidden layer neuron H
qBe output as h
q, hidden layer neuron H
qThreshold value be b
q, output layer neuron O
1Threshold value be b
o, the neuronic transfer function of hidden layer and output layer is respectively f
HAnd f
oWherein: p, q are natural number, p=1, and 2,3, q=1,2 ..., G
HThe pass of whole neural net input and output is:
Step 3: according to the shortest jumping distance, d of the anchor node in the wireless sensor network to other anchor node
Min, beeline jumping figure H
Cou, the D of beeline local density
SumAnd anchor node three-dimensional coordinate separately, the training sample of extraction neural net;
If 1. anchor node N
aAnd N
bCan carry out communication by the mode of direct or multi-hop, then according to N
aThree-dimensional coordinate (x
a, y
a, z
a) and N
bThree-dimensional coordinate (x
b, y
b, z
b) obtain N
aTo N
bActual range:
Wherein: a, b are natural number, and a ≠ b;
According to N
aTo N
bThe shortest jumping distance, d
Min(a, b), beeline jumping figure H
Cou(a, b), the D of beeline local density
Sum(a, b) and the actual range d between two anchor nodes
r(a b), obtains the training sample of neural net:
Sample
j={d
min(a,b),H
cou(a,b),D
sum(a,b),d
r(a,b)} (5)
Wherein,, Sample
jBe j training sample, j belongs to natural number.
If 2. anchor node N
aAnd N
bCan not carry out communication by the mode of direct or multi-hop, then can't be according to N
aAnd N
bPosition relation extract the training sample of neural net, this moment is according to other can extract the training sample of neural net by the position relation that the mode of direct or multi-hop is carried out the anchor node of communication in the wireless sensor network;
If the number of anchor node is G in the wireless sensor network
A, the unknown node number is G
U, can extract G at most according to the position relation of an anchor node and other anchor node
A-1 training sample, the maximum extractible training sample total quantity G of whole wireless sensor network
SFor: G
S=G
A(G
A-1)/2;
Step 4: utilize the training sample that obtains in the step 3 that neural net is trained;
Set the training precision ε of neural net
NMaximum iteration time N with neural metwork training
Max, the G that utilizes step 3 Chinese style (5) to obtain
SIndividual training sample is trained the neural net formula of setting up in the step 2 (3), makes neural net reach the training precision ε of predefined neural net
NOr reach the maximum iteration time N of neural metwork training
Max
Concrete steps are:
Corresponding to each sample Sample
j, j=1,2 ..., G
s, the neural net desired output is d
j, the actual d that is output as
j' (k
s), the mean square deviation function
For:
Neural net through type (7) to the connection weights of each interlayer and threshold value along the mean square deviation function
The negative gradient direction adjust, final satisfy
Required precision or iterations k
sReach maximum iteration time N
Max
Wherein,
Be the connection weight value matrix of current neural net,
Be the connection weight value matrix that next step iteration obtains,
Be the threshold matrix of current neural net,
Be the threshold matrix that next step iteration obtains,
It is current learning rate;
By training, neural net has write down the principal character of the geometry of wireless sensor network.
Step 5: according to the neural net that trains, each unknown node obtains self actual range to non-conterminous anchor node;
Interneuronal connection weights of each layer of neural net that trains in the step 4 and the neuronic threshold value of each layer are broadcast to whole wireless sensor network, set beeline jumping figure H
CouMaximum permissible value H
Max, unknown node N
xObtain self to anchor node N
iActual range d
r(x, i), x, i are natural number;
Be divided into following three kinds of situations:
(1) works as N
xTo N
iBeeline jumping figure H
Cou(x, i)=1, i.e. N
iBe N
xNeighbor node the time, N
xTo N
iThe shortest jumping distance, d
Min(x i) is N
xTo N
iActual range;
(2) as 1<H
Cou(x, i)≤H
MaxThe time, with N
xTo N
iThe shortest jumping distance, d
Min(x, i), beeline jumping figure H
Cou(x is i) with the D of beeline local density
Sum(x, i) substitution formula (3) obtains N
xTo N
iActual range;
(3) work as H
Cou(x, i)>H
MaxThe time, do not calculate N
xTo N
iActual range.
Step 6: unknown node obtains self three-dimensional coordinate according to the three-dimensional coordinate of anchor node and self to the actual range of anchor node;
When unknown node obtains self to arrive the actual range of four or more anchor node, set up the Euclidean distance equation group, obtain self three-dimensional coordinate;
Concrete steps are:
As unknown node N
xObtain self to four or more anchor node N
iActual range d
r(x, in the time of i), i=1,2 ..., n, n are natural number, n 〉=4, anchor node N
iThree-dimensional coordinate be (x
i, y
i, z
i), set up following Euclidean distance equation group:
Use obtains unknown node N based on the least square iterative method of Taylor expansion
xThree-dimensional coordinate, flow process realizes by following steps as shown in Figure 4:
1. get n anchor node N
iBarycenter or the maximum likelihood of formula (8) separate as unknown node N
xInitial three-dimensional coordinate (x
0, y
0, z
0), and establish iterations k
E=0;
2. with formula (8) at (x
0, y
0, z
0) locate to carry out Taylor series expansion, ignore the above component of second order, formula (8) is converted into system of linear equations:
Aδ=B (9)
Wherein:
r
iBe (x
0, y
0, z
0) to anchor node N
iDistance,
(Δ x, Δ y, Δ z) is respectively (x
0, y
0, z
0) increment;
3. use least square method, obtain the least square solution of formula (9): δ=(A
TA)
-1A
TB;
4. judge iteration stopping condition ‖ δ ‖
2≤ ε
EWhether set up ε
EUsually get greater than 0 and less than a decimal of 1;
If condition is set up, stop to calculate (x
0, y
0, z
0) be unknown node N
xThree-dimensional coordinate; Otherwise, make k
E=k
E+ 1;
5. judge k
E〉=K
MaxWhether set up K
MaxMaximum iteration time for the solving equation group set;
If set up, then stop to calculate (x
0, y
0, z
0) be node N
xThree-dimensional coordinate; Otherwise, get x '
0=x
0+ Δ x, y '
0=y
0+ Δ y, z '
0=z
0+ Δ z makes (x
0, y
0, z
0)=(x '
0, y '
0, z '
0), return step 2., up to obtaining N
xThree-dimensional coordinate.
When the 3-D wireless sensor network was the two dimensional wireless sensor network, the coordinate of node was a two-dimensional coordinate, with the three-dimensional coordinate of node in two-dimensional coordinate replacement the inventive method, in the step 6, as unknown node N
xObtain self to three or three above anchor node N
iActual range d
r(x, in the time of i), i=1,2 ..., n, n are natural number, the Euclidean distance equation group is set up in n 〉=3, replaces three-dimensional coordinate with two-dimensional coordinate, obtains unknown node N at last
xTwo-dimensional coordinate.
Embodiment:
As shown in Figure 5, in the three-dimensional spatial area of 200m * 200m * 200m, dispose 200 wireless sensor nodes at random by even distribution; Among the figure, anchor node is solid five-pointed star, and ratio is 20%, and ID is 1-40; Unknown node is a black circle, and ratio is 80%, and ID is 41-200.
Use node positioning method of the present invention and distance vector method to carry out node locating respectively, the position error of each node that obtains as shown in Figure 6, solid line is the position error of each node of using the inventive method and obtaining among the figure, average position error is 23.45%; The position error of each node that dotted line obtains for the service range vector method among the figure, average position error is 43.67%; With respect to the distance vector method, the inventive method can improve positioning accuracy about 20%; The service range vector method positions, and 51.25% node locating error surpasses 40%; And use the inventive method to position, have only about 10% node locating error above 40%.
Regulate the communication radius of node, at the network-in-dialing degree is 5~15 times, use node positioning method of the present invention and distance vector method to carry out node locating respectively, average position error as shown in Figure 7 under the heterogeneous networks degree of communication that obtains, solid line is the average position error of distance vector method, and dotted line is the average position error of the inventive method; Smaller or equal to 10 o'clock, the positioning accuracy of the inventive method was higher by about 20% than the positioning accuracy of distance vector method at the network-in-dialing degree; Greater than 10 o'clock, the positioning accuracy of the inventive method also positioning accuracy than distance vector method was high more than 10% at the network-in-dialing degree.This shows that node positioning method provided by the invention can effectively improve the positioning accuracy of 3-D wireless sensor network.
Claims (8)
1. 3-D wireless sensor network node method for self-locating based on neural net is characterized in that: realize by following steps:
Step 1: intiating radio sensor network;
Each node obtains node self local density by carrying out information interaction with self neighbor node, and measures self distance to neighbor node; Location information data frame by all anchor nodes is propagated in wireless sensor network, and all nodes obtain self the shortest jumping segment distance, beeline jumping figure and beeline local density to all anchor nodes; Described node self local density is meant the number of neighbor node in the node communication context; The shortest described jumping segment distance is meant the length of the shortest path that can communicate with one another between two nodes; Described beeline jumping figure is meant the jumping hop count order that the path at the shortest jumping segment distance place between two nodes comprises; Described beeline local density is meant self local density's sum of all nodes that the path at the shortest jumping segment distance place comprises;
Step 2: set up the neural net that is used to calculate actual range between non-conterminous node;
Foundation comprises the three-layer neural network of input layer I, output layer O and a hidden layer H; Wherein, input layer I comprises three neurons, is respectively neuron I
1, neuron I
2With neuron I
3, neuron I
1Be input as the shortest internodal jumping distance, d
Min, neuron I
2Be input as beeline jumping figure H
Cou, neuron I
3Be input as the D of beeline local density
SumOutput layer O comprises a neuron O
1, be output as internodal actual range d
rHidden layer H comprises G
HIndividual neuron is respectively H
1, H
2...,
, G
HNumber for hidden layer neuron;
If neural net input layer I
pWith hidden layer neuron H
qThe connection weights be w
Pq, hidden layer neuron H
qWith output layer neuron O
1The connection weights be v
Q1, hidden layer neuron H
qBe output as h
q, hidden layer neuron H
qThreshold value be b
q, output layer neuron O
1Threshold value be b
o, the neuronic transfer function of hidden layer and output layer is respectively f
HAnd f
oWherein: p, q are natural number, p=1, and 2,3, q=1,2 ..., G
HThe pass of whole neural net input and output is:
Step 3: according to the shortest jumping distance, d of the anchor node in the wireless sensor network to other anchor node
Min, beeline jumping figure H
Cou, the D of beeline local density
SumAnd anchor node three-dimensional coordinate separately, the training sample of extraction neural net;
If 1. anchor node N
aAnd N
bCan carry out communication by the mode of direct or multi-hop, then according to N
aThree-dimensional coordinate (x
a, y
a, z
a) and N
bThree-dimensional coordinate (x
b, y
b, z
b) obtain N
aTo N
bActual range:
Wherein: a, b are natural number, and a ≠ b;
According to N
aTo N
bThe shortest jumping distance, d
Min(a, b), beeline jumping figure H
Cou(a, b), the D of beeline local density
Sum(a, b) and the actual range d between two anchor nodes
r(a b), obtains the training sample of neural net:
Sample
j={d
min(a,b),H
cou(a,b),D
sum(a,b),d
r(a,b)}(3)
Wherein, Sample
jBe j training sample, j belongs to natural number;
If 2. anchor node N
aAnd N
bCan not carry out communication by the mode of direct or multi-hop, then can't be according to N
aAnd N
bPosition relation extract the training sample of neural net, this moment is according to other can extract the training sample of neural net by the position relation that the mode of direct or multi-hop is carried out the anchor node of communication in the wireless sensor network;
If the number of anchor node is G in the wireless sensor network
A, the unknown node number is G
U, can extract G at most according to the position relation of an anchor node and other anchor node
A-1 training sample, the maximum extractible training sample total quantity G of whole wireless sensor network
SFor: G
S=G
A(G
A-1)/2;
Step 4: utilize the training sample that obtains in the step 3 that neural net is trained;
Set the training precision ε of neural net
NMaximum iteration time N with neural metwork training
Max, the G that utilizes step 3 Chinese style (3) to obtain
SIndividual training sample is trained the neural net formula of setting up in the step 2 (1), makes neural net reach the training precision ε of predefined neural net
NOr reach the maximum iteration time N of neural metwork training
Max
Step 5: according to the neural net that trains, each unknown node obtains self actual range to non-conterminous anchor node;
Interneuronal connection weights of each layer of neural net that trains in the step 4 and the neuronic threshold value of each layer are broadcast to whole wireless sensor network, set beeline jumping figure H
CouMaximum permissible value H
Max, unknown node N
xObtain self to anchor node N
iActual range d
r(x, i), x, i are natural number;
Step 6: unknown node obtains self three-dimensional coordinate according to the three-dimensional coordinate of anchor node and self to the actual range of anchor node;
When unknown node obtains self to arrive the actual range of four or more anchor node, set up the Euclidean distance equation group, obtain self three-dimensional coordinate.
2. according to the described a kind of 3-D wireless sensor network node method for self-locating based on neural net of claim 1, it is characterized in that: the concrete steps of the intiating radio sensor network of described step 1 are:
(1) all node broadcasts comprise the request-reply Frame of self ID, transmission reply data frame was replied after neighbor node was received the request-reply Frame, the number of the reply data frame that each node statistics self is received obtains self local density, and measures self distance to neighbor node;
(2) all anchor nodes are broadcasted the location information data frame Frame that comprises self ID and self three-dimensional coordinate in wireless sensor network
m, m is a natural number; Frame
mForm as follows:
Frame
m={ID
m,x
m,y
m,z
m,H,d,D}(4)
Wherein, ID
mBe the ID of m anchor node, (x
m, y
m, z
m) be the three-dimensional coordinate of m anchor node; H is Frame
mThe jumping hop count order of process, d is Frame
mThe jumping segment distance sum of process, D is Frame
mLocal density's sum of all nodes of process; H, d are initialized as 0, and D is initialized as the local density of anchor node;
(3) node receives the anchor node location information data frame Frame that self neighbor node sends or transmits in wireless sensor network
mThe time, whether decision node self received the location information data frame Frame of this anchor node
m
1. work as node and do not receive this anchor node location information data frame Frame
mThe time, then need upgrade Frame
m, the anchor node location information data frame Frame after the renewal
m' be:
Frame
m′={ID
m,x
m,y
m,z
m,(H+1),(d+d′),(D+D′)}(5)
Wherein, d ' arrives to self sending or transmit Frame for node
mThe distance of neighbor node, D ' is the local density of node self, node is preserved the anchor node location information data frame Frame after upgrading
m', and with Frame
m' be broadcast to whole wireless sensor network;
2. work as node and received this anchor node location information data frame Frame
mThe time, judge Frame
mIn d add the jumping segment distance sum whether d ' has preserved less than node self to this anchor node;
If less than, then according to formula (5) this anchor node location information data frame Frame of preserving of new node self more
mBe Frame
m', and with Frame
m' be broadcast to whole wireless sensor network; Otherwise abandon this anchor node location information data frame Frame
m
When no longer including information interaction in the wireless sensor network, the anchor node location information data frame Frame of all nodes records
mIn d, H, D be self the shortest jumping distance, d to the respective anchors node
Min, beeline jumping figure H
CouWith the D of beeline local density
Sum
3. according to the described a kind of 3-D wireless sensor network node method for self-locating of claim 1, it is characterized in that based on neural net: in the step 2, hidden layer neuron number G
HThe 2-10 that gets the input layer number doubly.
4. according to the described a kind of 3-D wireless sensor network node method for self-locating based on neural net of claim 1, it is characterized in that: the concrete steps that the training sample that the utilizing of described step 4 obtains in the step 3 is trained neural net are:
Corresponding to each sample Sample
j, j=1,2 ..., G
s, the neural net desired output is d
j, the actual d ' that is output as
j(k
s), the mean square deviation function
For:
Neural net through type (7) to the connection weights of each interlayer and threshold value along the mean square deviation function
The negative gradient direction adjust, final satisfy
Required precision or iterations k
sReach maximum iteration time N
Max
Wherein,
Be the connection weight value matrix of current neural net,
Be the connection weight value matrix that next step iteration obtains,
Be the threshold matrix of current neural net,
Be the threshold matrix that next step iteration obtains,
It is current learning rate; By training, neural net has write down the principal character of the geometry of wireless sensor network.
5. according to the described a kind of 3-D wireless sensor network node method for self-locating of claim 1, it is characterized in that: unknown node N in the described step 5 based on neural net
xObtain self to anchor node N
iActual range d
r(x i) is divided into following three kinds of situations:
(1) works as N
xTo N
iBeeline jumping figure H
Cou(x, i)=1, i.e. N
iBe N
xNeighbor node the time, N
xTo N
iThe shortest jumping distance, d
Min(x i) is N
xTo N
iActual range;
(2) as 1<H
Cou(x, i)≤H
MaxThe time, with N
xTo N
iThe shortest jumping distance, d
Min(x, i), beeline jumping figure H
Cou(x is i) with the D of beeline local density
Sum(x, i) substitution formula (1) obtains N
xTo N
iActual range;
(3) work as H
Cou(x, i)>H
MaxThe time, do not calculate N
xTo N
iActual range.
6. according to claim 1,2,3,4 or 5 described a kind of 3-D wireless sensor network node method for self-locating based on neural net, it is characterized in that, when described 3-D wireless sensor network is the two dimensional wireless sensor network, the coordinate of node is a two-dimensional coordinate, replaces three-dimensional coordinate with two-dimensional coordinate.
7. according to the described a kind of 3-D wireless sensor network node method for self-locating of claim 1 based on neural net, it is characterized in that unknown node according to the three-dimensional coordinate of anchor node and self to the concrete steps that the distance of anchor node obtains self three-dimensional coordinate is in the described step 6:
As unknown node N
xObtain self to four or more anchor node N
iActual range d
r(x, in the time of i), i=1,2 ..., n, n are natural number, n 〉=4, anchor node N
iThree-dimensional coordinate be (x
i, y
i, z
i), set up following Euclidean distance equation group:
Use obtains unknown node N based on the least square iterative method of Taylor expansion
xThree-dimensional coordinate, realize by following steps:
1. get n anchor node N
iBarycenter or the maximum likelihood of formula (8) separate as unknown node N
xInitial three-dimensional coordinate (x
0, y
0, z
0), and establish iterations k
E=0;
2. with formula (8) at (x
0, y
0, z
0) locate to carry out Taylor series expansion, ignore the above component of second order, formula (8) is converted into system of linear equations:
Aδ=B (9)
r
iBe (x
0, y
0, z
0) to anchor node N
iDistance,
(Δ x, Δ y, Δ z) is respectively (x
0, y
0, z
0) increment;
3. use least square method, obtain the least square solution of formula (9): δ=(A
TA)
-1A
TB;
4. judge the iteration stopping condition || δ ||
2≤ ε
EWhether set up ε
EGet greater than 0 and less than a decimal of 1;
If condition is set up, stop to calculate (x
0, y
0, z
0) be unknown node N
xThree-dimensional coordinate; Otherwise, make k
E=k
E+ 1;
5. judge k
E〉=K
MaxWhether set up K
MaxMaximum iteration time for the solving equation group set;
If set up, then stop to calculate (x
0, y
0, z
0) be node N
xThree-dimensional coordinate; Otherwise, get x '
0=x
0+ Δ x, y '
0=y
0+ Δ y, z '
0=z
0+ Δ z makes (x
0, y
0, z
0)=(x '
0, y '
0, z '
0), return step 2., up to obtaining N
xThree-dimensional coordinate.
8. according to the described a kind of 3-D wireless sensor network node method for self-locating of claim 7, it is characterized in that when described 3-D wireless sensor network was the two dimensional wireless sensor network, the coordinate of node was a two-dimensional coordinate based on neural net; As unknown node N
xObtain self to three or three above anchor node N
iActual range d
r(x, in the time of i), i=1,2 ..., n, n are natural number, the Euclidean distance equation group is set up in n 〉=3, replaces three-dimensional coordinate with two-dimensional coordinate, obtains unknown node N at last
xTwo-dimensional coordinate.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN2009102363723A CN101695190B (en) | 2009-10-20 | 2009-10-20 | Three-dimensional wireless sensor network node self-locating method based on neural network |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN2009102363723A CN101695190B (en) | 2009-10-20 | 2009-10-20 | Three-dimensional wireless sensor network node self-locating method based on neural network |
Publications (2)
Publication Number | Publication Date |
---|---|
CN101695190A CN101695190A (en) | 2010-04-14 |
CN101695190B true CN101695190B (en) | 2011-07-27 |
Family
ID=42094113
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN2009102363723A Expired - Fee Related CN101695190B (en) | 2009-10-20 | 2009-10-20 | Three-dimensional wireless sensor network node self-locating method based on neural network |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN101695190B (en) |
Families Citing this family (18)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101853531B (en) * | 2010-05-25 | 2012-09-05 | 北京航空航天大学 | Helicopter flight state identification method based on presort technology and RBF (Radial Basis Function) neural network |
CN102158957A (en) * | 2011-03-11 | 2011-08-17 | 山东大学 | Positioning method for wireless sensor network in three-dimensional space through twice region segmentation and iteration |
CN102209383B (en) * | 2011-05-18 | 2014-02-26 | 华北电力大学 | Wireless sensor three-dimensional positioning method based on Euclidean algorithm |
CN103597413B (en) * | 2011-06-03 | 2017-01-18 | 西门子公司 | Method for the computer-supported generation of a data-driven model of a technical system, in particular of a gas turbine or wind turbine |
CN102288938B (en) * | 2011-06-28 | 2013-01-23 | 山东大学威海分校 | Effective three-dimensional positioning method for wireless sensor network node |
CN102238705B (en) * | 2011-07-04 | 2013-11-27 | 南京邮电大学 | Wireless sensor network topology control method based on artificial neural network (ANN) |
CN102300209B (en) * | 2011-08-24 | 2012-11-07 | 北京航空航天大学 | Wireless sensor network safety multi-hop positioning method |
CN102271377B (en) * | 2011-09-13 | 2013-11-20 | 河海大学常州校区 | Double-bounce multi-path route method of wireless sensor network based on energy collection technology |
CN102625446B (en) * | 2012-03-02 | 2014-09-03 | 黄东 | Method for positioning object tags in Internet of Things |
CN105811993B (en) * | 2016-03-11 | 2017-02-15 | 北京航空航天大学 | Data collection method based on compression dictionary learning in wireless sensor network |
CN106093849B (en) * | 2016-06-03 | 2018-05-25 | 清华大学深圳研究生院 | A kind of Underwater Navigation method based on ranging and neural network algorithm |
CN106912105B (en) * | 2017-03-08 | 2020-06-09 | 哈尔滨理工大学 | Three-dimensional positioning method based on PSO _ BP neural network |
CN108187206A (en) * | 2018-01-02 | 2018-06-22 | 佳木斯大学 | A kind of dual cavity bronchus vessel and control method |
CN108134980A (en) * | 2018-01-11 | 2018-06-08 | 南京华讯方舟通信设备有限公司 | A kind of fast distributed relative positioning method suitable for unmanned plane bee colony self-organizing network |
CN109709497A (en) * | 2019-02-16 | 2019-05-03 | 冀中能源峰峰集团有限公司 | A kind of high-accuracy position system and method |
CN110082717A (en) * | 2019-04-30 | 2019-08-02 | 上海海事大学 | A kind of underwater wireless sensor node positioning method |
CN112333634A (en) * | 2020-11-19 | 2021-02-05 | 江南大学 | Hybrid node positioning method based on UAV |
CN113111757B (en) * | 2021-04-01 | 2022-04-05 | 国能智慧科技发展(江苏)有限公司 | Accurate personnel positioning system based on image processing and intelligent Internet of things |
-
2009
- 2009-10-20 CN CN2009102363723A patent/CN101695190B/en not_active Expired - Fee Related
Also Published As
Publication number | Publication date |
---|---|
CN101695190A (en) | 2010-04-14 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN101695190B (en) | Three-dimensional wireless sensor network node self-locating method based on neural network | |
CN108696932B (en) | Outdoor fingerprint positioning method using CSI multipath and machine learning | |
CN101483818B (en) | Tri-dimensional positioning method for underwater wireless sensor network node | |
CN104363616B (en) | A kind of method of the indoor three dimensions field strength of outdoor to indoor propagation model prediction | |
CN102186242B (en) | Method for positioning mobile node of wireless sensor network in fixed area | |
Zhang et al. | A comprehensive study of bluetooth fingerprinting-based algorithms for localization | |
CN103096462B (en) | Non-ranging node locating method of wireless sensor network | |
CN109151750B (en) | LTE indoor positioning floor distinguishing method based on recurrent neural network model | |
CN101938832A (en) | Division and refinement-based node self-positioning method for wireless sensor network | |
CN106131797A (en) | A kind of water-saving irrigation monitoring network locating method based on RSSI range finding | |
CN104135766B (en) | A kind of Cooperative Localization Method between wireless network based on mesh generation | |
Lee et al. | Path loss exponent prediction for outdoor millimeter wave channels through deep learning | |
CN107727095A (en) | 3D indoor orientation methods based on spectral clustering and weighting reverse transmittance nerve network | |
CN102164408A (en) | Wireless sensor network annular positioning method based on transmission delay and improvement method thereof | |
CN103648164A (en) | Time difference of arrival and Gossip algorithm based wireless sensor network distributed positioning method | |
CN105676178A (en) | Wireless sensor network positioning method based on compressed sensing and BP neural networks | |
Chagas et al. | Genetic algorithms and simulated annealing optimization methods in wireless sensor networks localization using artificial neural networks | |
CN102497669A (en) | Wireless sensor network (WSN) node positioning method | |
CN110049441B (en) | WiFi indoor positioning method based on deep ensemble learning | |
CN104683949B (en) | It is a kind of to be applied to the mixing method for self-locating based on aerial array in Wireless Mesh network | |
CN107148079A (en) | Three-dimensional alignment by union and method for tracing in movable self-organization sensor network | |
Shit et al. | AI-enabled fingerprinting and crowdsource-based vehicle localization for resilient and safe transportation systems | |
CN106131951A (en) | RSSI based on equilateral triangle model weights distance-finding method | |
Monfared et al. | Range free localization of wireless sensor networks based on sugeno fuzzy inference | |
CN103037503B (en) | Wireless sensor network positioning method and wireless sensor network positioning system |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
C14 | Grant of patent or utility model | ||
GR01 | Patent grant | ||
C17 | Cessation of patent right | ||
CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20110727 Termination date: 20121020 |