CN106093849B - A kind of Underwater Navigation method based on ranging and neural network algorithm - Google Patents
A kind of Underwater Navigation method based on ranging and neural network algorithm Download PDFInfo
- Publication number
- CN106093849B CN106093849B CN201610389518.8A CN201610389518A CN106093849B CN 106093849 B CN106093849 B CN 106093849B CN 201610389518 A CN201610389518 A CN 201610389518A CN 106093849 B CN106093849 B CN 106093849B
- Authority
- CN
- China
- Prior art keywords
- node
- neural network
- output
- nodes
- distance
- 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.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 64
- 238000013528 artificial neural network Methods 0.000 title claims abstract description 61
- 238000004422 calculation algorithm Methods 0.000 title claims abstract description 51
- 239000013598 vector Substances 0.000 claims abstract description 43
- 238000012549 training Methods 0.000 claims abstract description 25
- 238000003062 neural network model Methods 0.000 claims abstract description 9
- 238000012546 transfer Methods 0.000 claims description 8
- 230000004913 activation Effects 0.000 claims description 4
- 238000011478 gradient descent method Methods 0.000 claims description 4
- 238000005259 measurement Methods 0.000 claims description 4
- 230000017105 transposition Effects 0.000 claims description 4
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 claims description 4
- 238000000926 separation method Methods 0.000 claims description 2
- 230000007935 neutral effect Effects 0.000 abstract 2
- 210000005036 nerve Anatomy 0.000 abstract 1
- 239000003643 water by type Substances 0.000 abstract 1
- 238000012706 support-vector machine Methods 0.000 description 13
- 230000006870 function Effects 0.000 description 7
- 238000004364 calculation method Methods 0.000 description 6
- 230000033001 locomotion Effects 0.000 description 6
- 230000000694 effects Effects 0.000 description 5
- 230000005540 biological transmission Effects 0.000 description 4
- 238000011161 development Methods 0.000 description 3
- 238000011160 research Methods 0.000 description 3
- 238000012360 testing method Methods 0.000 description 3
- 238000010586 diagram Methods 0.000 description 2
- 238000010801 machine learning Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 230000008569 process Effects 0.000 description 2
- 239000000523 sample Substances 0.000 description 2
- 238000006467 substitution reaction Methods 0.000 description 2
- 230000007547 defect Effects 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 230000001976 improved effect Effects 0.000 description 1
- 238000000691 measurement method Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
- G01S5/00—Position-fixing by co-ordinating two or more direction or position line determinations; Position-fixing by co-ordinating two or more distance determinations
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04W—WIRELESS COMMUNICATION NETWORKS
- H04W64/00—Locating users or terminals or network equipment for network management purposes, e.g. mobility management
- H04W64/006—Locating users or terminals or network equipment for network management purposes, e.g. mobility management with additional information processing, e.g. for direction or speed determination
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Artificial Intelligence (AREA)
- Biomedical Technology (AREA)
- Life Sciences & Earth Sciences (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Computer Networks & Wireless Communication (AREA)
- Signal Processing (AREA)
- Radar, Positioning & Navigation (AREA)
- Remote Sensing (AREA)
- Feedback Control In General (AREA)
- Measurement Of Velocity Or Position Using Acoustic Or Ultrasonic Waves (AREA)
Abstract
The invention discloses a kind of Underwater Navigation methods based on ranging and neural network algorithm, include the following steps:K anchor node and N number of node to be positioned are arranged in positioning waters, measure the distance between each node to be positioned and K anchor node;Defiber is trained and two stages of tuning on-line are positioned;The off-line training step comprises the following steps:N-th of node to be positioned and all anchor node distances are measured, distance vector is formed, distance vector is inputted into three-layer neural network:Neural network input layer, neutral net hidden layer and neutral net output layer;Prediction model parameters, output nerve network model parameter are updated by algorithm iteration;The tuning on-line stage comprises the following steps:N-th of node to be positioned and all anchor node current distances are measured, form distance vector;The neural network model parameter that distance vector input off-line training step is obtained;The position coordinates of node to be positioned is exported, reduces range error.
Description
Technical Field
The invention relates to the field of underwater wireless sensor networks, in particular to an underwater positioning method based on distance measurement and a neural network algorithm.
Background
In recent years, the world has increasingly emphasized the marine problem. With the development of marine environmental protection, resource development, marine engineering and other projects and the rapid development of land Wireless Sensor Network research, an Underwater Wireless Sensor Network (UWSN) has become a new research hotspot. The underwater wireless sensor network positioning problem is a research focus.
The existing UWSN positioning algorithm can be generally divided into a ranging-based algorithm and a ranging-free algorithm according to whether the distance of a node needs to be measured in the positioning process. In the ranging-based algorithm, the distance between an unknown node and an adjacent known node needs to be acquired before the node is positioned, or the azimuth angle of the unknown node relative to the known node is measured in a known coordinate system. The typical distance measurement methods mainly include the following four methods: (1) time of Arrival (TOA) ranging, requiring accurate Time synchronization between nodes; (2) time Difference of arrival (TDOA) based ranging, generally limited by the distance traveled by sound waves; (3) ranging based on Angle of arrival (AOA), requiring nodes to be equipped with special hardware and being easily interfered by external factors; (4) the Signal Received Strength (RSSI) ranging method has low power consumption and low cost, but has large ranging error. After the distance between the unknown node and the anchor node is obtained, common methods for calculating the position coordinates of the unknown node include Trilateration (Trilateration) and Least Squares (GLS). Trilateration generally requires only three anchor nodes, and the location coordinates of unknown nodes are calculated by intersecting three circles at a point. However, in practical situations, due to the existence of the ranging error, three circles usually cannot intersect at one point, so that the trilateration method is poor in positioning stability and positioning accuracy, and in such a situation, the least square method is required to be used for positioning. The basic principle of the least squares method is to find a point that minimizes the difference between the measured distance and the theoretically estimated distance as the coordinates of the unknown node. However, when the least square method is used for positioning, the number of anchor nodes is often more than three.
Although the least square method has a much improved effect in positioning accuracy compared with the trilateral positioning method, the underwater channel contains very serious interference problems of multipath effect, attenuation, noise and the like, the interference often causes a certain error between the measured distance and the real distance, and the least square method has a certain limit on the fault tolerance of the distance error, thereby causing the least square method to have poor performance in positioning accuracy and mean square error.
Disclosure of Invention
The invention aims to solve the problems of the inclusion of the interference, particularly the noise, and the position calculation of the positioning algorithm.
The technical problem of the invention is solved by the following technical scheme:
an underwater positioning method based on distance measurement and neural network algorithm is characterized by comprising the following steps:
a1, arranging K anchor nodes and N nodes to be positioned in a positioning water area, and measuring the distance between each node to be positioned and the K anchor nodes;
positioning in two stages of A2, separation line training and online positioning;
a3, the off-line training phase comprises the following steps:
a31, measuring the distance between the nth node to be positioned and all anchor nodes to form a distance vector dn;
A32, calculating the distance vector dnInputting three layers of neural networks: the device comprises a neural network input layer, a neural network hiding layer and a neural network output layer;
and A33, iteratively updating the prediction model parameters through an algorithm, and outputting the neural network model parameters.
A4, the online positioning stage comprises the following steps:
a41, measuring the current distance between the nth node to be positioned and all anchor nodes to form a distance vector sn;
A42, calculating the distance vector snInputting neural network model parameters obtained in an off-line training stage;
and A43, outputting the position coordinates of the node to be positioned.
According to another specific aspect of the present invention, the distance vector in step a31 is represented by the following formula:
dn=[dn,1,…,dn,k,…,dn,K]T
wherein: dn,kRepresents the distance between the nth node to be positioned and the kth anchor node,
n ∈ [1, N ], K ∈ [1, K ], T denotes the transposition operation.
