CN111479231B - Indoor fingerprint positioning method for millimeter wave large-scale MIMO system - Google Patents

Indoor fingerprint positioning method for millimeter wave large-scale MIMO system Download PDF

Info

Publication number
CN111479231B
CN111479231B CN202010307606.5A CN202010307606A CN111479231B CN 111479231 B CN111479231 B CN 111479231B CN 202010307606 A CN202010307606 A CN 202010307606A CN 111479231 B CN111479231 B CN 111479231B
Authority
CN
China
Prior art keywords
positioning
grained
coarse
fingerprint
fine
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
Application number
CN202010307606.5A
Other languages
Chinese (zh)
Other versions
CN111479231A (en
Inventor
范建存
王建鹏
罗新民
张莹
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Xian Jiaotong University
Original Assignee
Xian Jiaotong University
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Xian Jiaotong University filed Critical Xian Jiaotong University
Priority to CN202010307606.5A priority Critical patent/CN111479231B/en
Publication of CN111479231A publication Critical patent/CN111479231A/en
Application granted granted Critical
Publication of CN111479231B publication Critical patent/CN111479231B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04W—WIRELESS COMMUNICATION NETWORKS
    • H04W4/00—Services specially adapted for wireless communication networks; Facilities therefor
    • H04W4/30—Services specially adapted for particular environments, situations or purposes
    • H04W4/33—Services specially adapted for particular environments, situations or purposes for indoor environments, e.g. buildings
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04B—TRANSMISSION
    • H04B7/00—Radio transmission systems, i.e. using radiation field
    • H04B7/02—Diversity systems; Multi-antenna system, i.e. transmission or reception using multiple antennas
    • H04B7/04—Diversity systems; Multi-antenna system, i.e. transmission or reception using multiple antennas using two or more spaced independent antennas
    • H04B7/0413—MIMO systems
    • 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

Landscapes

  • Engineering & Computer Science (AREA)
  • Computer Networks & Wireless Communication (AREA)
  • Signal Processing (AREA)
  • Position Fixing By Use Of Radio Waves (AREA)

Abstract

The invention discloses an indoor fingerprint positioning method for a millimeter wave large-scale MIMO system. The method is based on the characteristics of a millimeter wave large-scale MIMO system, and combines a machine learning technology to realize a coarse-to-fine positioning process. In the coarse-grained location scheme, AOA information is used as fingerprint features, and a deep learning method is combined to adaptively extract feature quantities and establish a fingerprint library so as to realize coarse-grained location. In a fine-grained positioning scheme, in order to overcome high complexity caused by multiple antennas, an averaged multipath CSI amplitude is selected as a fingerprint feature for positioning, and a dynamic weighted K-nearest neighbor algorithm based on spatial mapping is provided to realize better positioning accuracy. Simulation shows that the invention can greatly improve the positioning precision. Compared with the existing positioning scheme based on the fingerprint, the invention adopts the single base station, has lower cost, simultaneously adopts the hierarchical positioning model, reduces the complexity of positioning operation, and can realize higher positioning precision in a larger positioning scene.

Description

Indoor fingerprint positioning method for millimeter wave large-scale MIMO system
Technical Field
The invention belongs to the technical field of communication, and particularly relates to an indoor fingerprint positioning method for a millimeter wave large-scale MIMO system.
Background
With the rapid development of wireless technology and the Internet, the Internet of Things (IoT) has become an indispensable part of people's daily life. Location-based services are one of the most attractive applications related to the internet of things. The rapid development of the internet of things technology has promoted the development of the application of many high-precision positioning services. For example: future automatic driving automobiles must reach centimeter-level positioning accuracy to ensure accurate high-speed driving of the vehicles. For certain indoor scenarios, such as large intelligent factories, goods picking or inventory by intelligent robots requires precise location information of the goods to speed up the picking process. In addition, baggage tracking at airports, personnel location for underground mining, smart home management and health monitoring all require high precision location.
The main positioning technology in the outdoor area is a Global Navigation Satellite System (GNSS) based positioning system, whereas in the indoor area, GNSS cannot provide satisfactory positioning performance due to its complex radio propagation environment. Therefore, in recent years, indoor positioning technology based on radio signals has been rapidly developed. Among these signals, the most common ones are WiFi and bluetooth, which are currently in smart devices. Fingerprint positioning technology has attracted extensive attention in academia and industry as one of the potential solutions for indoor positioning implemented by radio signals. The fingerprint positioning method comprises two stages: an off-line training phase and an on-line positioning phase. In the off-line stage, the professional samples the location of the location area, collects the wireless signal characteristics at each sampling location, and stores the wireless signal characteristics in a location-fingerprint database. In the on-line positioning stage, the user sends the wireless signal fingerprint of the position to the positioning server, the server matches the inquiry fingerprint with the database, and the position corresponding to the most similar fingerprint is used as the position estimation of the user and is returned to the user.
Received Signal Strength (RSS) is most commonly used for fingerprint location because it is simple and easy to measure. However, RSS is sensitive to environmental variations due to shadow fading and multipath effects. In addition, RSS can only embody coarse channel information. Unlike RSS, Channel State Information (CSI) may also reflect channel characteristics such as shadowing and multipath effects experienced by the received signal. By extracting CSI, fine-grained physical layer information of the system can be obtained, which is why CSI is widely used for indoor fingerprint positioning in recent years. Furthermore, millimeter wave (mmWave) technology is considered one of the key technologies for 5G networks and can be utilized to determine the real-time location of a vehicle. The characteristics of high bandwidth and high path loss of millimeter waves can better observe frequency selective fading caused by multipath in a frequency domain, and improve the resolution of a receiver on multipath signals corresponding to a time domain. In addition, millimeter waves can provide larger effective signal bandwidth in a high-frequency band, and the theoretical boundary of positioning accuracy is improved.
Disclosure of Invention
The invention aims to provide an indoor fingerprint positioning method for a millimeter wave large-scale MIMO system, which adopts a fingerprint positioning technology, respectively uses CSI amplitude information and AOA information obtained through CSI estimation as fingerprint characteristics, and is carried out in two stages of coarse-grained positioning and fine-grained positioning to realize coarse-to-fine positioning.
In order to achieve the purpose, the invention is realized by adopting the following technical scheme:
an indoor fingerprint positioning method for a millimeter wave large-scale MIMO system is characterized in that a receiving end acquires a CSI multipath signal with a separable time domain of the millimeter wave large-scale MIMO system, and performs data collection and preprocessing on the CSI multipath signal; in the coarse-grained positioning scheme, AOA information estimated by a multipath CSI signal is used as fingerprint characteristics, a deep learning method is combined, characteristic quantities are extracted in a self-adaptive mode, and a fingerprint database is established to achieve coarse-grained positioning; in order to overcome the high complexity caused by multiple antennas, in a fine-grained positioning scheme, positioning is carried out by taking averaged multipath CSI (channel state information) as fingerprint characteristics; meanwhile, the problem that the traditional WKNN algorithm depends on a pre-selected fixed K value to a great extent is considered, and a dynamic weighting K nearest neighbor algorithm is adopted to achieve better positioning accuracy.
The invention is further improved in that the method is as follows:
firstly, in a coarse-grained positioning stage, estimating acquired multipath CSI fingerprint data by using a classical MUSIC algorithm to obtain multipath AOA information as fingerprint characteristics, extracting the fingerprint characteristics in a self-adaptive manner through a multilayer convolutional neural network to construct a fingerprint database, and then using a position label with the maximum output probability as a final estimation position in the coarse-grained positioning stage through a softmax classifier;
then, in a fine-grained positioning stage, on the basis of the coarse-grained positioning labels obtained through estimation, fine-grained labels are generated by taking the coarse-grained positioning labels as a core, and considering that the more the number of the fine-grained labels is, the finer the fine-grained positioning result is, the averaged CSI amplitude information is selected as the fingerprint characteristic of the fine-grained positioning stage, and a dynamic weighted K nearest neighbor algorithm based on space mapping is provided for fine-grained positioning;
finally, in order to improve the positioning accuracy, at the fine-grained stage, the following facts are considered: the same CSI difference may correspond to different geometric distances, that is, for CSI differences of different magnitudes, the signal distances represented by the CSI difference may correspond to different geometric distances; therefore, the relationship between the signal distance space and the geometric distance space is established by training the ELM to realize the spatial mapping and prevent the influence of the distance mismatch from causing the reduction of the positioning precision.
The invention has the further improvement that the data collection and pretreatment in the coarse-grained location stage specifically comprises the following steps:
step 1: uniformly dividing cells into N1A block having its geometric center as a classification position label;
step 2: in the coarse-grained location stage, the selected coarse-grained location labels are sparsely distributed and are few in number, so that the N divided is1Multipath CSI information respectively collected in the blocks is estimated through a classical MUSIC estimation algorithm to obtain multipath AOA information of all coarse-grained position label points, and the multipath AOA information is used as fingerprint characteristics;
and step 3: adding the sample characteristics into a label, wherein the sample in the coarse-grained positioning stage is represented as follows:
Figure BDA0002456329910000031
wherein phi is0AOA, phi of direct path1For the AOA of the first scatter path,
Figure BDA0002456329910000032
is the NthrayThe AOA, T of the bar scattering path is matrix transpose.
The invention has the further improvement that the task of the coarse-grained location stage is to train the parameters of the deep convolutional neural network and the regression classifier network according to the acquired labeled training data, and the training aim is to minimize the mean square error output by the training label and the network;
the off-line stage training process of the coarse-grained location link is as follows:
for a deep learning network, a deep convolutional neural network is adopted, a ReLU function is adopted as an excitation function of each layer of nodes, after training data are input into the network, the output of each layer is obtained according to the excitation function and is used as the input of the next layer, and finally, the network output is obtained through layer-by-layer forward propagation; constructing a penalty function according to a minimum mean square error principle, updating and iterating by using a random gradient descent algorithm to obtain a final training parameter, and storing a trained weight W, b as a part of a fingerprint library;
and then, taking the training output data of the neural network as the input of a softmax classifier, then dividing the training output data into C classes, taking the probability that the input data belongs to each class as the output of the classifier, constructing a penalty function according to a minimum mean square error principle, updating and iterating by using a random gradient descent algorithm to obtain a final training parameter, and combining W, b and theta together to form a fingerprint library, wherein the theta is a classifier parameter.
A further improvement of the invention is that the output of the classifier is as follows:
Figure BDA0002456329910000041
wherein the content of the first and second substances,
Figure BDA0002456329910000042
is a C1 matrix, each term is represented in
Figure BDA0002456329910000043
In the given case of the situation where,
Figure BDA0002456329910000044
the probability of belonging to each of the classes,
Figure BDA0002456329910000045
training output for neural networksI.e., the input to the regression classifier, and θ is the parameter of the classifier.
The further improvement of the invention is that the on-line stage positioning process of the coarse grain positioning link is as follows:
step 1: after receiving CSI information from users at unknown positions, the MUSIC algorithm estimates the received AOA information, namely
Figure BDA0002456329910000046
The probability that the unknown data belong to each to-be-determined position is obtained through forward propagation of a machine learning network and classification of a regression classifier;
step 2: and using a probability method to take the position label with the maximum output probability as the final position estimation of coarse-grained positioning.
The invention has the further improvement that the data collection and pretreatment in the fine-grained positioning stage specifically comprises the following steps:
step 1: based on the coarse-grained positioning result, the coarse-grained positioning result is used as a core, the coarse-grained positioning result is expanded outwards at equal intervals to generate a fine-grained label called an expanded sliding window, and N is generated in the sliding window according to the method2Fine-grained location tags;
step 2: for a fine-grained positioning stage, in order to further reduce data dimension and operation complexity, sampling CSI data of each path, then averaging amplitude channel matrixes on different antennas, and grouping and numbering the amplitude data according to the corresponding relation between known CSI information and user positions;
and step 3: adding the sample characteristics into a label, wherein the sample in the fine-grained positioning stage is represented as follows after the label is added:
Figure BDA0002456329910000051
wherein the content of the first and second substances,
Figure BDA0002456329910000052
N2the position label number of the fine-grained positioning stage.
The invention has the further improvement that the task of the fine-grained positioning stage is to outwards extend a sliding window which is full of fine-grained virtual labels by taking the estimated position of the fine-grained positioning stage as the center on the basis of coarse-grained positioning; considering that phase shift exists in CSI, and in order to overcome high complexity brought by multiple antennas, averaged multipath CSI amplitude information is used as a position label, and a space mapping-based adaptive dynamic weighting K nearest neighbor algorithm comprises the following algorithm steps:
step 1: calculating the signal distance between all sample points of the fingerprint library:
Figure BDA0002456329910000053
step 2: deleting singular distance, and setting threshold T ═ alpha × D1Keeping the signal distance which is not more than the threshold value T in the formula, and sequentially marking as D from small to large1,...,DSS-1, 2, …, S represents the number of retained neighboring reference points;
and step 3: calculate the mean distance difference of the retention points from each other:
Figure BDA0002456329910000054
wherein, Δ dj,sRepresents DjAnd DsThe distance difference of (a);
and 4, step 4: estimating the final position, the dynamic weight value is expressed as:
Figure BDA0002456329910000055
wherein Δ D ═ DK-D1(ii) a In particular, when DK＝D1Time, omega j1 is ═ 1; defining:
Figure BDA0002456329910000056
the specific form of the final position estimate is then as follows:
Figure BDA0002456329910000061
wherein (x)j,yj) The coordinates of the jth remaining location tag are represented.
The further improvement of the present invention is that, in the proposed dynamic weighted K-nearest neighbor algorithm step (1) based on spatial mapping, when calculating the signal distance, in consideration of the fact that the signal distance and the physical distance are not matched at different CSI amplitude levels, a spatial mapping method based on an extreme learning machine is adopted, and this problem is described as follows:
Figure BDA0002456329910000062
wherein S isDAnd SCRespectively signal space distance and geometric space distance.
The invention has at least the following beneficial technical effects:
the good positioning method does not only pursue positioning accuracy, but pursues the target of meeting application requirements and adapting to environmental characteristics. Therefore, the present invention is directed to scenarios with a large localization range, such as: in indoor scenes such as large indoor parking lots and large intelligent factories, an indoor fingerprint positioning method for a millimeter wave large-scale MIMO system is designed by adopting a fingerprint positioning technology based on wireless signals, so that a coarse-to-fine positioning process is realized, and higher positioning accuracy can be achieved. The core of the invention is that a receiving end acquires the CSI multipath signals of the millimeter wave large-scale MIMO system, which can be separated in time domain, and performs data collection and pretreatment on the multipath signals. In the coarse-grained positioning scheme, AOA information estimated by multipath CSI signals is used as fingerprint characteristics, a deep learning method is combined, characteristic quantities are extracted in a self-adaptive mode, and a fingerprint database is established to achieve coarse-grained positioning. In order to overcome the high complexity caused by multiple antennas, in a fine-grained positioning scheme, the averaged multipath CSI is used as fingerprint characteristics for positioning. Meanwhile, a dynamic weighting K nearest neighbor algorithm is designed in consideration of the problem that the traditional WKNN algorithm depends on a pre-selected fixed K value to a great extent, so that better positioning accuracy is realized. In summary, the present invention has the following advantages:
1. the invention provides an indoor fingerprint positioning method based on a hierarchical model, which is used for a millimeter wave large-scale MIMO system, realizes a coarse-to-fine positioning process, reduces the complexity of the system, and reduces the positioning cost by adopting a single base station.
2. The method adopts multipath AOA information as coarse-grained fingerprints, and utilizes a convolutional neural network and a softmax regression classifier to adaptively extract fingerprint characteristics to construct a fingerprint database. Because the indoor environment has rich multi-path and the direct path does not exist in most cases, the precision of the traditional AOA estimation algorithm depends on the direct path to a great extent, so the estimation precision is greatly reduced in an indoor scene, and the small error of the AOA can cause serious positioning error, especially in a scene with a large positioning range. The invention takes the multipath AOA information as the fingerprint characteristics, realizes the fingerprint positioning by utilizing the uniqueness of the multipath AOA at different positions, not only overcomes the difficult problem of high-precision AOA estimation in multipath environment, but also improves the positioning precision. In addition, the convolutional neural network has excellent performance in the aspect of video image classification, the convolutional neural network is used for realizing feature extraction, a fingerprint database is constructed, compared with other networks such as a BP (back propagation) network, the method has richer extracted features, and the positioning performance is favorably improved.
3. In the fine-grained positioning stage, in order to reduce the operation complexity, the invention only selects the amplitude value after the channel state information is averaged as the fingerprint characteristic, and uses the averaged multipath CSI information as the fingerprint characteristic for positioning. Meanwhile, the problem that the traditional WKNN algorithm depends on the selected fixed K value to a great extent is considered, and a dynamic weighting K nearest neighbor algorithm is provided so as to achieve better positioning accuracy. The traditional WKNN algorithm depends on a preselected fixed K value, and the dynamic weighted K nearest neighbor algorithm provided by the invention ensures that each target is allocated with a proper K value when being positioned, so that the positioning performance is greatly improved. In addition, the invention also considers the problem that the signal distance in the algorithm is not matched with the corresponding real physical distance, researches the mapping relation of the signal distance and the real physical distance, and realizes the mapping of the signal space and the physical space by utilizing a machine learning method (ELM), thereby realizing high-precision positioning. As far as i know, the space mapping is realized by utilizing a machine learning algorithm for the first time, and good positioning performance is obtained.
4. The invention aims at indoor scenes, especially scenes with a large positioning range, such as large-scale indoor factories, intelligent indoor parking lots and the like, the positioning advantages are more obvious, and the positioning precision is effectively improved. The simulation proves that the method is 120 multiplied by 70 multiplied by 3m3The indoor environment of (2) can realize the positioning accuracy of decimeter level.
Drawings
Fig. 1 is a schematic diagram of a 3D indoor positioning scene;
FIG. 2 is a schematic view of a hierarchical positioning system flowchart;
FIG. 3 is a schematic diagram of a data training process of a convolutional neural network;
FIG. 4 is a schematic diagram of a virtual tag in a fine-grained positioning scheme;
FIG. 5 is a schematic diagram of a positioning effect of randomly selecting 120 test points;
FIG. 6 is a comparison of the performance of the present invention and a positioning scheme;
FIG. 7 is a graph of the effect of refining the number of virtual tags on the average positioning error;
fig. 8 illustrates the effect of the threshold coefficient a on the average positioning error.
Detailed Description
The technical solution of the present invention is further described in detail by the accompanying drawings and embodiments.
Positioning environment referring to fig. 1, the 3D indoor positioning environment includes a base station (Access point) with a known location and multiple antennas and a device (MP) to be positioned with an unknown location as a receiver. Since in mmWave channels, the direct path (LOS) dominates, while the non-direct path (NLOS) is severely attenuated after one or two reflections. Thus, only single-hop and double-hop reflections are considered in the considered positioning environment. Since the receiver (MP) is located at a certain height above the ground, the single hop takes into account the ground reflections in addition to the reflections from the four walls.
The positioning system flow diagram is shown in fig. 2. Aiming at a scene with a large positioning range, the positioning process is divided into two stages of coarse-grained positioning (on the left side of a figure) and fine-grained positioning (on the right side of the figure).
Firstly, in a coarse-grained positioning stage, multipath AOA information is obtained by estimating collected multipath CSI fingerprint data by using a classical MUSIC algorithm and is used as fingerprint characteristics, and then the fingerprint characteristics are extracted in a self-adaptive manner through a multilayer convolutional neural network to construct a fingerprint database. And then, through a softmax classifier, taking the position label with the maximum output probability as a final estimated position of the coarse-grained positioning stage.
And then, in a fine-grained positioning stage, on the basis of the coarse-grained positioning labels obtained through estimation, fine-grained labels are generated by taking the coarse-grained positioning labels as a core, and considering that the more the number of the fine-grained labels is, the finer the fine-grained positioning result is, and the calculation cost of AOA estimation is higher, so that the averaged CSI amplitude information is selected as the fingerprint characteristics of the fine-grained positioning stage, and a dynamic weighting K nearest neighbor algorithm based on spatial mapping is provided for fine-grained positioning.
1) The specific process of the coarse grain positioning stage is as follows:
(1) the data collection and pretreatment in the coarse-grained positioning stage specifically comprise the following steps:
step 1: uniformly dividing cells into N1And the geometric center of the block serves as a classification position label. See fig. 4.
Step 2: in the coarse-grained location stage, the selected coarse-grained location labels are sparsely distributed and less in number, so that the N is divided1Multipath CSI information respectively collected in the blocks is estimated through a classical MUSIC algorithm to obtain multipath AOA information of all coarse-grained position label points, and the multipath AOA information is used as fingerprint characteristics;
and step 3: sample features are added to the label. The samples in the coarse-grained location stage after adding the labels can be expressed as:
Figure BDA0002456329910000091
wherein the content of the first and second substances,φ0AOA, phi of direct path1For the AOA of the first scatter path,
Figure BDA0002456329910000092
is the NthrayThe AOA, T of the bar scattering path is matrix transpose.
(2) The off-line stage training process of the coarse-grained location link is as follows:
for a deep learning network, a deep convolutional neural network is adopted, a ReLU function is adopted as an excitation function of each layer of nodes, after training data are input into the network, the output of each layer is obtained according to the excitation function and is used as the input of the next layer, and finally, the network output is obtained through layer-by-layer forward propagation; constructing a penalty function according to a minimum mean square error principle, updating and iterating by using a random gradient descent algorithm to obtain a final training parameter, and storing a trained weight W, b as a part of a fingerprint library;
and then, taking the training output data of the neural network as the input of a softmax classifier, then dividing the training output data into C classes, taking the probability that the input data belongs to each class as the output of the classifier, constructing a penalty function according to a minimum mean square error principle, updating and iterating by using a random gradient descent algorithm to obtain a final training parameter, and combining W, b and theta together to form a fingerprint library, wherein the theta is a classifier parameter.
The output of the classifier is as follows:
Figure BDA0002456329910000093
wherein the content of the first and second substances,
Figure BDA0002456329910000094
is a C1 matrix, each term is represented in
Figure BDA0002456329910000095
In the given case of the situation where,
Figure BDA0002456329910000096
the probability of belonging to each of the classes,
Figure BDA0002456329910000101
and theta is a parameter of the classifier, and is a training output of the neural network, namely an input of the regression classifier.
The process of coarse-grained location phase data training using the Deep Convolutional Neural Network (DCNN) is shown in fig. 3. First, AOA data is obtained as input, and AOA values for each position are estimated by MUSIC. Then, an AOA tensor is established for each location, which is simple and convenient for DCNN to process in all layers. In the first convolutional layer and the downsampling layer, a convolution operation is performed using 18 convolution kernels of size 2 × 1 to obtain the same number of feature maps as the input. In order to ensure invariance of the feature maps, downsampling is performed at a size of 2 × 1 to obtain the same number of feature maps. For all layers in the designed DCNN, convolution and downsampling with convolution kernels of the same size are chosen for the convolutional and pooling layers for simplicity. Except for the step sizes of the second and fourth pooling layers. In addition, it is noted that a Dropout layer is introduced into the designed DCNN, that is, in the training process, the weight of the DCNN is temporarily discarded with a certain probability and is not updated temporarily, so that overfitting can be effectively prevented.
(3) The on-line stage positioning process of the coarse grain positioning link is as follows:
step 1: after receiving CSI information from users at unknown positions, the MUSIC algorithm estimates the received AOA information, namely
Figure BDA0002456329910000102
And (4) obtaining the probability that the unknown data belongs to each to-be-determined position through forward propagation of a machine learning network and classification of a regression classifier.
Step 2: and using a simple probability method, and taking the position label with the maximum output probability as the final position estimation of coarse-grained positioning.
2) The task of the fine-grained positioning stage is to outwards extend a sliding window full of fine-grained virtual labels by taking the estimated position of the fine-grained positioning stage as the center on the basis of coarse-grained positioning. Considering that phase shift exists in CSI, and in order to overcome high complexity brought by multiple antennas, averaged multipath CSI amplitude information is used as a position label, and a space mapping-based adaptive dynamic weighting K nearest neighbor algorithm is provided. The virtual tag building process in the fine-grained positioning scheme is shown in fig. 4. Wherein, the left side of the figure is a coarse-grained position label distribution schematic diagram, and the five-pointed star marks a coarse-grained position label; and the right side of the diagram is a fine-grained position label distribution schematic diagram, a circular mark represents a fine label for fine-grained positioning, and a quadrangle star mark represents a target position to be positioned.
(1) The data collection and pretreatment of the fine-grained positioning stage specifically comprise the following steps:
step 1: on the basis of the coarse-grained positioning result, the coarse-grained positioning result is used as a core, and the fine-grained label is generated by outwards expanding the coarse-grained positioning result at equal intervals, as shown in fig. 4, the fine-grained label is called as an expanded sliding window. According to this method, N is generated in the sliding window2Fine-grained location tags.
Step 2: for the fine-grained positioning stage, in order to further reduce the data dimension and reduce the operation complexity, the CSI data of each path are sampled, then amplitude channel matrixes on different antennas are averaged, and the amplitude data are grouped and numbered according to the corresponding relation between the known CSI information and the user position.
And step 3: sample features are added to the label. The samples in the fine-grained location stage can be represented as follows after the tags are added:
Figure BDA0002456329910000111
wherein the content of the first and second substances,
Figure BDA0002456329910000112
N2the position label number of the fine-grained positioning stage. The above is the input data of the two phases.
(2) Aiming at the problem that the traditional WKNN algorithm depends on a selected fixed K value to a great extent, a dynamic weighting K nearest neighbor algorithm based on space mapping is provided in a fine-grained positioning stage so as to realize better positioning accuracy. The algorithm comprises the following steps:
step 1: calculating the signal distance between all sample points of the fingerprint library:
Figure BDA0002456329910000113
step 2: and deleting the singular distance. Setting threshold value T ═ α × D1Keeping the signal distance which is not more than the threshold value T in the formula, and sequentially marking as D from small to large1,...,DSAnd S is 1,2, …, S represents the number of retained neighboring reference points.
And step 3: calculate the mean distance difference of the retention points from each other:
Figure BDA0002456329910000114
wherein, Δ dj,sRepresents DjAnd DsThe distance difference of (a).
And 4, step 4: the position is estimated. The dynamic weight value is expressed as:
Figure BDA0002456329910000121
wherein Δ D ═ DK-D1. In particular, when DK＝D1Time, omega j1. Defining:
Figure BDA0002456329910000122
the specific form of the final position estimate is then as follows:
Figure BDA0002456329910000123
wherein (x)j,yj) The coordinates of the jth remaining location tag are represented.
In the proposed dynamic weighted K-nearest neighbor algorithm step (1) based on spatial mapping, a spatial mapping method based on an Extreme Learning Machine (ELM) is proposed in consideration of the fact that the signal distance and the physical distance are not matched at different CSI amplitude levels when calculating the signal distance. This problem is described as follows:
Figure BDA0002456329910000124
wherein S isDAnd SCRespectively signal space distance and geometric space distance. The purpose of (1) is to try to find a mapping relation between the two. Therefore, an ELM-based adaptive spatial mapping method is designed. This is known to be the first time that machine learning is used for spatial mapping of signal spatial distances and geometric spatial distances, and effectively improves significantly to the accuracy of the positioning.
3) In order to verify the performance of the hierarchical fingerprint positioning method provided by the invention, the following simulation is carried out:
to better evaluate the performance of the proposed method, the indoor dimensions are 70 × 120 × 4m at 28GHz millimeter wave channel3The experiment was performed in a propagation environment. The position of the base station (AP) is set to (47.5m,57.5m,4.0 m). A single base station with 256 antennas is used at the transmitter and 32 antennas are configured at the receiver. The distance between adjacent antennas and the radius of each uniform circular array are respectively half wavelength and double wavelength of the millimeter wave carrier frequency. In addition, for the millimeter wave indoor propagation environment, a near free space path loss model (the close-in free space reference path loss model) is adopted, that is:
Figure BDA0002456329910000131
wherein, with dlDenotes the length of the l-th path, γ and
Figure BDA0002456329910000132
path loss factor and shadow fading, FSPL (f, d), respectively0)＝20log10(4 π f/c) is the free space path loss.
The performance of the experiment is mainly evaluated by three aspects of Root Mean Square Error Distance (RMSED), median Error value and Cumulative Distribution Function (CDF), wherein RMSED is the Mean Square Error between the actual position and the estimated position:
Figure BDA0002456329910000133
firstly, 400 test points are selected for testing, and 120 test points are randomly selected from the test points to draw a positioning effect graph. Referring to fig. 5, a hexagram is marked as the actual position and a circle is marked as the estimated position. As is evident from the figure, the estimated positions of all test points are very close to, or even coincide with, the actual positions. And further shows that the proposed hierarchical positioning system has good positioning performance.
Then, in order to further verify the validity of the proposed scheme, the positioning performances of the different positioning methods are compared. As shown in table 1, the second column is the average error distance, the third column is the median error value, and the last column is the maximum error distance. It can be seen from the table that the proposed positioning method can finally achieve a positioning error of 0.4738 m. The median error is 0.2619m, and the maximum error distance is also significantly smaller than other comparison methods.
TABLE 1 comparison of results of different positioning methods
Figure BDA0002456329910000134
The performance of the present invention is compared to other positioning schemes, see figure 6. As can be seen from the figure, the positioning system is obviously superior to other algorithms, the positioning error of about 77 percent of the test points is below 0.5m, and the positioning error of 88 percent of the test points is below 1 m. And test points with the positioning error of other algorithms within 1m are only 55%, 60% and 64% respectively.
The CDF comparison of the proposed algorithm with different refinement label numbers ( n 7,9,11,13) is shown in fig. 7. It can be seen from the figure that as the number of refinement labels in the sliding window increases, the CDF value gradually increases, indicating that the positioning effect is better and better. Especially when n is 13, the test points with an error of less than 1m account for about 97%. Furthermore, as n increases, the maximum error distance decreases by about 0.5m at a time. This is because the fixed interval of adjacent thinning labels is set to 0.5m, and when n is changed from 7 to 9, the side length of the sliding window increases by 1m, but in reality the sliding window extends only 0.5m in four directions, so the maximum error decreases by approximately 0.5 m. The above analysis shows that the error level can be controlled within a certain range by reasonably setting the number of the refined labels.
Table 2 shows the comparison of the positioning effect under different thinning label numbers, and it can be seen that the average positioning error, the median error value, and the maximum error are all gradually reduced with the increase of n. When n is 13, the mean error value is further reduced to 0.3101 m. This is because, as the number of the thinning labels increases, the size of the sliding window increases, and thus the positioning performance becomes better.
TABLE 2 comparison of positioning effects under different refined label numbers
Refining tag count Mean(m) Median(m) Max(m)
n＝7 0.7006 0.3653 3.9613
n＝9 0.4738 0.2619 3.4169
n＝11 0.3729 0.2506 2.9777
n＝13 0.3101 0.2315 2.4903
In the proposed SMWKNN algorithm, before calculating the average signal distance, candidate labels are screened by setting a threshold value, and label points corresponding to the singular signal distance are deleted. Therefore, in evaluating the performance, the influence of the threshold coefficient α on the positioning error is compared, and in order to select the threshold coefficient with the best general applicability, experiments are performed under the condition of different numbers of refinement labels, see fig. 8. Obviously, when α is 2, the localization performance is optimal for all refinement tag numbers set experimentally, so a threshold coefficient of 2 is chosen.
Therefore, in summary, the indoor fingerprint positioning method based on the hierarchical model provided by the invention can effectively reduce the positioning complexity, improve the positioning precision and reduce the positioning cost.
While the invention has been described in further detail with reference to specific preferred embodiments thereof, it will be understood by those skilled in the art that various changes in form and details may be made therein without departing from the spirit and scope of the invention as defined by the appended claims.

Claims (4)

1. An indoor fingerprint positioning method for a millimeter wave large-scale MIMO system is characterized in that a receiving end acquires a CSI multipath signal of the millimeter wave large-scale MIMO system, which can be separated in time domain, and performs data collection and preprocessing on the CSI multipath signal; in the coarse-grained positioning scheme, AOA information estimated by a multipath CSI signal is used as fingerprint characteristics, a deep learning method is combined, characteristic quantities are extracted in a self-adaptive mode, and a fingerprint database is established to achieve coarse-grained positioning; in order to overcome the high complexity caused by multiple antennas, in a fine-grained positioning scheme, positioning is carried out by taking averaged multipath CSI (channel state information) as fingerprint characteristics; meanwhile, considering the problem that the traditional WKNN algorithm depends on a preselected fixed K value to a great extent, a dynamic weighted K nearest neighbor algorithm is adopted to realize better positioning accuracy; the method comprises the following specific steps:
firstly, in a coarse-grained positioning stage, estimating acquired multipath CSI fingerprint data by using a classical MUSIC algorithm to obtain multipath AOA information as fingerprint characteristics, extracting the fingerprint characteristics in a self-adaptive manner through a multilayer convolutional neural network to construct a fingerprint database, and then using a position label with the maximum output probability as a final estimation position in the coarse-grained positioning stage through a softmax classifier;
the data collection and pretreatment in the coarse-grained positioning stage specifically comprise the following steps:
step 1: uniformly dividing cells into N1A block having its geometric center as a classification position label;
step 2: in the coarse-grained location stage, the selected coarse-grained location labels are sparsely distributed and are few in number, so that the N divided is1Multipath CSI information respectively collected in the blocks is estimated through a classical MUSIC estimation algorithm to obtain multipath AOA information of all coarse-grained position label points, and the multipath AOA information is used as fingerprint characteristics;
and step 3: adding the sample characteristics into a label, wherein the sample in the coarse-grained positioning stage is represented as follows:
Figure FDA0003037190100000011
wherein phi is0AOA, phi of direct path1Scattering for the first stripeThe AOA of the path is determined,
Figure FDA0003037190100000012
is the NthrayThe AOA and T of the scattering paths are matrix transposes;
the task of the coarse-grained positioning stage is to train parameters of a deep convolutional neural network and a regression classifier network according to the acquired labeled training data, and the training aim is to minimize the mean square error output by a training label and the network;
the off-line stage training process of the coarse-grained location link is as follows:
for a deep learning network, a deep convolutional neural network is adopted, a ReLU function is adopted as an excitation function of each layer of nodes, after training data are input into the network, the output of each layer is obtained according to the excitation function and is used as the input of the next layer, and finally, the network output is obtained through layer-by-layer forward propagation; constructing a penalty function according to a minimum mean square error principle, updating and iterating by using a random gradient descent algorithm to obtain a final training parameter, and storing a trained weight W, b as a part of a fingerprint library;
then, taking neural network training output data as input of a softmax classifier, then dividing the neural network training output data into C classes, taking the probability that the input data belongs to each class as output of the classifier, constructing a penalty function according to a minimum mean square error principle, updating and iterating by using a random gradient descent algorithm to obtain final training parameters, and forming W, b and theta together into a fingerprint library, wherein the theta is a classifier parameter;
the output of the classifier is as follows:
Figure FDA0003037190100000021
wherein the content of the first and second substances,
Figure FDA0003037190100000022
is a C1 matrix, each term is represented in
Figure FDA0003037190100000023
In the given case of the situation where,
Figure FDA0003037190100000024
the probability of belonging to each of the classes,
Figure FDA0003037190100000025
the method comprises the following steps of (1) inputting training output of a neural network, namely input of a regression classifier, and theta is a parameter of the classifier;
the on-line stage positioning process of the coarse grain positioning link is as follows:
step 1: after receiving CSI information from users at unknown positions, the MUSIC algorithm estimates the received AOA information, namely
Figure FDA0003037190100000026
The probability that the unknown data belong to each to-be-determined position is obtained through forward propagation of a machine learning network and classification of a regression classifier;
step 2: using a probability method to take the position label with the maximum output probability as the final position estimation of coarse-grained positioning;
then, in a fine-grained positioning stage, on the basis of the coarse-grained positioning labels obtained through estimation, fine-grained labels are generated by taking the coarse-grained positioning labels as a core, and considering that the more the number of the fine-grained labels is, the finer the fine-grained positioning result is, the averaged CSI amplitude information is selected as the fingerprint characteristic of the fine-grained positioning stage, and a dynamic weighted K nearest neighbor algorithm based on space mapping is provided for fine-grained positioning;
finally, in order to improve the positioning accuracy, at the fine-grained stage, the following facts are considered: the same CSI difference may correspond to different geometric distances, that is, for CSI differences of different magnitudes, the signal distances represented by the CSI difference may correspond to different geometric distances; therefore, the relationship between the signal distance space and the geometric distance space is established by training the ELM to realize the spatial mapping and prevent the influence of the distance mismatch from causing the reduction of the positioning precision.
2. The indoor fingerprint positioning method for the millimeter wave massive MIMO system as claimed in claim 1, wherein the data collection and preprocessing at the fine-grained positioning stage comprises the following steps:
step 1: based on the coarse-grained positioning result, the coarse-grained positioning result is used as a core, the coarse-grained positioning result is expanded outwards at equal intervals to generate a fine-grained label called an expanded sliding window, and N is generated in the sliding window according to the method2Fine-grained location tags;
step 2: for a fine-grained positioning stage, in order to further reduce data dimension and operation complexity, sampling CSI data of each path, then averaging amplitude channel matrixes on different antennas, and grouping and numbering the amplitude data according to the corresponding relation between known CSI information and user positions;
and step 3: adding the sample characteristics into a label, wherein the sample in the fine-grained positioning stage is represented as follows after the label is added:
Figure FDA0003037190100000031
wherein the content of the first and second substances,
Figure FDA0003037190100000032
N2the position label number of the fine-grained positioning stage.
3. The indoor fingerprint positioning method for the millimeter wave massive MIMO system as claimed in claim 2, wherein the task of the fine-grained positioning stage is to extend outward to form a sliding window full of the refined virtual tags with the estimated position as the center on the basis of the coarse-grained positioning; considering that phase shift exists in CSI, and in order to overcome high complexity brought by multiple antennas, averaged multipath CSI amplitude information is used as a position label, and a space mapping-based adaptive dynamic weighting K nearest neighbor algorithm comprises the following algorithm steps:
step 1: calculating the signal distance between all sample points of the fingerprint library:
Figure FDA0003037190100000033
step 2: deleting singular distance, and setting threshold T ═ alpha × D1Keeping the signal distance which is not more than the threshold value T in the formula, and sequentially marking as D from small to large1,...,DSS-1, 2, …, S represents the number of retained neighboring reference points;
and step 3: calculate the mean distance difference of the retention points from each other:
Figure FDA0003037190100000041
wherein, Δ dj,sRepresents DjAnd DsThe distance difference of (a);
and 4, step 4: estimating the final position, the dynamic weight value is expressed as:
Figure FDA0003037190100000042
wherein Δ D ═ DK-D1(ii) a In particular, when DK＝D1Time, omegaj1 is ═ 1; defining:
Figure FDA0003037190100000043
the specific form of the final position estimate is then as follows:
Figure FDA0003037190100000044
wherein (x)j,yj) The coordinates of the jth remaining location tag are represented.
4. The indoor fingerprint positioning method for mmwave massive MIMO system as claimed in claim 3, wherein in the proposed dynamic weighted K-nearest neighbor algorithm step (1) based on spatial mapping, when calculating the signal distance, considering the fact that the signal distance and the physical distance do not match at different CSI amplitude levels, the spatial mapping method based on extreme learning machine is adopted, and this problem is described as follows:
Figure FDA0003037190100000045
wherein S isDAnd SCRespectively signal space distance and geometric space distance.
CN202010307606.5A 2020-04-17 2020-04-17 Indoor fingerprint positioning method for millimeter wave large-scale MIMO system Active CN111479231B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202010307606.5A CN111479231B (en) 2020-04-17 2020-04-17 Indoor fingerprint positioning method for millimeter wave large-scale MIMO system

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202010307606.5A CN111479231B (en) 2020-04-17 2020-04-17 Indoor fingerprint positioning method for millimeter wave large-scale MIMO system

Publications (2)

Publication Number Publication Date
CN111479231A CN111479231A (en) 2020-07-31
CN111479231B true CN111479231B (en) 2021-07-13

Family

ID=71753994

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202010307606.5A Active CN111479231B (en) 2020-04-17 2020-04-17 Indoor fingerprint positioning method for millimeter wave large-scale MIMO system

Country Status (1)

Country Link
CN (1) CN111479231B (en)

Families Citing this family (12)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112040394B (en) * 2020-08-05 2022-06-10 深圳市微能信息科技有限公司 Bluetooth positioning method and system based on AI deep learning algorithm
CN111918388A (en) * 2020-08-17 2020-11-10 南京邮电大学 CSI fingerprint passive positioning method based on depth separable convolution
CN112040400B (en) * 2020-08-25 2021-08-13 西安交通大学 Single-site indoor fingerprint positioning method based on MIMO-CSI, storage medium and equipment
CN112712557B (en) * 2020-12-17 2023-04-07 上海交通大学 Super-resolution CIR indoor fingerprint positioning method based on convolutional neural network
CN112770390B (en) * 2020-12-30 2022-07-12 东南大学 Robust fingerprint extraction method based on large-scale MIMO single station system
CN112964960B (en) * 2021-02-04 2024-04-12 国网上海市电力公司 Power grid fault diagnosis method based on multi-source data fusion of scheduling fault fingerprint library
CN112995892B (en) * 2021-02-08 2022-11-29 东南大学 Large-scale MIMO fingerprint positioning method based on complex neural network
US11885899B2 (en) 2021-05-07 2024-01-30 Qualcomm Incorporated Generating indoor maps based on radio frequency sensing
CN113676857B (en) * 2021-08-19 2023-05-26 重庆邮电大学 Channel state information feature extraction method for indoor Wi-Fi positioning
CN117062002B (en) * 2023-08-01 2024-04-30 哈尔滨工程大学 5G NR indoor positioning method and system based on lightweight TRANSFORMER
CN116828596B (en) * 2023-08-28 2023-11-10 四川思凌科微电子有限公司 High-precision positioning method using RSSI
CN117576225B (en) * 2023-11-28 2024-05-31 吉林化工学院 Indoor visible light positioning method and system based on received signal strength ratio

Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106332277A (en) * 2016-09-05 2017-01-11 中南大学 Indoor positioning method based on distribution of channel state information
CN107241700A (en) * 2017-04-23 2017-10-10 西安电子科技大学 Indoor orientation method based on CSI Space-frequency characters and reference point locations clustering algorithm
CN107517446A (en) * 2017-09-28 2017-12-26 知谷（上海）网络科技有限公司 Indoor orientation method and device based on Wi Fi focuses
CN107769828A (en) * 2017-09-27 2018-03-06 中国电子科技集团公司第五十四研究所 A kind of the CSI MIMO indoor orientation methods and system of the extraction of feature based value
CN108696932A (en) * 2018-04-09 2018-10-23 西安交通大学 It is a kind of using CSI multipaths and the outdoor fingerprint positioning method of machine learning
CN110381436A (en) * 2019-06-25 2019-10-25 东南大学 A kind of quick fingerprint positioning method based on the mono- station system of extensive MIMO
US10531421B1 (en) * 2018-12-06 2020-01-07 At&T Intellectual Property I, L.P. Systems and methods for locating user equipment during disasters

Family Cites Families (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106685590B (en) * 2016-12-08 2019-10-29 浙江工业大学 Indoor human body orientation identification method based on channel state information and KNN
CN108304826A (en) * 2018-03-01 2018-07-20 河海大学 Facial expression recognizing method based on convolutional neural networks

Patent Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106332277A (en) * 2016-09-05 2017-01-11 中南大学 Indoor positioning method based on distribution of channel state information
CN107241700A (en) * 2017-04-23 2017-10-10 西安电子科技大学 Indoor orientation method based on CSI Space-frequency characters and reference point locations clustering algorithm
CN107769828A (en) * 2017-09-27 2018-03-06 中国电子科技集团公司第五十四研究所 A kind of the CSI MIMO indoor orientation methods and system of the extraction of feature based value
CN107517446A (en) * 2017-09-28 2017-12-26 知谷（上海）网络科技有限公司 Indoor orientation method and device based on Wi Fi focuses
CN108696932A (en) * 2018-04-09 2018-10-23 西安交通大学 It is a kind of using CSI multipaths and the outdoor fingerprint positioning method of machine learning
US10531421B1 (en) * 2018-12-06 2020-01-07 At&T Intellectual Property I, L.P. Systems and methods for locating user equipment during disasters
CN110381436A (en) * 2019-06-25 2019-10-25 东南大学 A kind of quick fingerprint positioning method based on the mono- station system of extensive MIMO

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
Xuyu Wang.CSI Phase Fingerprinting for Indoor Localization With a Deep Learning Approach.<IEEE>.2016, *
王亚涛.基于位置指纹的WiFi室内定位研究.《中国优秀硕士学位论文全文数据库（电子期刊） 信息科技辑》.2019, *

Also Published As

Publication number Publication date
CN111479231A (en) 2020-07-31

Similar Documents

Publication Publication Date Title
CN111479231B (en) Indoor fingerprint positioning method for millimeter wave large-scale MIMO system
Sun et al. Fingerprint-based localization for massive MIMO-OFDM system with deep convolutional neural networks
CN108696932B (en) Outdoor fingerprint positioning method using CSI multipath and machine learning
CN106851571B (en) Decision tree-based rapid KNN indoor WiFi positioning method
CN107071743B (en) Rapid KNN indoor WiFi positioning method based on random forest
CN110381436A (en) A kind of quick fingerprint positioning method based on the mono- station system of extensive MIMO
CN114189809B (en) Indoor positioning method based on convolutional neural network and high-dimensional 5G observation feature
CN110049549A (en) More fusion indoor orientation methods and its system based on WiFi fingerprint
CN107727095A (en) 3D indoor orientation methods based on spectral clustering and weighting reverse transmittance nerve network
CN112616184B (en) Mobile equipment position estimation method based on multi-base station channel state information fusion
CN109640262B (en) Positioning method, system, equipment and storage medium based on mixed fingerprints
CN108924756A (en) Indoor orientation method based on WiFi double frequency-band
CN110536245A (en) A kind of indoor wireless positioning method and system based on deep learning
CN109922427A (en) Utilize the intelligent radio positioning system and method for large scale array antenna
CN113207089A (en) Position fingerprint positioning method based on CSI and crowdsourcing migration self-calibration updating
CN111918388A (en) CSI fingerprint passive positioning method based on depth separable convolution
CN109379711A (en) A kind of localization method
CN113596989A (en) Indoor positioning method and system for intelligent workshop
CN114386321A (en) AOA and TOF joint estimation method and device for indoor positioning and storage medium
Kia et al. A CNN approach for 5G mm wave positioning using beamformed CSI measurements
Tan et al. UAV localization with multipath fingerprints and machine learning in urban NLOS scenario
CN111523571B (en) Semi-supervised learning positioning method for distributed large-scale multi-antenna system
CN114269014B (en) Large-scale MIMO dynamic environment fingerprint positioning method based on domain adaptive network
Nie et al. Joint access point fuzzy rough set reduction and multisource information fusion for indoor Wi-Fi positioning
Chen et al. Multipath-based CSI fingerprinting localization with a machine learning approach

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant