CN107027148B - Radio Map classification positioning method based on UE speed - Google Patents

Radio Map classification positioning method based on UE speed Download PDF

Info

Publication number
CN107027148B
CN107027148B CN201710240569.9A CN201710240569A CN107027148B CN 107027148 B CN107027148 B CN 107027148B CN 201710240569 A CN201710240569 A CN 201710240569A CN 107027148 B CN107027148 B CN 107027148B
Authority
CN
China
Prior art keywords
rsrp
point
cqt
sampling
sampling points
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Expired - Fee Related
Application number
CN201710240569.9A
Other languages
Chinese (zh)
Other versions
CN107027148A (en
Inventor
马琳
金宁迪
徐玉滨
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.)
Harbin Institute of Technology
Original Assignee
Harbin Institute of Technology
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 Harbin Institute of Technology filed Critical Harbin Institute of Technology
Priority to CN201710240569.9A priority Critical patent/CN107027148B/en
Publication of CN107027148A publication Critical patent/CN107027148A/en
Application granted granted Critical
Publication of CN107027148B publication Critical patent/CN107027148B/en
Expired - Fee Related legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04W—WIRELESS COMMUNICATION NETWORKS
    • H04W24/00—Supervisory, monitoring or testing arrangements
    • H04W24/10—Scheduling measurement reports ; Arrangements for measurement reports
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
    • G06F18/2413—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on distances to training or reference patterns
    • G06F18/24147—Distances to closest patterns, e.g. nearest neighbour classification
    • 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)
  • Theoretical Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • Signal Processing (AREA)
  • Computer Networks & Wireless Communication (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Evolutionary Biology (AREA)
  • Evolutionary Computation (AREA)
  • Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Artificial Intelligence (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Mobile Radio Communication Systems (AREA)

Abstract

The invention discloses a Radio Map classification positioning method based on UE (user equipment) speed, and relates to a Radio Map classification positioning method based on UE speed. The invention aims to solve the problems that in the background of mass user LTE positioning, the positioning accuracy is low due to the fact that a signal fingerprint is short when UE moves at a low speed and the fingerprint mismatch phenomenon is serious when the UE moves at a high speed. The specific process is as follows: firstly, acquiring DT/CQT/MDT sampling points; secondly, extracting RSRP symbiotic vectors from the acquired DT/CQT/MDT sampling points; thirdly, designing a strong classification function based on an Adaboost classification algorithm according to the second step; fourthly, the method comprises the following steps: constructing an offline Radio Map by using a strong classification function obtained by training in the step three; and fifthly, carrying out online positioning by using the strong classification function obtained by training in the third step and the offline Radio Map obtained in the fourth step. The invention is used in the field of positioning.

Description

Radio Map classification positioning method based on UE speed
Technical Field
The invention relates to a Radio Map classification positioning method based on UE speed.
Background
In recent years, with rapid development of communication technology and information technology and rapid popularization of various smart devices, location-based services have received much attention. 80% of the information in daily life is related to location, enough to see the importance of location. On the other hand, the concept of 'all things interconnection' is revolutionarily proposed by 5G, and for the internet of things and the internet of vehicles, the most basic characteristic of the object is 'moving', and it is very important to find the motion of the object and to know the position of the object in real time. Therefore, high precision positioning, especially high precision outdoor positioning, is an important focus of increasing researchers' attention. GNSS (global navigation Satellite system) is a most widely used outdoor positioning technology, and although IT can achieve high positioning accuracy, IT is sensitive to obstruction, weather change, and the like, on the other hand, obtaining a positioning result of a user's GNSS requires permission of a user terminal, and is easily available for manufacturers having terminal software products in the IT industry, and for manufacturers providing basic services such as operators, IT is not easy to obtain this part of data, so that a mobile terminal positioning method without user participation needs to be researched to replace GNSS. With the popularization of the fourth generation mobile communication technology, the full coverage of LTE has been realized in urban areas and most of rural areas, so the positioning system based on LTE is gradually a research hotspot for outdoor positioning.
The LTE-based positioning system uses a wide range of technologies, which mainly include a location fingerprint method, a toa (time of arrival) method, an aoa (angle of attach) method, a tdoa (time Difference of arrival) method, and the like. The position fingerprint method is used as a supervised positioning method, a position fingerprint Map (Radio Map) of the method is established based on actually acquired data, and the characteristics of an actual signal propagation environment, such as barrier shielding, non-line-of-sight propagation and the like, are completely reserved, namely the positioning method based on the position fingerprint is more suitable for an outdoor complex environment, so the positioning method based on the position fingerprint is adopted. The traditional location method based on location fingerprint carries out matching location by RSRP (reference Signal RecevingPower) characteristics and a database. Location based positioning comprises the following two phases: establishing an off-line stage of a radio map database and an on-line fingerprint matching and positioning stage. Establishing a position fingerprint database (Radio Map) containing Reference Point (RP) positions and RSRP fingerprints at an off-line stage; in the on-line stage, generally, the euclidean distance of the signal space between the reference point RSRP in the Radio Map and the on-line RSRP is calculated, and a positioning result is obtained by using an NN (Nearest Neighbor) or k-NN (k-Nearest Neighbor) algorithm.
The position fingerprint positioning of mass users in the LTE system is different from the position fingerprint positioning in the traditional meaning, and has the characteristics of large positioning area, more positioning users, wide positioning scene and the like. The invention aims at the problem that the positioning accuracy is influenced by the special terminal movement speed under the background of LTE mass user positioning, and improves the traditional position fingerprint positioning algorithm.
The signal fingerprints in the MDT reports reported by the UE at different motion speeds are greatly different. In an LTE positioning system based on location fingerprints, in the process of establishing a Radio Map by using an mdt (minimization Drive test) measurement report, it is assumed that signal fingerprint information, location information, and reported time information in the measurement report are strictly synchronized. However, in practical applications, this assumption is difficult to be made, especially in case of high-speed UE movement. Fig. 1 is a diagram illustrating location, fingerprint and time offset of periodically triggered MDT reports.
Three coordinate axes in the figure are respectively position information updating time, signal fingerprint information updating time and MDT measurement report reporting time. It can be seen that for the periodically triggered MDT, the reporting time of the MDT measurement report is strictly equal, and the location and signal fingerprint in the report are not measured at the MDT reporting time, but are the result of the last measurement, which results in time misalignment between different types of information in the MDT measurement report. For the UE moving at low speed, because the change rate of the position and the signal fingerprint is small, the influence of the misalignment phenomenon on the measurement report is not obvious, and for the UE moving at high speed, the misalignment phenomenon will seriously affect the accuracy of the Radio Map.
Two measurement reporting modes for MDT of E-UTRAN are specified in 3GPP TS 32.422: logged MDT and immediatate MDT. For the Logged MDT, when the UE is in the Idle state, the UE performs MDT measurement and storage according to MDT configuration information issued by the network side, and then enters a connection state and reports to the network when the network side requests a measurement report. And the Immediate MDT measures in a UE connection state and immediately reports a measurement report to the network after the measurement is finished. Because the requirement on the real-time performance of positioning is high, an Immediate MDT mode is adopted in an LTE positioning system based on the position fingerprint. There are three triggering modes for the Immediate MDT mode: periodic measurements, a2 event trigger, and a2 event trigger. The a2 event is an intra-frequency system event that indicates when the quality of service is worse than a threshold. The UE triggers an a2 event when the condition is satisfied and leaves the a2 event when the condition is satisfied.
Ms+Hys＜Thresh (1)
Ms-Hys＞Thresh (2)
Wherein Ms is the measurement result of the serving cell, and has a unit of dBm when the Ms represents RSRP and a unit of dB when the Ms represents RSRQ or RS-SINR; hys is a hysteresis parameter for the A2 event, in dB; thresh is the threshold parameter for the a2 event (configured in reportConfigEUTRA), and has the same units as the measurement Ms. When the UE moves at a high speed, due to the influence of Doppler effect, RSRQ and RS _ SINR are poor, and therefore an A2 event is triggered, more same-frequency neighbor cell measurement is started, and the obtained signal fingerprint is long. And under the condition of low-speed movement, an A2 event is not triggered generally, the number of measured adjacent regions is small, and the length of the signal fingerprint is short.
Therefore, it can be concluded that for a user moving at a high speed, the time, location information and signal fingerprint information in the report have serious fingerprint mismatch phenomena, but the signal fingerprint is usually long; the fingerprint mismatch phenomenon of a user moving at a low speed also exists, but the accuracy of the fingerprint is not greatly influenced, and the signal fingerprint is generally short. If the two types of fingerprints can be separately positioned, the fingerprint length advantage of a moving user and the fingerprint precision advantage of a low-speed moving user can be exerted, and therefore the positioning precision is improved. Currently, only manual high-speed dt (drive test) tests and low-speed cqt (call Quality test) tests can confirm the approximate movement situation of the sampling point, while the speed information of the user in the MDT measurement report reported by the ordinary user is unknown. Therefore, it is necessary to divide the user speed based on the information in the MDT measurement report, and further separate the two types of fingerprints to establish Radio Map and positioning.
Disclosure of Invention
The invention aims to solve the problems that in the background of LTE positioning of a large number of users, the positioning accuracy is low due to the fact that a signal fingerprint is short when UE moves at a low speed and the fingerprint mismatch phenomenon is serious when the UE moves at a high speed, and provides a Radio Map classification positioning method based on the UE speed.
A Radio Map classification positioning method based on UE speed comprises the following specific processes:
step one, obtaining DT/CQT/MDT sampling points;
the DT is a drive test; CQT is call quality test, MDT is minimization of drive test;
step two, extracting RSRP symbiotic vectors from DT/CQT/MDT sampling points obtained in the step one;
the RSRP is reference signal received power;
step three, designing a strong classification function based on an Adaboost classification algorithm according to the step two;
step four: constructing an offline Radio Map by using a strong classification function obtained by training in the third step;
the Radio Map is a position fingerprint Map;
and step five, carrying out online positioning by using the strong classification function obtained by training in the step three and the offline Radio Map obtained in the step four.
The invention has the beneficial effects that:
in a UE speed Classification training module, firstly adding a sliding time window to continuous MDT, extracting RSRP symbiotic vectors from RSRP values in each time window, And training an Adaboost Classification model taking a Classification Regression Tree (CART) as a weak classifier. In the UE speed classification module, a sliding time window with the same length is added to the continuous MDT, RSRP symbiotic vectors are extracted from the RSRP value in one time window, and the speed classification corresponding to the current time window is obtained based on an Adaboost model established offline. In the off-line stage, two Radio maps of DT and CQT are established based on the classification result; and in the online stage, matching the corresponding Radio Map based on the classification result and obtaining the classification result by adopting a KNN algorithm. The problem of when UE low-speed motion signal fingerprint is shorter, when UE high-speed motion fingerprint mismatch phenomenon is serious, lead to positioning accuracy low is solved.
Under the background of mass user LTE positioning, the positioning method provided by the invention can distinguish high-speed and low-speed sampling points with high precision, give play to respective fingerprint advantages and effectively improve the positioning precision.
The low-speed detection probability of 99.92 percent and the high-speed detection probability of 93.59 percent in the UE speed classification by adopting the method are obtained by combining the table 2; the low-speed false alarm probability is 0.47%, and the high-speed false alarm probability is 1.17%;
with reference to fig. 3, when the positioning error is 10m, the cumulative probability distribution of the present invention is 0.4, and the cumulative probability distribution of the KNN algorithm of the prior art is 0.38; when the positioning error is 60m, the cumulative probability distribution of the method is 0.87, and the cumulative probability distribution of the KNN algorithm in the prior art is 0.82; when the positioning error is 100m, the cumulative probability distribution of the invention is 0.89, and the cumulative probability distribution of the KNN algorithm in the prior art is 0.84.
Drawings
FIG. 1 is a diagram illustrating position, fingerprint and time offset of MDT report triggered periodically, where MDT is Minimization of Drive Test (MDT), Δ t is time interval of MDT report, t is time axis1、t2、t3、t4、t5Reporting time for MDT report with equal interval;
FIG. 2 is a block diagram of an LTE fingerprint positioning system of the present invention, wherein LTE is a fourth generation communication technology (Long TermEsolution);
fig. 3 is a schematic diagram of comparison of CDF curves of simulation positioning accuracy of the present invention, where CDF is cumulative probability distribution, and KNN algorithm is K nearest neighbor classification algorithm.
Detailed Description
The first embodiment is as follows: the Radio Map classified positioning method based on the UE speed of the embodiment comprises the following specific processes:
step one, acquiring DT, CQT and MDT sampling points;
the DT is a drive test; CQT is call quality test, MDT is minimization of drive test;
step two, extracting RSRP symbiotic vectors from DT/CQT/MDT sampling points obtained in the step one;
the RSRP is reference signal received power;
step three, designing a strong classification function based on an Adaboost classification algorithm according to the step two;
step four: constructing an offline Radio Map by using a strong classification function obtained by training in the third step;
the Radio Map is a position fingerprint Map;
and step five, carrying out online positioning by using the strong classification function obtained by training in the step three and the offline Radio Map obtained in the step four.
The second embodiment is as follows: the first difference between the present embodiment and the specific embodiment is: acquiring DT/CQT/MDT sampling points in the first step; the specific process is as follows:
the method comprises the following steps: acquiring a DT/CQT/MDT sampling point data packet from LTE network optimization data, marking the DT sampling point as high-speed moving UE, marking the CQT sampling point as low-speed moving UE or static UE, and enabling the MDT sampling point to have no moving label; the UE is User Equipment (User Equipment);
the UE speed is higher than or equal to 30km/h and is high, and the UE speed is lower than 30km/h and is low;
the first step is: the latitude and longitude of DT, CQT and MDT sampling points with positions
Figure GDA0002247844090000041
Carrying out position gridding, fixing the sampling point position at a grid node with the nearest distance, and setting the grid interval to be 1 meter so as to enable the effective digits of longitude and latitude to be matched with the positioning precision of a GPS; the longitude difference and latitude difference corresponding to 1 meter are calculated by formula (3), the difference is related to the user position, taking Harland city (126 degree E,45 degree N) as an example, the longitude difference is about 1.3X 10-6Degree, difference in latitude of about 0.9X 10-6°；
Figure GDA0002247844090000051
Where λ is the longitude of the sampling point UE,
Figure GDA0002247844090000052
for sampling point UE latitude, R is earth radius (6371 km in this patent), lambdaAThe longitude of the point a is taken as the point a,
Figure GDA0002247844090000053
is latitude, λ, at point ABIs the longitude of the point B, and the point B,
Figure GDA0002247844090000054
latitude of point B, LABA, B, the distance between the two points is 1 m;
step one is three: extracting IMSI (International Mobile Subscriber identity) of the UE from the DT/CQT/MDT sampling point data packet acquired in the step one to uniquely identify the UE;
the IMSI is an international mobile subscriber identity;
step one is: extracting a current test sampling point measurement time stamp (Timestamp) from the DT/CQT/MDT sampling point data packet acquired in the step one by one;
step one and five: extracting global cell identification code of service cell from DT/CQT/MDT sampling point data packet obtained from step one by one, and recording the global cell identification code as CGI0(ii) a Extracting the RSRP measurement result of the serving cell and marking as RSRP0；
Step one is six: extracting PCI (physical Cell identifier) of the neighboring Cell from DT/CQT/MDT sampling point data packets acquired in the step one by one, wherein the neighboring Cell PCI (physical Cell identifier) at most comprises 6 neighboring cells and is marked as the PCI1～PCI6(ii) a Obtaining RSRP measurement result corresponding to PCI of adjacent cell and marking as RSRP1～RSRP6；RSRP1Receiving power for reference signals of adjacent cells 1; RSRP6Receiving power of reference signals of adjacent cells 6; PCI is physical cell identification; RSRP is reference signal received power;
step one and step seven: comparing the PCI of the neighboring cell with the I-parameter, and when the PCI of the neighboring cell is completely consistent with the I-parameter, obtaining a CGI of the neighboring cell, and recording the CGI as the CGI1～CGI6；CGI1Is a global cell identity, CGI, of a neighboring cell 16Is a global cell identification code of a neighboring cell 6; CGI is a global cell identity of a neighboring cell;
the working parameters comprise base station positions, physical cell identifiers, carrier frequency points and global cell identifiers, and are provided by operators, and the LTE network optimization personnel update the working parameters every day according to network optimization projects.
Step one eight: after the steps one-to-one-seven, all information required for position fingerprint positioning is acquired, see table 1 below, and the steps one-to-one-seven are repeated until all DT/CQT/MDT sampling points are acquired:
TABLE 1 sampling point information and corresponding table of meaning
Figure GDA0002247844090000055
Figure GDA0002247844090000061
Other steps and parameters are the same as those in the first embodiment.
The second embodiment of the invention aims at the design of a TDD-LTE fingerprint positioning system and an FDD-LTE fingerprint positioning system, and the data acquisition mode in the second embodiment can be popularized and applied to various mobile communication network fingerprint positioning systems such as a GSM fingerprint positioning system, a TD-SCDMA fingerprint positioning system, a WCDMA fingerprint positioning system, a CDMA2000 fingerprint positioning system and the like.
The third concrete implementation mode: the present embodiment differs from the first or second embodiment in that: extracting RSRP symbiotic vectors from DT/CQT/MDT sampling points obtained in the step one in the step two; the specific process is as follows:
step two, firstly: screening sampling points of the same IMSI from the DT/CQT/MDT sampling points obtained in the step one, namely sampling points reported from the same terminal; arranging the sampling points of the same IMSI in an ascending order according to the Timestamp; adding a sliding time window on the sampling point sequence of the same IMSI, wherein the window length is 60s, and the sliding is carried out at an interval of 12 s;
the IMSI is an international mobile subscriber identity;
the Timestamp is the current test sampling point measurement Timestamp;
step two: selecting from a plurality of main areas or adjacent areas detected by the UE in a time windowDetecting the cell c with the highest proportion, and extracting the RSRP vector R of the cell cc，
Figure GDA0002247844090000062
n is the detected times of the cell c in the current time window, and the value is a positive integer;
Figure GDA0002247844090000063
is a real number;
step two and step three: r is obtainedcMinimum value r of elements in vectorcminMaximum value of the element rcmaxThen the span of RSRP is m-rcmax-rcmin+ 1; to RcThe following treatments were carried out:
R′c＝Rc-rcmin+1 (4)
wherein r iscmin＝[rcminrcmin… rcmin]n×1，1＝[1 1 … 1]n×1R 'thus obtained'cWherein the elements are positive integers and the minimum value is 1; r'cIs an intermediate variable;
step two, four: initializing RSRP co-occurrence matrix fetching
Figure GDA0002247844090000064
R′cAny one element of r'ciAnd the next time element r'ci+1To obtain a signal intensity variation pair (r'ci,r′ci+1) Let GMMiddle r'ciLine r'ci+1Column element + 1; r'cAll signal strength change pairs taken over, then GMThe construction is completed; m is the span of RSRP, and the value is a positive integer;
step two and step five: r'cIn total, n-1 signal intensity variation pairs, pair GMCarrying out normalization processing to obtain a normalized RSRP co-occurrence matrix
Figure GDA0002247844090000071
Figure GDA0002247844090000072
Step two, step six: co-occurrence matrix P of RSRPMSumming elements in the main diagonal direction, enabling the length of a new vector obtained by summing to be 11, and if the length of the vector obtained by summing is smaller than 11, supplementing 0 with the same number at two ends of the vector; if the length of the vector obtained by summation is larger than 11, truncating 0s with the same number at the two ends of the vector to obtain RSRP symbiotic vector with equal length
Figure GDA0002247844090000073
As shown in fig. 2, the LTE fingerprint positioning system based on Radio Map classification of UE speed mainly includes a UE speed classification training module and a UE speed classification module.
Other steps and parameters are the same as those in the first or second embodiment.
The fourth concrete implementation mode: the difference between this embodiment mode and one of the first to third embodiment modes is: designing a strong classification function based on an Adaboost classification algorithm according to the step two in the step three; the specific process is as follows:
step three, firstly: extracting RSRP symbiotic vectors from all DT/CQT sampling points acquired in the step one according to the method in the step two;
step three: initializing DT/CQT sample point weights to D1(i) (ii) 0.5/w (class (i)), where w (class (i)) is the total number of DT/CQT sample points of the tag to which sample point i belongs;
class (i) is the prior class label of sample point i; i is the ith of DT/CQT sampling points;
step three: using CART as a weak classifier; when each weak classifier CART is trained, the prediction output of each weak classifier CART is obtained, and the prediction sequence g is obtainedt(x) The sum of prediction errors, which is calculated by the formula:
Figure GDA0002247844090000074
wherein i satisfies
Figure GDA0002247844090000075
gt(xi) To predict value, yiIs a desired value; phi is a threshold value and takes a value of 0-0.5; dt(i) Weighting the DT/CQT sampling points of the t iteration;
step three and four: according to the predicted sequence gt(x) Is predicted by a prediction error etCalculating the predicted sequence gt(x) Weight of atThe calculation formula is as follows:
Figure GDA0002247844090000076
step three and five: according to the predicted sequence gt(x) Weight of atAnd obtaining the weight of the next round of DT/CQT sampling points, wherein the adjustment formula is as follows:
Figure GDA0002247844090000077
wherein Z istIs a normalization factor so that
Figure GDA0002247844090000081
Dt+1(i) Sample weight for the t +1 th iteration; w is the total number w of DT/CQT sampling points which is w (-1) + w (+1), and the value is a positive integer;
step three and six: after the steps three to three five, Adaboost iteration is completed. Repeating the third step, the third step and the fifth step until the upper limit T of the iteration times is reached to obtain T weak classification functions f (g)t(x),at) Combining them to obtain a strong classification function h (x):
Figure GDA0002247844090000082
in the formula, atWeak classifier weights for the t-th iteration; the upper limit T of the iteration times is set for people, the larger the iteration times is, the more accurate the iteration times is, the larger the calculated amount is, the longer the time consumption is, and the value is 10-1000; f (g)t(x),at) The weak classifier for the t-th iteration.
Other steps and parameters are the same as those in one of the first to third embodiments.
The fifth concrete implementation mode: the difference between this embodiment and one of the first to fourth embodiments is: in the fourth step, an offline Radio Map is constructed by using a strong classification function obtained by training in the third step; the specific process is as follows:
step four, firstly: classifying the motion states of the sampling points in the off-line stage by using a strong classification function obtained by training in the third step;
the sampling points in the off-line stage are sampling points with positions in all the MDT sampling points acquired in the step one;
the length of the sliding time window is 60s, the interval is 12s, then each offline stage sampling point can obtain 5 classification results, the state label of the offline stage sampling point moving at a low speed is set as-1, the state label of the offline stage sampling point moving at a high speed is set as +1, and then the state label of each offline stage sampling point is expressed as:
Figure GDA0002247844090000083
hi(x) Predicting a result for the Adaboost strong classifier;
when the state label of the sampling point in the off-line stage moving at low speed is-1, marking the sampling point as a CQT sampling point; when the status label of the off-line stage sampling point moving at high speed is +1, marking as the sampling point of DT;
step four and step two: merging the sampling points marked as DT to construct DT offline Radio Map;
merging
Figure GDA0002247844090000084
Obtaining an RP by the same sampling points marked as DT, and averaging the RSRP measured by each sampling point marked as DT for the same CGI; repeating the steps until all the sampling points marked as DT are combined, and obtaining a Radio Map, wherein the Radio Map comprises three parts of information of position, cell number and signal intensity; assuming j RPs, the number of APs corresponding to each RP may be different because the number of neighboring cells in the measurement report is not fixed; lambda [ alpha ]Is the sampling point UE longitude;
Figure GDA0002247844090000091
the latitude of the UE is taken as a sampling point; j is a positive integer, RP is a reference point, and AP is an access point;
step four and step three: combining the sampling points marked as CQT to construct CQT off-line Radio Map,
merging
Figure GDA0002247844090000092
Obtaining an RP by the same sampling points marked as CQTs, and averaging the RSRPs measured by the sampling points marked as the CQTs for the same CGI; repeating the steps until all the sampling points marked as CQTs are combined, and obtaining a Radio Map, wherein the Radio Map comprises three parts of information of positions, cell numbers and signal intensity; if d RPs are provided, the number of APs corresponding to each RP may be different because the number of neighboring cells in the measurement report is not fixed; λ is the sampling point UE longitude;
Figure GDA0002247844090000093
the latitude of the UE is taken as a sampling point; d is a positive integer;
and the DT offline Radio Map and the CQT offline Radio Map form an offline Radio Map.
Other steps and parameters are the same as in one of the first to fourth embodiments.
The sixth specific implementation mode: the difference between this embodiment and one of the first to fifth embodiments is: in the fifth step, the strong classification function obtained by training in the third step and the offline Radio Map obtained in the fourth step are utilized for online positioning; the specific process is as follows:
step five, first: classifying the motion states of the sampling points at the online stage by using a strong classification function obtained by training in the third step; the method is the same as the fourth step.
The online stage sampling points are sampling points without positions in all the MDT sampling points acquired in the step one;
the length of the sliding time window is 60s, the interval is 12s, each online stage sampling point can obtain 5 classification results, the state label of the online stage sampling point moving at a low speed is set as-1, the state label of the online stage sampling point moving at a high speed is set as +1, and the state label of each online stage sampling point is expressed as:
Figure GDA0002247844090000094
in the formula, hi(x) Predicting a result for the Adaboost strong classifier;
when the state label of the sampling point in the online stage moving at low speed is-1, marking as the sampling point of CQT; when label is +1, marking as DT sampling point;
step five two: matching the corresponding off-line Radio Map in the off-line Radio Map obtained in the step four according to the on-line stage sampling point motion state classification result obtained in the step five; screening with inclusion of CGI0～CGInCalculating the Euclidean distance of the signal space of the reference point in the online RSRP and the offline Radio Map; suppose that the pth reference point contains all online CGIs simultaneously, and the corresponding CGI in the reference point is the CGIp0～CGIpnThe corresponding RSRP is RSRPp0～RSRPpnAnd the Euclidean distance d of the signal space of the p-th reference point in the RSRP and the Radio Map of the online sampling pointpComprises the following steps:
Figure GDA0002247844090000101
the reference point is a point obtained by combining sampling points with positions in all the MDT sampling points obtained in the step one; p is positive integer;
in the formula | · | non-conducting phosphor2Is the L2 norm of the matrix; RSRPpRSRP measurements for reference point p for 1 reference point; RSRP is reference signal received power; CGI0A global cell identity for a serving cell for 1 reference point; CGInA global cell identity for neighbor cell n for 1 reference point; n is 1-6; CGIp0A global cell identity of a serving cell that is reference point p; CGIpnA global cell identity of a neighbor cell n being a reference point p; RSRPp0RSRP measurement results for a serving cell that is reference point p; RSRPpnRSRP measurement results of neighbor cells n that are reference points p; p is a positive integer;
step five and step three: selection of Euclidean distance dpA minimum of k reference points, the test point location is estimated by:
Figure GDA0002247844090000102
in the formula (I), the compound is shown in the specification,
Figure GDA0002247844090000103
estimating a longitude for the online reference point;
Figure GDA0002247844090000104
estimating a latitude for the online reference point; k is 3-5; lambda [ alpha ]lLongitude for the selected ith reference point;
Figure GDA0002247844090000105
the latitude of the selected first reference point; dpThe Euclidean distance of the signal space between the RSRP of the online sampling point and the p reference point.
Other steps and parameters are the same as in one of the first to fourth embodiments.
The following examples were used to demonstrate the beneficial effects of the present invention:
the first embodiment is as follows:
the Radio Map classified positioning method based on the UE speed is specifically prepared according to the following steps:
the selected experimental environment is a dense urban area in Harbin city with the length of 2.91km and the width of 2.47km, and the experimental data is obtained from real data obtained by a 4G LTE base station. The detection probability and the false alarm probability of each class of sampling points are shown in a table 2.
TABLE 2 UE speed Classification Effect
Low speed (CQT) High speed (DT)
Probability of detection 99.92％ 93.59％
Probability of false alarm 0.47％ 1.17％
The positioning method provided by the invention can distinguish high-speed and low-speed sampling points with high precision, give play to respective fingerprint advantages and effectively improve the positioning precision. The detection probability of the method at low speed is 99.92 percent and the detection probability of the method at high speed is 93.59 percent according to the combination of the table 2; the low-speed false alarm probability is 0.47%, and the high-speed false alarm probability is 1.17%;
the experimental simulation positioning accuracy CDF curve is shown in fig. 3.
With reference to fig. 3, when the positioning error is 10m, the cumulative probability distribution of the present invention is 0.4, and the cumulative probability distribution of the KNN algorithm of the prior art is 0.38; when the positioning error is 60m, the cumulative probability distribution of the method is 0.87, and the cumulative probability distribution of the KNN algorithm in the prior art is 0.82; when the positioning error is 100m, the cumulative probability distribution of the invention is 0.89, and the cumulative probability distribution of the KNN algorithm in the prior art is 0.84.
The positioning method provided by the invention can distinguish high-speed and low-speed sampling points with high precision, give play to respective fingerprint advantages and effectively improve the positioning precision.
For ease of understanding, all variables and physical meanings of the present invention are tabulated below:
Figure GDA0002247844090000111
Figure GDA0002247844090000121
the present invention is capable of other embodiments and its several details are capable of modifications in various obvious respects, all without departing from the spirit and scope of the present invention.

Claims (1)

1. A Radio Map classified positioning method based on UE speed is characterized in that the method comprises the following specific processes:
step one, obtaining DT/CQT/MDT sampling points;
the DT is a drive test; CQT is call quality test, MDT is minimization of drive test;
step two, extracting RSRP symbiotic vectors from DT/CQT/MDT sampling points obtained in the step one;
the RSRP is reference signal received power;
step three, designing a strong classification function based on an Adaboost classification algorithm according to the step two;
step four: constructing an offline Radio Map by using a strong classification function obtained by training in the third step;
the Radio Map is a position fingerprint Map;
fifthly, carrying out online positioning by using the strong classification function obtained by training in the third step and the offline Radio Map obtained in the fourth step;
acquiring DT/CQT/MDT sampling points in the first step; the specific process is as follows:
the method comprises the following steps: acquiring a DT/CQT/MDT sampling point data packet from LTE network optimization data, marking the DT sampling point as high-speed moving UE, marking the CQT sampling point as low-speed moving UE or static UE, and enabling the MDT sampling point to have no moving label; the UE is user terminal equipment;
the UE speed is higher than or equal to 30km/h and is high, and the UE speed is lower than 30km/h and is low;
the first step is: the latitude and longitude of DT, CQT and MDT sampling points with positions
Figure FDA0002247844080000011
Carrying out position gridding, fixing the positions of sampling points to grid nodes closest to each other, and setting grid intervals to be 1 meter; the longitude difference and the latitude difference corresponding to 1 meter are obtained by calculation according to the formula (3),
Figure FDA0002247844080000012
where λ is the longitude of the sampling point UE,
Figure FDA0002247844080000013
for sampling point UE latitude, R is earth radius, lambdaAThe longitude of the point a is taken as the point a,
Figure FDA0002247844080000014
is latitude, λ, at point ABIs the longitude of the point B, and the point B,
Figure FDA0002247844080000015
latitude of point B, LABA, B, the distance between the two points is 1 m;
step one is three: extracting IMSI of the UE from DT/CQT/MDT sampling point data packets obtained in the step one by one to uniquely identify the UE;
the IMSI is an international mobile subscriber identity;
step one is: extracting a measurement time stamp of a current test sampling point from the DT/CQT/MDT sampling point data packet acquired in the step one by one;
step one and five: extracting global cell identification code of service cell from DT/CQT/MDT sampling point data packet obtained from step one by one, and recording the global cell identification code as CGI0(ii) a Extracting the RSRP measurement result of the serving cell and marking as RSRP0；
Step one is six: extracting PCI of the neighboring cell from DT/CQT/MDT sampling point data packets obtained in the steps one by one, wherein the PCI of the neighboring cell at most comprises 6 neighboring cellsZone, denoted PCI1～PCI6(ii) a Obtaining RSRP measurement result corresponding to PCI of adjacent cell and marking as RSRP1～RSRP6；RSRP1Receiving power for reference signals of adjacent cells 1; RSRP6Receiving power of reference signals of adjacent cells 6; PCI is physical cell identification; RSRP is reference signal received power;
step one and step seven: comparing the PCI of the neighboring cell with the I-parameter, and when the PCI of the neighboring cell is completely consistent with the I-parameter, obtaining a CGI of the neighboring cell, and recording the CGI as the CGI1～CGI6；CGI1Is a global cell identity, CGI, of a neighboring cell 16Is a global cell identification code of a neighboring cell 6; CGI is a global cell identity of a neighboring cell;
the working parameters comprise base station positions, physical cell identifiers, carrier frequency points and global cell identification codes;
step one eight: repeating the steps one by one to seven until all DT/CQT/MDT sampling points are obtained;
extracting RSRP symbiotic vectors from DT/CQT/MDT sampling points obtained in the step one in the step two; the specific process is as follows:
step two, firstly: screening sampling points of the same IMSI from the DT/CQT/MDT sampling points obtained in the step one, namely sampling points reported from the same terminal; arranging the sampling points of the same IMSI in an ascending order according to the Timestamp; adding a sliding time window on the sampling point sequence of the same IMSI, wherein the window length is 60s, and the sliding is carried out at an interval of 12 s;
the IMSI is an international mobile subscriber identity;
the Timestamp is the current test sampling point measurement Timestamp;
step two: in a time window, a cell c with the highest detection ratio is selected from a plurality of main regions or adjacent regions detected by UE, and an RSRP vector R of the cell c is extractedc，
Figure FDA0002247844080000021
n is the detected times of the cell c in the current time window, and the value is a positive integer;
Figure FDA0002247844080000022
is a real number;
step two and step three: r is obtainedcMinimum value r of elements in vectorcminMaximum value of the element rcmaxThen the span of RSRP is m-rcmax-rcmin+ 1; to RcThe following treatments were carried out:
R′c＝Rc-rcmin+1 (4)
wherein r iscmin＝[rcminrcmin…rcmin]n×1，1＝[1 1…1]n×1R 'thus obtained'cWherein the elements are positive integers and the minimum value is 1; r'cIs an intermediate variable;
step two, four: initializing RSRP co-occurrence matrix fetching
Figure FDA0002247844080000023
R′cAny one element of r'ciAnd the next time element r'ci+1To obtain a signal intensity variation pair (r'ci,r′ci+1) Let GMMiddle r'ciLine r'ci+1Column element + 1; r'cAll signal strength change pairs taken over, then GMThe construction is completed; m is the span of RSRP, and the value is a positive integer;
step two and step five: r'cIn total, n-1 signal intensity variation pairs, pair GMCarrying out normalization processing to obtain a normalized RSRP co-occurrence matrix
Figure FDA0002247844080000024
Figure FDA0002247844080000025
Step two, step six: co-occurrence matrix P of RSRPMSumming elements in the main diagonal direction, enabling the length of a new vector obtained by summing to be 11, and if the length of the vector obtained by summing is smaller than 11, supplementing 0 with the same number at two ends of the vector; if the length of the summed vector is greater than 11, truncating the same number of 0s at both ends of the vectorObtaining the equal-length RSRP symbiotic vector
Figure FDA0002247844080000031
Designing a strong classification function based on an Adaboost classification algorithm according to the step two in the step three; the specific process is as follows:
step three, firstly: extracting RSRP symbiotic vectors from all DT/CQT sampling points acquired in the step one according to the method in the step two;
step three: initializing DT/CQT sample point weights to D1(i) (ii) 0.5/w (class (i)), where w (class (i)) is the total number of DT/CQT sample points of the tag to which sample point i belongs;
class (i) is the prior class label of sample point i; i is the ith of DT/CQT sampling points;
step three: using CART as a weak classifier; when each weak classifier CART is trained, the prediction output of each weak classifier CART is obtained, and the prediction sequence g is obtainedt(x) The sum of prediction errors, which is calculated by the formula:
Figure FDA0002247844080000032
wherein i satisfies
Figure FDA0002247844080000033
gt(xi) To predict value, yiIs a desired value; phi is a threshold value and takes a value of 0-0.5; dt(i) Weighting the DT/CQT sampling points of the t iteration;
step three and four: according to the predicted sequence gt(x) Is predicted by a prediction error etCalculating the predicted sequence gt(x) Weight of atThe calculation formula is as follows:
Figure FDA0002247844080000034
step three and five: according to the predicted sequence gt(x) Weight of atAnd obtaining the weight of the next round of DT/CQT sampling points, wherein the adjustment formula is as follows:
Figure FDA0002247844080000035
wherein Z istIs a normalization factor so that
Figure FDA0002247844080000036
Dt+1(i) Sample weight for the t +1 th iteration; w is the total number w of DT/CQT sampling points which is w (-1) + w (+1), and the value is a positive integer;
step three and six: repeating the third step, the third step and the fifth step until the upper limit T of the iteration times is reached to obtain T weak classification functions f (g)t(x),at) Combining them to obtain a strong classification function h (x):
Figure FDA0002247844080000041
in the formula, atWeak classifier weights for the t-th iteration; the upper limit T of the iteration times is set by a person and is 10-1000; f (g)t(x),at) A weak classifier for the t-th iteration;
in the fourth step, an offline Radio Map is constructed by using a strong classification function obtained by training in the third step; the specific process is as follows:
step four, firstly: classifying the motion states of the sampling points in the off-line stage by using a strong classification function obtained by training in the third step;
the sampling points in the off-line stage are sampling points with positions in all the MDT sampling points acquired in the step one;
the length of the sliding time window is 60s, the interval is 12s, then each offline stage sampling point can obtain 5 classification results, the state label of the offline stage sampling point moving at a low speed is set as-1, the state label of the offline stage sampling point moving at a high speed is set as +1, and then the state label of each offline stage sampling point is expressed as:
Figure FDA0002247844080000042
hi(x) Predicting a result for the Adaboost strong classifier;
when the state label of the sampling point in the off-line stage moving at low speed is-1, marking the sampling point as a CQT sampling point; when the status label of the off-line stage sampling point moving at high speed is +1, marking as the sampling point of DT;
step four and step two: merging the sampling points marked as DT to construct DT offline Radio Map;
merging
Figure FDA0002247844080000043
Obtaining an RP by the same sampling points marked as DT, and averaging the RSRP measured by each sampling point marked as DT for the same CGI; repeating the steps until all the sampling points marked as DT are combined, and obtaining a Radio Map, wherein the Radio Map comprises three parts of information of position, cell number and signal intensity; assuming j RPs, the number of APs corresponding to each RP may be different because the number of neighboring cells in the measurement report is not fixed; λ is the sampling point UE longitude;
Figure FDA0002247844080000044
the latitude of the UE is taken as a sampling point; j is a positive integer, RP is a reference point, and AP is an access point;
step four and step three: combining the sampling points marked as CQT to construct CQT off-line Radio Map,
merging
Figure FDA0002247844080000045
Obtaining an RP by the same sampling points marked as CQTs, and averaging the RSRPs measured by the sampling points marked as the CQTs for the same CGI; repeating the steps until all the sampling points marked as CQTs are combined, and obtaining a Radio Map, wherein the Radio Map comprises three parts of information of positions, cell numbers and signal intensity; if d RPs are provided, the number of APs corresponding to each RP may be different because the number of neighboring cells in the measurement report is not fixed; λ is the sampling point UE longitude;
Figure FDA0002247844080000046
the latitude of the UE is taken as a sampling point; d is a positive integer;
the DT offline Radio Map and the CQT offline Radio Map form an offline Radio Map;
in the fifth step, the strong classification function obtained by training in the third step and the offline Radio Map obtained in the fourth step are utilized for online positioning; the specific process is as follows:
step five, first: classifying the motion states of the sampling points at the online stage by using a strong classification function obtained by training in the third step;
the online stage sampling points are sampling points without positions in all the MDT sampling points acquired in the step one;
the length of the sliding time window is 60s, the interval is 12s, each online stage sampling point can obtain 5 classification results, the state label of the online stage sampling point moving at a low speed is set as-1, the state label of the online stage sampling point moving at a high speed is set as +1, and the state label of each online stage sampling point is expressed as:
Figure FDA0002247844080000051
in the formula, hi(x) Predicting a result for the Adaboost strong classifier;
when the state label of the sampling point in the online stage moving at low speed is-1, marking as the sampling point of CQT; when label is +1, marking as DT sampling point;
step five two: matching the corresponding off-line Radio Map in the off-line Radio Map obtained in the step four according to the on-line stage sampling point motion state classification result obtained in the step five; screening with inclusion of CGI0～CGInCalculating the Euclidean distance of the signal space of the reference point in the online RSRP and the offline Radio Map; suppose that the pth reference point contains all online CGIs simultaneously, and the corresponding CGI in the reference point is the CGIp0～CGIpnThe corresponding RSRP is RSRPp0～RSRPpnAnd the Euclidean distance d of the signal space of the p-th reference point in the RSRP and the Radio Map of the online sampling pointpComprises the following steps:
Figure FDA0002247844080000052
the reference point is a point obtained by combining sampling points with positions in all the MDT sampling points obtained in the step one; p is positive integer;
in the formula | · | non-conducting phosphor2Is the L2 norm of the matrix; RSRPpRSRP measurements for reference point p for 1 reference point; RSRP is reference signal received power; CGI0A global cell identity for a serving cell for 1 reference point; CGInA global cell identity for neighbor cell n for 1 reference point; n is 1-6; CGIp0A global cell identity of a serving cell that is reference point p; CGIpnA global cell identity of a neighbor cell n being a reference point p; RSRPp0RSRP measurement results for a serving cell that is reference point p; RSRPpnRSRP measurement results of neighbor cells n that are reference points p; p is a positive integer;
step five and step three: selection of Euclidean distance dpA minimum of k reference points, the test point location is estimated by:
Figure FDA0002247844080000061
in the formula (I), the compound is shown in the specification,
Figure FDA0002247844080000062
estimating a longitude for the online reference point;
Figure FDA0002247844080000063
estimating a latitude for the online reference point; k is 3-5; lambda [ alpha ]lLongitude for the selected ith reference point;
Figure FDA0002247844080000064
the latitude of the selected first reference point; dpThe Euclidean distance of the signal space between the RSRP of the online sampling point and the p reference point.
CN201710240569.9A 2017-04-13 2017-04-13 Radio Map classification positioning method based on UE speed Expired - Fee Related CN107027148B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201710240569.9A CN107027148B (en) 2017-04-13 2017-04-13 Radio Map classification positioning method based on UE speed

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201710240569.9A CN107027148B (en) 2017-04-13 2017-04-13 Radio Map classification positioning method based on UE speed

Publications (2)

Publication Number Publication Date
CN107027148A CN107027148A (en) 2017-08-08
CN107027148B true CN107027148B (en) 2020-04-14

Family

ID=59527819

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201710240569.9A Expired - Fee Related CN107027148B (en) 2017-04-13 2017-04-13 Radio Map classification positioning method based on UE speed

Country Status (1)

Country Link
CN (1) CN107027148B (en)

Families Citing this family (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110418285A (en) * 2018-04-26 2019-11-05 中移(苏州)软件技术有限公司 A kind of method for building up and device of radio map
CN111818634B (en) * 2019-04-11 2021-12-28 上海华为技术有限公司 Positioning method, positioning platform and user terminal in 5G scene
CN110290490B (en) * 2019-05-06 2020-12-08 广东工业大学 Construction method of indoor mixed fingerprint map
CN112543470B (en) * 2019-09-23 2023-04-07 中国移动通信集团重庆有限公司 Terminal positioning method and system based on machine learning
CN114035910B (en) * 2020-09-09 2024-05-10 东华大学 Double-parallel machine scheduling method for shortening emergency response time
CN114679749A (en) * 2020-12-25 2022-06-28 展讯半导体(南京)有限公司 Neighbor cell measuring method and device and computer readable storage medium
CN114488247A (en) * 2022-02-19 2022-05-13 北京中科智易科技有限公司 Method for analyzing mobility of equipment based on high-precision Beidou differential positioning

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103167605A (en) * 2013-03-07 2013-06-19 哈尔滨工业大学 WiFi (wireless fidelity) outdoor positioning method for building/updating satellite assisted radio map
CN103889053A (en) * 2014-03-26 2014-06-25 哈尔滨工业大学 Automatic establishing method of self-growing-type fingerprint
CN104602342A (en) * 2015-01-13 2015-05-06 浙江大学 IBeacon device based efficient indoor positioning method
CN105044662A (en) * 2015-05-27 2015-11-11 南京邮电大学 Fingerprint clustering multi-point joint indoor positioning method based on WIFI signal intensity

Family Cites Families (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US9143176B2 (en) * 2013-04-08 2015-09-22 Wavion, Ltd. Method and system for multipath fingerprinting by maximum discrimination techniques

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103167605A (en) * 2013-03-07 2013-06-19 哈尔滨工业大学 WiFi (wireless fidelity) outdoor positioning method for building/updating satellite assisted radio map
CN103889053A (en) * 2014-03-26 2014-06-25 哈尔滨工业大学 Automatic establishing method of self-growing-type fingerprint
CN104602342A (en) * 2015-01-13 2015-05-06 浙江大学 IBeacon device based efficient indoor positioning method
CN105044662A (en) * 2015-05-27 2015-11-11 南京邮电大学 Fingerprint clustering multi-point joint indoor positioning method based on WIFI signal intensity

Non-Patent Citations (5)

* Cited by examiner, † Cited by third party
Title
A Comparative Study on Machine Learning Algorithms for Indoor Positioning;Sinem Bozkurt,etc;《2015 International Symposium on Innovations in Intelligent SysTems and Applications (INISTA)》;20150904;全文 *
Improved AdaBoost-based Fingerprint Algorithm for WiFi Indoor Localization;Yu Feng，etc;《2014 IEEE 7th Joint International Information Technology and Artificial Intelligence Conference》;20141221;全文 *
Performance Evaluation of MDT Assisted LTE RF Fingerprint Framework;Riaz Mondal，etc;《2014 Seventh International Conference on Mobile Computing and Ubiquitous Networking (ICMU)》;20140108;全文 *
Radio map updated method based on subscriber locations in indoor WLAN localization;Lin Ma，etc;《Journal of Systems Engineering and Electronics》;20151231;全文 *
基于动态Radio Map的粒子滤波室内无线定位算法;林以明等;《计算机研究与发展》;20110131;全文 *

Also Published As

Publication number Publication date
CN107027148A (en) 2017-08-08

Similar Documents

Publication Publication Date Title
CN107027148B (en) Radio Map classification positioning method based on UE speed
CN108696932B (en) Outdoor fingerprint positioning method using CSI multipath and machine learning
CN105491661B (en) Indoor locating system and method based on improved Kalman filter algorithm
CN106604228B (en) A kind of fingerprint positioning method based on LTE signaling data
CN103068035A (en) Wireless network location method, device and system
US11419088B2 (en) Radiation source positioning method based on radio spectrum monitoring big data processing
Lakmali et al. Database correlation for GSM location in outdoor & indoor environments
CN109151750B (en) LTE indoor positioning floor distinguishing method based on recurrent neural network model
CN109803274B (en) Antenna azimuth angle optimization method and system
CN106954187A (en) A kind of indoor orientation method based on heterogeneous network
Zhang et al. Machine learning-based integrated wireless sensing and positioning for cellular network
Fang et al. An accurate and real-time commercial indoor localization system in LTE networks
Strelkovskaya et al. Using quadratic complex planar splines in solving local positioning problems
CN108574927B (en) Mobile terminal positioning method and device
Li et al. Outdoor location estimation using received signal strength feedback
Huai et al. Multi-Feature based Outdoor Fingerprint Localization with Accuracy Enhancement for Cellular Network
Zhang et al. Gaussian process based radio map construction for LTE localization
Meniem et al. Relative RSS-based GSM localization technique
Zhang et al. Prnet: Outdoor position recovery for heterogenous telco data by deep neural network
Khalaf-Allah et al. Mobile location in GSM networks using database correlation with bayesian estimation
Lin et al. Dccp: Deep convolutional neural networks for cellular network positioning
Zhao et al. Improved fingerprint location methods for cellular networks
Xu et al. Keeping track of position and cell residual dwell time of cellular networks using HSMM structure and cell-ID information
Lemelson et al. Automatic identification of fingerprint regions for quick and reliable location estimation
Ng et al. Footfall count estimation techniques using mobile data

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
CF01 Termination of patent right due to non-payment of annual fee

Granted publication date: 20200414

Termination date: 20210413

CF01 Termination of patent right due to non-payment of annual fee