In accordance with another particular aspect of the present invention, in step a32,
the neural network input layer: the total number of the nodes is K +1, the first K nodes are active nodes, the first K nodes are correspondingly input nodes, and the input and the output of the kth node are dn,k,k∈[1,K]The K +1 th node of the input layer is a bias node, and the output is
The neural network hidden layer: the total number of the nodes is U +1, the first U are active nodes, wherein the input of the U-th active node isOutput is asu∈[1,U],For the transfer parameter vector from the input layer to the U-th active node of the hidden layer, f (x) represents the active function, and the output vectors of all the U active nodes areThe U +1 th node of the hidden layer is a bias node, and the output is
The neural network output layer: only 1 node, with inputs ofOutput is asWherein,for the hidden layer to output node transfer parameter vector, (x)n,yn) Is the position coordinate of the nth node to be positioned.
According to another particular aspect of the invention, said activation function f (x) is a sigmoid function, i.e. f (x) 1/(1+ exp (-x)).
According to another specific aspect of the present invention, in step a33, the batch gradient descent method is used, and the prediction model parameters converging to the optimal values are obtained through multiple iterative updates:
according to another specific aspect of the present invention, the partial derivatives are calculated using a back propagation algorithm in the plurality of iterative updates.
According to another specific aspect of the present invention, the distance vector in step a41 is represented by the following formula:
sn=[sn,1,…,sn,k,…,sn,K]T
wherein: sn,kRepresents the current distance between the nth node to be positioned and the kth anchor node,
n∈[1,N],k∈[1,K]。
according to another specific aspect of the present invention, the distance vector s is processed in step A42nInputting the spirit obtained in the off-line training stageParameters of network modelThe method comprises the following steps:
a421, a neural network input layer: the input and output of the kth node are sn,k,k∈[1,K]The bias node output is
A422, a neural network hidden layer: the u-th active node is input asOutput is asu∈[1,U]All U active nodes output vectors ofBiased node output is
A423, a neural network output layer: input is asOutput is as(xn,yn) Is the position coordinate of the nth node to be positioned.
According to another specific aspect of the invention, each of said nodes to be located is measured for distance from all anchor nodes using a TDOA method.
Compared with the prior art, the invention has the advantages that:
the positioning method provided by the invention organically combines a neural network algorithm with a positioning algorithm based on ranging, overcomes the defects of certain limitation, poor positioning precision and poor stability of the traditional position calculation method such as a least square method on the fault tolerance of the distance error by adopting a TDOA ranging method, utilizes the characteristic that the neural network algorithm usually has excellent error fault tolerance, and performs positioning based on ranging and by using the neural network algorithm to replace the traditional position calculation algorithm, thereby reducing the ranging error.
Drawings
FIG. 1 is a flow chart of the off-line training phase of the present invention;
FIG. 2 is a diagram of a neural network architecture according to the present invention;
FIG. 3 is a flow chart of the online location phase of the present invention;
FIG. 4 is a diagram of a network model architecture of the present invention;
FIG. 5 is a graph comparing the performance of three algorithms when the number of anchor nodes changes in accordance with the present invention;
FIG. 6 is a graph comparing the performance of three algorithms when the training set size is changed according to the present invention;
FIG. 7 is a graph comparing the performance of three algorithms when the range error of the present invention is changed.
Detailed Description
The positioning method provided by the invention organically combines a neural network algorithm and a positioning algorithm based on distance measurement. The underwater channel usually has a ranging error caused by serious multipath effect, interference, noise and other factors, and the traditional position calculation method such as the least square method has certain limitation on the fault tolerance of the ranging error, so that the positioning accuracy and the stability (mean square error) of the underwater channel are poor. Neural network algorithms typically have excellent error tolerance. Therefore, the core of the present invention is based on ranging and uses neural network algorithms instead of traditional position calculation algorithms for positioning.
Consider that a certain underwater wireless sensor network has N nodes to be positioned and K anchor nodes. The algorithm for positioning by using the neural network algorithm is divided into two stages: an off-line training phase and an on-line positioning phase. The invention provides detailed positioning algorithm steps:
1. an off-line training stage: sequentially constructing a position estimation prediction model for each node to be positioned according to the following steps,
as shown in fig. 1:
a) measuring the distance between the nth node to be positioned and all anchor nodes by using a TDOA method to form a distance vector dn=[dn,1,…,dn,k,…,dn,K]TWherein d isn,kRepresents the distance between the nth node to be positioned and the kth anchor node, and N belongs to [1, N ∈],k∈[1,K]T denotes a transposition operation;
b) will distance vector dnInputting a three-layer neural network;
1) the input layer of the neural network (the first column counted on the left side of the figure 2) has K +1 nodes in total, the first K nodes are activation nodes, the first K nodes are corresponding to input nodes, and the input and the output of the kth node are dn,k,k∈[1,K]The K +1 th node of the input layer is a bias node, and the output is
2) The hidden layer of the neural network (the second column from the left in FIG. 2) has U +1 nodes, the first U nodes are active nodes, and the input of the U-th active node isOutput is asu∈[1,U],For the transfer parameter vector from the input layer to the U-th active node of the hidden layer, f (x) represents the active function, where a sigmoid function is used, i.e., f (x) ═ 1/(1+ exp (-x)), and the output vectors of all U active nodes areThe U +1 th node of the hidden layer is a bias node, and the output is
3) The output layer of the neural network (third column from left in FIG. 2) has only 1 node, and its input isOutput is asWherein,for the hidden layer to output node transfer parameter vector, (x)n,yn) Is the position coordinate of the nth node to be positioned.
c) The invention uses a batch gradient descent method to update the parameters of the prediction model through multiple iterationsUntil it converges to an optimal value, wherein the key step is to calculate the partial derivatives, the present invention uses a Back Propagation (BP) algorithm to perform fast and accurate calculation of the partial derivatives.
2. And (3) in an online positioning stage: neural network model parameters obtained from an offline training phaseSequentially calculating the coordinates of each node to be positioned, as shown in fig. 3:
a) measuring using TDOA methodThe nth node to be positioned and the current distances of all anchor nodes form a distance vector sn=[sn,1,…,sn,k,…,sn,K]TWherein s isn,kRepresents the current distance between the nth node to be positioned and the kth anchor node, and belongs to [1, N ∈],k∈[1,K];
b) Will distance vector snInputting a three-layer neural network; the network structure is shown in FIG. 2, and the network parameters areThe method comprises the following specific steps:
1) neural network input layer: the input and output of the kth node are sn,k,k∈[1,K]The bias node output is
2) Neural network hidden layer: the u-th active node is input asOutput is asu∈[1,U]All U active nodes output vectors ofBiased node output is
3) The neural network output layer: input is asOutput is as(xn,yn) Is the position coordinate of the nth node to be positioned.
Detailed Description
K anchor nodes and N nodes to be positioned are deployed in a positioning area. In order to acquire a certain amount of data (more than or equal to 100) for training a neural network model, the invention considers that a detector moves along a fixed track underwater with a constant motion speed of v m/s, and an acoustic signal is used for information transmission underwater, and the motion speed v of the detector is far less than the transmission speed of the acoustic signal underwater. And measuring and recording the distances between the current point and the K anchor nodes and the coordinates of the current point once at fixed distance intervals by using a certain distance measuring mode, such as TOA, TDOA, AOA and the like, in the moving process of the detector until the whole track is completely followed.
The invention selects the positioning area in a 200m by 200m water area, and sets 12 anchor nodes in the positioning area. The invention considers that a detector moves along a fixed track underwater with a movement speed v equal to 1m/s and uses an acoustic wave signal for information transmission underwater, the detector measures distance by using a TDOA method (ranging based on signal arrival time difference) during movement, data acquisition is carried out every 1m, 729 data are acquired in total, and a specific network model is shown in FIG. 4.
1. Off-line training phase
Sequentially constructing a position estimation prediction model for each node to be positioned according to the following steps, as shown in FIG. 1:
a) measuring the distance between the nth node to be positioned and all anchor nodes by using a TDOA method to form a distance vector dn=[dn,1,…,dn,k,…,dn,K]TWherein d isn,kRepresents the distance between the nth node to be positioned and the kth anchor node, and N belongs to [1, N ∈],k∈[1,K]T denotes a transposition operation;
b) will distance vector dnInput three-layer neural network;
1) The input layer of the neural network (the first column counted on the left side of the figure 2) has K +1 nodes in total, the first K nodes are activation nodes, the first K nodes are corresponding to input nodes, and the input and the output of the kth node are dn,k,k∈[1,K]The K +1 th node of the input layer is a bias node, and the output is
2) The hidden layer of the neural network (the second column from the left in FIG. 2) has U +1 nodes, the first U nodes are active nodes, and the input of the U-th active node isOutput is asu∈[1,U],For the transfer parameter vector from the input layer to the U-th active node of the hidden layer, f (x) represents the active function, where a sigmoid function is used, i.e., f (x) ═ 1/(1+ exp (-x)), and the output vectors of all U active nodes areThe U +1 th node of the hidden layer is a bias node, and the output is
3) The output layer of the neural network (third column from left in FIG. 2) has only 1 node, and its input isOutput is asWherein,for the hidden layer to output node transfer parameter vector, (x)n,yn) Is the position coordinate of the nth node to be positioned.
c) The invention updates the prediction model parameters through multiple iterations by using a batch gradient descent methodUntil it converges to an optimal value, wherein the key step is to calculate the partial derivatives, the present invention uses a Back Propagation (BP) algorithm to perform fast and accurate calculation of the partial derivatives.
2. On-line positioning stage
And (3) in an online positioning stage: neural network model parameters obtained from an offline training phaseSequentially calculating the coordinates of each node to be positioned, as shown in fig. 3:
a) measuring the current distances between the nth node to be positioned and all anchor nodes by using a TDOA (time difference of arrival) method to form a distance vector sn=[sn,1,…,sn,k,…,sn,K]TWherein s isn,kRepresents the current distance between the nth node to be positioned and the kth anchor node, and belongs to [1, N ∈],k∈[1,K];
b) Will distance vector snInputting a three-layer neural network; the network structure is shown in FIG. 2, and the network parameters areThe method comprises the following specific steps:
1) neural network input layer: the input and output of the kth node are sn,k,k∈[1,K]The bias node output is
2) Neural network hidden layer: the u-th active node is input asOutput is asu∈[1,U]All U active nodes output vectors ofBiased node output is
3) The neural network output layer: input is asOutput is as(xn,yn) Is the position coordinate of the nth node to be positioned.
The performance of the invention was analyzed as follows:
1.1 Environment and parameters
The invention selects the positioning area in a 200m by 200m water area, and sets 12 anchor nodes in the positioning area. A probe is used to move along a fixed track underwater with a movement velocity v equal to 1m/s, and an acoustic wave signal is used for information transmission underwater, the probe uses a TDOA method (ranging based on signal arrival time difference) for ranging during movement, and data acquisition is performed every 1m, 729 pieces of data are acquired in total, and a specific network model is shown in FIG. 4.
1.2 Performance index
In the invention, in order to compare the positioning performance of the three methods, the concept of Mean Squared Error (MSE) is introduced, and the MSE in positioning not only represents the advantages and disadvantages of different positioning methods in positioning precision, but also represents the difference of different positioning methods in positioning stability. It is defined as follows:
wherein,for the estimated coordinates of the node to be located, (x)n,yn) Is the real coordinate of the node to be positioned.
1.3 Performance analysis
In the following analysis, the number of training sets is 500, the number of test sets is 100, the least squares method uses all 600 points for position calculation, and the distance error obeys N (0, σ)2) Normal distribution of (a) ("a")214. The performance of these three positioning algorithms was compared by varying the number of anchor nodes, and the results are shown in fig. 5. The performance of the neural network algorithm is best of the three algorithms when the number of anchor nodes is changed, e.g., the MSE of the neural network algorithm is about 5m lower than that of the least squares method when only 7 anchor nodes are used2And 10m lower than that of the support vector machine2Left and right; when all 12 anchor nodes are used, the performance of the least square method and the performance of the support vector machine are comparable, and the MSE of the neural network algorithm is 8m lower than that of the least square method and the support vector machine2Left and right. When the anchor nodes are changed, the performance of the support vector machine and the least square method is good and bad respectively, for example, when the number of the anchor nodes is less than 10, the performance of the least square method is better than that of the support vector machine, and when the number of the anchor nodes is more than 10, the performance of the least square method is worse than that of the support vector machine.
Then, the invention fixes the number of anchor nodes, uniformly selects 12 anchor nodes, and the distance error still obeys N (0, sigma)2) Normal distribution of (a) ("a")2The performance of the three algorithms was compared by varying the size of the training set, as shown in fig. 6. From FIG. 6, the performance of the neural network algorithm is optimal among the three algorithms, while the support vector machineThe performance of (2) is not satisfactory, for example, when the training set has only 100 points, the error of the support vector machine is as high as 25.7m2And the error of the least squares method is centered at 13.3m2The error of the neural network algorithm is that the best error is only 8.5m2. And when the training set is large, the advantages of the machine learning algorithm are shown. For example, when the training set size is 600, the traditional least square method has the worst positioning effect, which is 12.6m2Left and right, the support vector machine performance is centered at 10.5m2On the left and right, the same neural network algorithm has the highest performance, only 5.3m2。
Finally, the invention fixes the sizes of the anchor nodes, the training sets and the test sets, wherein the number of the anchor nodes is 12, the number of the training sets is 500, the number of the test sets is 100, the least square method uses all 600 points to calculate the position, and the performances of the three algorithms are compared by changing the distance error variance, as shown in fig. 7. As can be seen from fig. 7, the neural network algorithm still performs the best among the three algorithms, and the performance of the support vector machine is comparable to that of the least square method. For example, when the variance of the range error is 9, the MSE of the neural network algorithm is only 3.6m2While the MSE of the least squares method and the support vector machine is as high as 10.6m2And 12.2m2When the variance of the range error rises to 39, the MSE of the neural network algorithm is also only 10m2At this time, the MSE of the least square method and the support vector machine is as high as 16.9m2And 15.4m2. In addition, in comparison of the performance of the least square method and the support vector machine, the performance of the least square method is better when the variance of the distance error is smaller, and the performance of the least square method is worse when the variance of the distance error increases to a certain degree, but the performance of the support vector machine is better. This shows that, in the underwater positioning, when the measured distance between the unknown node and the anchor node is very accurate, the least square method is selected to receive good effect, but when the obtained distance value error is large, the positioning effect of the traditional least square method is poor, but the method for positioning by using the machine learning algorithm provided by the invention has tolerance to the distance errorIs relatively strong.
The foregoing is a more detailed description of the invention in connection with specific/preferred embodiments and is not intended to limit the practice of the invention to those descriptions. It will be apparent to those skilled in the art that various substitutions and modifications can be made to the described embodiments without departing from the spirit of the invention, and such substitutions and modifications are to be considered as within the scope of the invention.
Claims (7)
1. An underwater positioning method based on distance measurement and neural network algorithm is characterized by comprising the following steps:
a1, arranging K anchor nodes and N nodes to be positioned in a positioning water area, and measuring the distance between each node to be positioned and the K anchor nodes;
positioning in two stages of A2, separation line training and online positioning;
a3, the off-line training phase comprises the following steps:
a31, measuring the distance between the nth node to be positioned and all anchor nodes to formDistance vector dn;
A32, calculating the distance vector dnInputting three layers of neural networks: the device comprises a neural network input layer, a neural network hiding layer and a neural network output layer;
a33, iteratively updating the prediction model parameters through an algorithm, and outputting the neural network model parameters;
a4, the online positioning stage comprises the following steps:
a41, measuring the current distance between the nth node to be positioned and all anchor nodes to form a distance vector sn;
A42, calculating the distance vector snInputting neural network model parameters obtained in an off-line training stage;
a43, outputting the position coordinates of the node to be positioned;
wherein:
the distance vector d in step A31nRepresented by the following formula:
dn=[dn,1,…,dn,k,…,dn,K]T
wherein: dn,kRepresents the distance between the nth node to be positioned and the kth anchor node,
n belongs to [1, N ], K belongs to [1, K ], and T represents transposition operation;
in the step a32, the method comprises the steps of,
the neural network input layer: the total number of the nodes is K +1, the first K nodes are active nodes, the first K nodes are correspondingly input nodes, and the input and the output of the kth node are dn,k,k∈[1,K]The K +1 th node of the input layer is a bias node, and the output is
The neural network hidden layer: the total number of the nodes is U +1, the first U are active nodes, wherein the input of the U-th active node isOutput is asu∈[1,U],For the transfer parameter vector from the input layer to the U-th active node of the hidden layer, f (x) represents the active function, and the output vectors of all the U active nodes areThe U +1 th node of the hidden layer is a bias node, and the output is
The neural network output layer: only 1 node, with inputs ofOutput is asWherein,for the hidden layer to output node transfer parameter vector, (x)n,yn) Is the position coordinate of the nth node to be positioned.
2. The underwater positioning method of claim 1,
the activation function f (x) is a sigmoid function, i.e. f (x) 1/(1+ exp (-x)).
3. The underwater positioning method of claim 1,
in step a33, a batch gradient descent method is used, and a prediction model parameter converging to an optimal value is obtained through multiple iterative updates:
4. underwater positioning method according to claim 3,
and calculating partial derivatives by using a back propagation algorithm in the multiple iterative updating.
5. The underwater positioning method of claim 1,
the distance vector in step a41 is represented by the following formula:
sn=[sn,1,…,sn,k,…,sn,K]T
wherein: sn,kRepresents the current distance between the nth node to be positioned and the kth anchor node,
n∈[1,N],k∈[1,K]。
6. underwater positioning method according to claim 1, wherein the distance vector s is used in step A42nInputting neural network model parameters obtained in an off-line training stageThe method comprises the following steps:
a421, a neural network input layer: the input and output of the kth node are sn,k,k∈[1,K]The bias node output is
A422, a neural network hidden layer: the u-th active node is input asOutput is asu∈[1,U]All U active nodes output vectors ofBiased node output is
A423, a neural network output layer: input is asOutput is as(xn,yn) Is the position coordinate of the nth node to be positioned.
7. The underwater positioning method as claimed in claim 1, wherein the distance of each node to be positioned from all anchor nodes is measured using a TDOA method.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610389518.8A CN106093849B (en) | 2016-06-03 | 2016-06-03 | A kind of Underwater Navigation method based on ranging and neural network algorithm |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610389518.8A CN106093849B (en) | 2016-06-03 | 2016-06-03 | A kind of Underwater Navigation method based on ranging and neural network algorithm |
Publications (2)
Publication Number | Publication Date |
---|---|
CN106093849A CN106093849A (en) | 2016-11-09 |
CN106093849B true CN106093849B (en) | 2018-05-25 |
Family
ID=57447546
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201610389518.8A Active CN106093849B (en) | 2016-06-03 | 2016-06-03 | A kind of Underwater Navigation method based on ranging and neural network algorithm |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN106093849B (en) |
Families Citing this family (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107135541A (en) * | 2017-04-21 | 2017-09-05 | 广东技术师范学院 | UWB indoor localization method based on OPTICS Density Clusterings and BP neural network |
CN108491674B (en) * | 2018-05-28 | 2023-07-18 | 西南交通大学 | Underwater discharge forming working machine detection system and positioning error prediction method |
CN109696698B (en) * | 2019-03-05 | 2021-03-12 | 湖南国科微电子股份有限公司 | Navigation positioning prediction method and device, electronic equipment and storage medium |
CN109996172A (en) * | 2019-03-14 | 2019-07-09 | 北京工业大学 | One kind being based on BP neural network precision indoor positioning system and localization method |
CN110082717A (en) * | 2019-04-30 | 2019-08-02 | 上海海事大学 | A kind of underwater wireless sensor node positioning method |
CN110798885A (en) * | 2019-11-03 | 2020-02-14 | 桂林电子科技大学 | Underwater positioning method based on ranging and neural network algorithm |
CN111624586B (en) * | 2020-05-27 | 2022-09-23 | 苏州大学 | Underwater acoustic target ranging method based on feature extraction and neural network |
CN114216464B (en) * | 2021-11-16 | 2023-08-25 | 湖南国天电子科技有限公司 | Intelligent underwater target positioning method |
CN114563763B (en) * | 2022-01-21 | 2022-10-21 | 青海师范大学 | Underwater sensor network node distance measurement positioning method based on return-to-zero neurodynamics |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101695190A (en) * | 2009-10-20 | 2010-04-14 | 北京航空航天大学 | Three-dimensional wireless sensor network node self-locating method based on neural network |
CN102883430A (en) * | 2012-09-12 | 2013-01-16 | 南京邮电大学 | Range-based wireless sensing network node positioning method |
US9129223B1 (en) * | 2011-03-30 | 2015-09-08 | Amazon Technologies, Inc. | Sound localization with artificial neural network |
Family Cites Families (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP6576624B2 (en) * | 2014-09-24 | 2019-09-18 | 五洋建設株式会社 | Underwater positioning system and underwater positioning method |
-
2016
- 2016-06-03 CN CN201610389518.8A patent/CN106093849B/en active Active
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101695190A (en) * | 2009-10-20 | 2010-04-14 | 北京航空航天大学 | Three-dimensional wireless sensor network node self-locating method based on neural network |
US9129223B1 (en) * | 2011-03-30 | 2015-09-08 | Amazon Technologies, Inc. | Sound localization with artificial neural network |
CN102883430A (en) * | 2012-09-12 | 2013-01-16 | 南京邮电大学 | Range-based wireless sensing network node positioning method |
Non-Patent Citations (1)
Title |
---|
基于神经网络的TDOA定位改进算法研究;乔梁等;《河南师范大学学报(自然科学版)》;20140731;第42卷(第4期);139-143 * |
Also Published As
Publication number | Publication date |
---|---|
CN106093849A (en) | 2016-11-09 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN106093849B (en) | A kind of Underwater Navigation method based on ranging and neural network algorithm | |
CN106912105B (en) | Three-dimensional positioning method based on PSO _ BP neural network | |
CN110940949B (en) | Method for estimating DOA of reciprocal array based on quantum penguin search mechanism in strong impact noise environment | |
CN108896962B (en) | Iterative positioning method based on sound position fingerprint | |
CN109581281B (en) | Moving target positioning method based on arrival time difference and arrival frequency difference | |
CN108845325A (en) | Towed linear-array sonar submatrix error misfits estimation method | |
CN110132281B (en) | Underwater high-speed target high-precision autonomous acoustic navigation method based on inquiry response mode | |
WO2016112758A1 (en) | Method and apparatus for locating terminal | |
CN105445718A (en) | DOA estimation method for distributed multi-carrier-warship BVR (beyond visual range) radar based on array reconfiguration | |
CN112986907B (en) | Moving target positioning method under clock deviation and clock drift conditions | |
CN111505580B (en) | Multi-platform cooperative target positioning method based on azimuth angle and Doppler information | |
CN110278525B (en) | High-precision indoor wireless positioning method | |
CN113189541B (en) | Positioning method, device and equipment | |
CN108761384B (en) | Target positioning method for robust sensor network | |
CN108924734B (en) | Three-dimensional sensor node positioning method and system | |
Yan et al. | UAM-RDE: an uncertainty analysis method for RSSI-based distance estimation in wireless sensor networks | |
CN110703187A (en) | Acoustic beacon direction finding and positioning method based on underwater mobile platform | |
CN111263295A (en) | WLAN indoor positioning method and device | |
CN107197519B (en) | Underwater target positioning method based on improved least square support vector machine | |
CN113970762B (en) | Multistage interference source positioning method and system | |
CN115469314A (en) | Uniform circular array steady underwater target azimuth tracking method and system | |
CN111856400B (en) | Underwater target sound source positioning method and system | |
CN114679683A (en) | Indoor intelligent positioning method based on derivative fingerprint migration | |
CN112437397A (en) | Distributed sensor node positioning method based on alternative correction Newton method | |
Yong et al. | C-VoNNI: a precise fingerprint construction for indoor positioning systems using natural neighbor methods with clustering-based Voronoi diagrams |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |