CN111751785B - Vehicle visible light positioning method in tunnel environment - Google Patents

Vehicle visible light positioning method in tunnel environment Download PDF

Info

Publication number
CN111751785B
CN111751785B CN202010631380.4A CN202010631380A CN111751785B CN 111751785 B CN111751785 B CN 111751785B CN 202010631380 A CN202010631380 A CN 202010631380A CN 111751785 B CN111751785 B CN 111751785B
Authority
CN
China
Prior art keywords
led
vehicle
distance
positioning
photoelectric detector
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
CN202010631380.4A
Other languages
Chinese (zh)
Other versions
CN111751785A (en
Inventor
宋翔
阎舜
李玲
李丽萍
蒋慧琳
张磊
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.)
Nanjing Xiaozhuang University
Original Assignee
Nanjing Xiaozhuang 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 Nanjing Xiaozhuang University filed Critical Nanjing Xiaozhuang University
Priority to CN202010631380.4A priority Critical patent/CN111751785B/en
Publication of CN111751785A publication Critical patent/CN111751785A/en
Application granted granted Critical
Publication of CN111751785B publication Critical patent/CN111751785B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
    • G01S5/00—Position-fixing by co-ordinating two or more direction or position line determinations; Position-fixing by co-ordinating two or more distance determinations
    • G01S5/16—Position-fixing by co-ordinating two or more direction or position line determinations; Position-fixing by co-ordinating two or more distance determinations using electromagnetic waves other than radio waves
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01C—MEASURING DISTANCES, LEVELS OR BEARINGS; SURVEYING; NAVIGATION; GYROSCOPIC INSTRUMENTS; PHOTOGRAMMETRY OR VIDEOGRAMMETRY
    • G01C21/00—Navigation; Navigational instruments not provided for in groups G01C1/00 - G01C19/00
    • G01C21/20—Instruments for performing navigational calculations
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N20/00—Machine learning
    • G06N20/10—Machine learning using kernel methods, e.g. support vector machines [SVM]

Landscapes

  • Engineering & Computer Science (AREA)
  • Radar, Positioning & Navigation (AREA)
  • Remote Sensing (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Software Systems (AREA)
  • Data Mining & Analysis (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Evolutionary Computation (AREA)
  • Medical Informatics (AREA)
  • Artificial Intelligence (AREA)
  • Computing Systems (AREA)
  • General Engineering & Computer Science (AREA)
  • Mathematical Physics (AREA)
  • Automation & Control Theory (AREA)
  • Electromagnetism (AREA)
  • Navigation (AREA)
  • Traffic Control Systems (AREA)

Abstract

The invention discloses a vehicle visible light positioning method in a tunnel environment based on a support vector machine and extended Kalman filtering. In a tunnel environment where a GPS (Global Positioning System) signal is blocked and cannot be located, an LED (Light Emitting Diode) lamp is arranged in the tunnel, a visible Light communication technology is used, a Least Square Support Vector Machine (LSSVM) algorithm is used to accurately estimate a distance between each LED lamp and a photodetector according to an incident intensity value of a Light signal sent by a vehicle-mounted photodetector receiving the LED lamp, and further the distance is used as an observed quantity, and the known coordinates of each LED lamp are combined, and a kalman filter is used to reliably estimate position information of a vehicle in the tunnel environment in real time.

Description

Vehicle visible light positioning method in tunnel environment
Technical Field
The invention relates to a vehicle visible light positioning method in a tunnel environment based on a support vector machine and extended Kalman filtering, which aims to utilize the visible light communication technology and adopt a least square support vector machine and an extended Kalman filter to realize accurate estimation of the position of a vehicle in a tunnel so as to solve the problem that a global positioning system in the tunnel is difficult to accurately position due to shielding.
Background
In recent years, intelligent Transportation systems its (intelligent Transportation systems) have been rapidly developed worldwide, which effectively and comprehensively apply various advanced technologies such as information technology, data communication transmission technology, electronic control technology, sensor technology, computer processing technology and the like to the Transportation system, thereby realizing more accurate, real-time and efficient comprehensive management and control of the Transportation system and maximally realizing harmony and unity among people, vehicles and roads. The intelligent traffic system can effectively solve traffic problems such as traffic jam, frequent traffic accidents, deteriorated traffic environment and the like, and becomes one of the most main development directions of modern traffic. The vehicle navigation positioning plays an important role in an intelligent traffic system, and by means of accurate positioning information, the ITS can effectively improve the operation efficiency and improve the driving safety, meets the requirements of public security management and vehicle-related resource application service, and is a research hotspot for determining the position of a vehicle quickly, in real time and accurately.
Currently, the most applied technology in the field of vehicle navigation and Positioning is the Global Positioning System (GPS) technology. Under the non-shielding traffic environment, the GPS can provide information such as three-dimensional position, speed, time and the like for the dynamic carrier all weather in real time, thereby being widely applied. However, with the rapid development of three-dimensional traffic, more and more underground tunnels are put into use, in a closed tunnel environment, multipath effect and shielding are serious, a GPS is prone to failure for a long time, and particularly in some long tunnels, failure of several minutes or even ten minutes is often caused, and accurate and reliable continuous positioning cannot be achieved.
In order to make up for the deficiency of GPS positioning, deep research is carried out at home and abroad aiming at the vehicle combined positioning technology based on multi-sensor fusion. The combination of GPS/INS (Inertial Navigation System) and GPS/DR (Dead-reckoning) is the most applied vehicle-mounted combined positioning technology, and when the GPS fails, the INS or DR is used to calculate the positioning. The INS and the DR both comprise an inertial device, namely a gyroscope, and due to the cost, the inertial device widely used in the field of civil vehicle-mounted positioning is basically manufactured by a micro-mechanical MEMS (micro-electromechanical systems), so that larger drift and deviation exist under the limitation of the technical level, once the calculation time is longer, larger errors are accumulated, and the compensation and correction effects are limited.
Besides GPS, other wireless positioning technologies, such as cellular mobile positioning, RFID, UWB, WLAN and Bluetooth, have been developed rapidly in recent years, but mainly applied to indoor positioning, and related positioning algorithms and positioning technologies have been advanced greatly, and some researchers have applied UWB, RFID, WLAN and other technologies to outdoor environment to achieve positioning. However, these techniques require a large number of tags to be arranged in the environment or complex positioning equipment to be built, and thus, the energy consumption and the cost are high, so that the techniques are difficult to popularize on a large scale.
In recent years, with the development of Visible Light Communication (VLC) technology, researchers have proposed a positioning method based on the Visible Light Communication technology, which uses white Light emitted from a semiconductor Light Emitting Diode (LED) as a carrier of a positioning signal to perform positioning. The indoor positioning technology based on the LED visible light communication has the advantages of low energy consumption, high safety and no need of building complex indoor communication equipment. In the existing indoor positioning scheme based on visible light communication, a common basic idea is to use a certain physical quantity transmitted to a photoelectric detector by a measuring LED lamp as a positioning basis so as to calculate the position of the photoelectric detector, wherein the current common physical quantity comprises a distance, an angle and the like, the distance is the most common, namely, the certain direct quantity is measured as the positioning basis, the distance is converted into a distance value, and then the position is calculated by a related algorithm. Common measurements in indoor positioning include Time of Arrival (TOA), Time Difference of Arrival (TDOA), Angle of Arrival (AOA), and Received Signal Strength (RSS), where TOA and TDOA require strict clock synchronization between LEDs, and positioning accuracy is limited by clock accuracy and is susceptible to multipath effects. The AOA needs to deploy an image sensor, the hardware cost is high, the algorithm complexity is high, and the RSS algorithm calculates the transmission distance according to the optical power of the receiving end to finally obtain the position coordinate of the terminal. Among these quantities, the RSS method does not require additional hardware, is lower in cost, is less affected by multipath effects, is simple and easy to implement, and can work in various complex environments, compared to other methods.
The existing visible light positioning based on RSS method is mostly used for indoor positioning, and vehicle positioning applied to tunnel environment is not yet seen, in the indoor positioning, the distance between an LED lamp and a photoelectric detector is generally calculated by an RSS value, and then a target position is calculated, the target position comprises a distance estimation algorithm and a position estimation algorithm, the distance estimation algorithm generally adopts a direct current channel propagation model to obtain a relation function between RSS and the distance, but in the tunnel environment and other environments, due to the reflection, scattering or shielding phenomenon of signals, interference is generally generated on the received signal strength, an accurate conversion model between RSS and the distance is lacked, the distance measurement precision is lower, the positioning performance is influenced, the position estimation algorithm is most widely applied to a multipoint positioning method, but for vehicle positioning, due to the complex and changeable driving environment and the limitation of the existing technical level, the stability and reliability of the optical signal transmission within the effective detection range of the photoelectric detector cannot be guaranteed, and in addition, the influence of scattering, diffraction and multipath effects, the estimated distance value still contains large errors, so that the positioning precision is influenced.
Disclosure of Invention
The invention aims to overcome the defects of the prior art, provides a vehicle visible light positioning method in a tunnel environment based on a support vector machine and extended Kalman filtering, and realizes accurate and reliable estimation of the vehicle position when a GPS (global positioning system) cannot be positioned due to shielding in the tunnel environment. The method adopts a least square support vector machine algorithm to estimate the distance between the LED lamp and the photoelectric detector, utilizes the incident signal intensity as an input quantity, can accurately estimate the distance between the LED lamp and the photoelectric detector, can effectively overcome the defect that the distance information is difficult to estimate by adopting an accurate mathematical model in the tunnel environment, has lower operation complexity, good real-time performance, higher precision and better generalization capability to adapt to changeable application environments; and then an extended Kalman filter is adopted as a positioning algorithm, the distance between the LED lamp and the photoelectric detector, which is acquired by a least square support vector machine, is used as an observed quantity, noise can be effectively filtered, the defects of poor accuracy and dynamic performance of a multipoint positioning algorithm are overcome, accurate and continuous output of positioning information is guaranteed, and the method has the remarkable advantages of high precision, low cost, good real-time performance and the like, and can meet the requirement of automobile positioning and navigation.
A vehicle visible light positioning method in a tunnel environment based on a support vector machine and extended Kalman filtering is characterized in that: in a tunnel environment where a GPS (Global Positioning System) signal is blocked and cannot be positioned, arranging LED (Light Emitting Diode) lamps in the tunnel, utilizing a visible Light communication technology, accurately estimating distances between each LED lamp and a photoelectric detector by adopting a Least Square Support Vector Machine (LSSVM) according to an incident intensity value of a Light signal emitted by a vehicle-mounted photoelectric detector and receiving the incident intensity value of the Light signal emitted by the LED lamp, and further estimating position information of a vehicle in the tunnel environment in real time and reliably by combining known coordinates of each LED lamp by using the distances as observed quantities and expanding a Kalman filter; the method comprises the following specific steps:
1) layout of LED lamp and photoelectric detector
Aiming at the vehicle-mounted environment, the LED lamps are arranged at the two sides of the top layer of the tunnel at equal intervals of 5-10m, and each LED lamp sends an LED optical signal; the method comprises the steps that a geographical coordinate system xoy is established by taking the east as ox direction, the north as oy direction and the origin as o, the origin o is selected on a fixed point on the earth surface, the coordinate position of each LED lamp under the geographical coordinate system can be obtained in advance, a photoelectric detector is installed on the roof of a running vehicle, the photoelectric detector can receive the incident intensity value of an LED optical signal and the position coordinate of the LED lamp and moves along with the vehicle, the position coordinate of the photoelectric detector under the geographical coordinate system can be regarded as the position coordinate of the vehicle under the geographical coordinate system, and the photoelectric detector and the LED lamp can be approximately regarded as being approximately positioned on the same plane;
2) location required information acquisition
According to the layout in step 1, at each discrete time k (k is 1,2,3, … in the present invention), the photodetector can stably receive s nearest to itIn order to ensure the positioning accuracy of the light information emitted by the LED lamp, s is more than or equal to 4, and the information read by the photoelectric detector comprises the incident intensity value RSS of the light emitted by the LED lamp through a visible light communication technologyi(k) (i ═ 1,2.. s) and the position coordinates (X) of the LED luminaire in a geographical coordinate systemi(k),Yi(k))(i＝1,2,...s)；
3) Distance estimation based on LSSVM
The distance between the photoelectric detector and the LED lamp is estimated through a Least Square Support Vector Machine (LSSVM) algorithm, and the incident intensity value RSS of the LED visible light emitted by the LED lamp i (i is 1,2.. s) at the discrete time ki(k) The input quantity of the LSSVM is the output quantity of the LSSVM, the output quantity is the distance between the k photoelectric detector and the LED lamp i at the discrete moment, and training data are acquired from different environments through tests; the training process is completely off-line, and the trained LSSVM model is used for estimating the distance between the photoelectric detector and the LED lamp in real time; in order to ensure the effectiveness of the LSSVM model, the same group of photoelectric detectors and LED lamps are adopted for data acquisition and real-time application, and the richness and typicality of training data are ensured; using LSSVM to establish incident intensity RSSi(k) The steps of the relationship with the distance are as follows:
for a given training data set { x }m,ymN is the total number of samples, where x is 1,2mTo input an RSS vector, ymIs the output distance vector, xm,ym∈R1,R1For a one-dimensional vector space, the LSSVM model can be represented in the feature space as follows:
Figure BDA0002568944100000041
in the formula (1), the reaction mixture is,
Figure BDA0002568944100000042
is an adjustable hyperplane weight vector that,
Figure BDA0002568944100000046
representing a high-dimensional feature space (n)h> 1), superscript' denotes transposing the matrix, b is a scalar threshold, nonlinear mapping
Figure BDA0002568944100000043
Representing a mapping from the low-dimensional input vector to a high-dimensional feature space, such that a linear regression in the high-dimensional feature space corresponds to a non-linear regression in the low-dimensional input space; the objective of the regression problem is to determine the optimal function
Figure BDA0002568944100000044
So that f (x) the unknown input vector can be correctly regressed with as high a probability as possible; the weight ω and the threshold b are quantities to be solved, and the unknown quantity can be determined by the following optimization problem:
Figure BDA0002568944100000045
the constraint conditions are as follows:
Figure BDA00025689441000000511
in the formula, emRepresenting error variable, gamma is more than or equal to 0 and is a regularization constant, and in order to solve the optimization problem, a Lagrangian function L is definedLS-SVM：
Figure BDA0002568944100000051
Wherein alpha ismLagrangian function L for Lagrangian factors according to Karush Kuhn Tucker (KKT) optimization conditionsLS-SVMShould satisfy:
Figure BDA0002568944100000052
in the formula (5), the reaction mixture is,
Figure BDA0002568944100000053
represents LLS-SVMThe derivation of omega is carried out on the basis of,
Figure BDA0002568944100000054
represents LLS-SVMThe result of the derivation of b is obtained,
Figure BDA0002568944100000055
represents LLS-SVMTo emThe derivation is carried out by the derivation,
Figure BDA0002568944100000056
represents LLS-SVMFor alphamDerivation is carried out; the solved optimization problem is thus converted into solving a linear equation:
Figure BDA0002568944100000057
wherein Y ═ Y1,…,yN]′,1N＝[1,…,1]′，α＝[α1,…,αN]′，INFor an N × N unit array, Ω is the kernel function:
Figure BDA0002568944100000058
commonly used kernel functions include linear kernel functions, polynomial kernel functions, gaussian kernel functions, multi-layered perceptron functions, etc., and the present invention employs Radial Basis Functions (RBFs) as kernel functions K (·, ·):
Figure BDA0002568944100000059
wherein
Figure BDA00025689441000000510
Being the square of the euclidean distance between the two space vectors, σ being the width of the RBF, the LSSVM model is approximated as:
Figure BDA0002568944100000061
in the formula, alphamAnd B is a solution of formula (6) such that B ═ Ω + INA/gamma is then alpham＝B-1(Y-b1N)，
Figure BDA0002568944100000062
The values of gamma and sigma can be determined through a large amount of training, and through off-line training, the formula (9) can be used for estimating the distance between the photoelectric detector and the LED lamp i at the discrete moment k on line;
4) vehicle positioning based on extended Kalman filtering
At the discrete time k, the photoelectric detector can measure the incident intensity values of the LED lights emitted by s (s is more than or equal to 4) LED lamps, and the distance r between the photoelectric detector and the LED lamps can be estimated in real time according to the step 31(k),r2(k)...rs(k) Then, the real-time vehicle position can be obtained by establishing an extended Kalman filter; establishing a discrete state equation and a measurement equation as follows:
X(k)＝AX(k-1)+W(k-1)
Z(k)＝h(X(k))+V(k) (10)
where k denotes a discrete time, x (k) ═ pe(k) pn(k)]'is a state vector, superscript' denotes transposing the matrix, pe(k) And pn(k) Respectively representing the east and north positions of the photoelectric detector in the geographical coordinate system set in the step 1) and state transition vectors
Figure BDA0002568944100000063
W and V are respectively system state noise and measurement noise vectors which meet the assumption of Gaussian white noise, and the covariance matrixes of the W and V are respectively Q and R; z ═ r1(k) r2(k) ... rs(k)]' is the observation vector, h is the corresponding observation function:
Figure BDA0002568944100000064
(Xi, Yi) are the coordinates of the known ith LED luminaire,
Figure BDA0002568944100000065
representing a corresponding observed noise vector;
the standard extended kalman filtering process includes time update and measurement update:
and (3) time updating:
Figure BDA0002568944100000066
P(k,k-1)＝A(k,k-1)P(k-1)A′(k,k-1)+Q(k-1)
and (3) measurement updating:
K(k)＝P(k,k-1)H′1(k)[H(k)P(k,k-1)H′(k)+R(k)]-1
Figure BDA0002568944100000067
P(k)＝[I-K(k)H(k)]P(k,k-1)
where I represents the identity matrix and H is the Jacobian matrix of H (-) over X;
Figure BDA0002568944100000071
Figure BDA0002568944100000072
through filtering calculation, the vehicle position at each moment can be calculated, and vehicle positioning under the tunnel environment is realized.
The invention has the advantages and obvious effects that:
1. the invention provides a vehicle positioning method based on the combination of a least square support vector machine and extended Kalman filtering, which adopts a visible light communication technology, is applied to positioning of vehicles in a tunnel environment, only needs visible light information, and has the advantages of high precision, low cost, good instantaneity and reliability and the like.
2. The distance measurement method disclosed by the invention aims at the defect that the traditional indoor distance measurement method is difficult to adapt to the interference of the actual tunnel environment, so that the accuracy is low, modeling and fitting are realized on a nonlinear function by utilizing a least square support vector machine through off-line training of a large amount of test data acquired in different environments, the operation complexity is low, the real-time performance is good, the precision is high, and the adaptability and the generalization performance to different actual use environments are good, so that the accuracy of distance information is ensured.
3. The positioning method adopts the extended Kalman filtering method, can effectively overcome the defect that a common multipoint positioning method in indoor positioning has larger error, has higher positioning precision, only takes distance information as observed quantity, and has high instantaneity and low cost.
4. The visible light communication positioning can be compatible with the LED lighting lamp in the tunnel, so that no or little additional hardware equipment investment is needed, the cost is low, and the large-scale popularization is facilitated.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, it is obvious that the drawings in the following description are only embodiments of the present invention, and for those skilled in the art, other drawings can be obtained according to the provided drawings without creative efforts.
FIG. 1 is a block diagram of a process flow of the proposed invention;
FIG. 2 is a layout of an LED light fixture and a photodetector;
FIG. 3 is a comparison of the distance estimation effect of the distance measurement method of the present invention and the conventional propagation model;
FIG. 4 is a comparison of the ranging method of the present invention with the distance estimation error of the conventional propagation model;
FIG. 5 is a comparison of east and north position errors of the position estimates of the positioning method of the present invention and the conventional multi-point positioning method.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
In recent years, intelligent Transportation systems its (intelligent Transportation systems) have been rapidly developed worldwide, which effectively and comprehensively apply various advanced technologies such as information technology, data communication transmission technology, electronic control technology, sensor technology, computer processing technology and the like to the Transportation system, thereby realizing more accurate, real-time and efficient comprehensive management and control of the Transportation system and maximally realizing harmony and unity among people, vehicles and roads. The intelligent traffic system can effectively solve traffic problems such as traffic jam, frequent traffic accidents, deteriorated traffic environment and the like, and becomes one of the most main development directions of modern traffic. The vehicle navigation positioning plays an important role in an intelligent traffic system, and by means of accurate positioning information, the ITS can effectively improve the operation efficiency and improve the driving safety, meets the requirements of public security management and vehicle-related resource application service, and is a research hotspot for determining the position of a vehicle quickly, in real time and accurately.
Currently, the most applied technology in the field of vehicle navigation and Positioning is the Global Positioning System (GPS) technology. Under the non-shielding traffic environment, the GPS can provide information such as three-dimensional position, speed, time and the like for the dynamic carrier all weather in real time, thereby being widely applied. However, with the rapid development of three-dimensional traffic, more and more underground tunnels are put into use, in a closed tunnel environment, multipath effect and shielding are serious, a GPS is prone to failure for a long time, and particularly in some long tunnels, failure of several minutes or even ten minutes is often caused, and accurate and reliable continuous positioning cannot be achieved.
In order to make up for the deficiency of GPS positioning, deep research is carried out at home and abroad aiming at the vehicle combined positioning technology based on multi-sensor fusion. The combination of GPS/INS (Inertial Navigation System) and GPS/DR (Dead-reckoning) is the most applied vehicle-mounted combined positioning technology, and when the GPS fails, the INS or DR is used to calculate the positioning. The INS and the DR both comprise an inertial device, namely a gyroscope, and due to the cost, the inertial device widely used in the field of civil vehicle-mounted positioning is basically manufactured by a micro-mechanical MEMS (micro-electromechanical systems), so that larger drift and deviation exist under the limitation of the technical level, once the calculation time is longer, larger errors are accumulated, and the compensation and correction effects are limited.
Besides GPS, other wireless positioning technologies, such as cellular mobile positioning, RFID, UWB, WLAN and Bluetooth, have been developed rapidly in recent years, but mainly applied to indoor positioning, and related positioning algorithms and positioning technologies have been advanced greatly, and some researchers have applied UWB, RFID, WLAN and other technologies to outdoor environment to achieve positioning. However, these techniques require a large number of tags to be arranged in the environment or complex positioning equipment to be built, and thus, the energy consumption and the cost are high, so that the techniques are difficult to popularize on a large scale.
In recent years, with the development of Visible Light Communication (VLC) technology, researchers have proposed a positioning method based on the Visible Light Communication technology, which uses white Light emitted from a semiconductor Light Emitting Diode (LED) as a carrier of a positioning signal to perform positioning. The LED-based visible light communication technology has the advantages of low energy consumption, high safety and no need of building complex indoor communication equipment. In the existing indoor positioning scheme based on visible light communication, a common basic idea is to use a certain physical quantity transmitted to a photoelectric detector by a measuring LED lamp as a positioning basis so as to calculate the position of the photoelectric detector, wherein the current common physical quantity comprises a distance, an angle and the like, the distance is the most common, namely, the certain direct quantity is measured as the positioning basis, the distance is converted into a distance value, and then the position is calculated by a related algorithm. Currently, the more common physical quantities that can be taken as the basis for ranging-based positioning in indoor positioning include:
1) time of Arrival (TOA). According to the method, the distance is measured by measuring the propagation time of a visible light signal between the LED lamp and the photoelectric detector, and then the positioning is realized by utilizing a correlation algorithm. In this method, the distance between the reading LED lamp and the photodetector is calculated by multiplying the TOA by the speed of light, and therefore a high-precision clock is required and time synchronization of the receiving and transmitting devices is ensured. The method has the advantages that the precision is limited by the clock precision, the resolution ratio is limited, and in the signal transmission process, the method is easily influenced by the multipath effect.
2) The Time Difference of Arrival (TDOA) method. The method is an improvement on the time of arrival method of the signal. The method is characterized in that the time difference of a visible light signal reaching a plurality of photoelectric detectors is measured, and then the position coordinates of an object to be determined are calculated by utilizing a hyperbolic model. TDOA does not require time synchronization between measurement devices, thus greatly reducing system complexity, as compared to TOA positioning, but its positioning accuracy is limited to clock accuracy, as with the TOA method.
3) Angle of Arrival (AOA) of the signal. The method calculates the distance by measuring the propagation signals of a plurality of signals, can measure the visible light signals emitted by the target to be positioned by combining an image sensor or a plurality of receivers, judges the direction in which the visible light signals are positioned, and can determine the position of the target according to the values of two or more arrival angles. However, the hardware cost of this method is high, the algorithm complexity is also high, and the method is greatly influenced by external environments, such as noise, non-line-of-sight, and the like, when the distance between the photoelectric detectors is large, the positioning performance is obviously reduced, and under the condition of non-line-of-sight, a large positioning error can occur due to the shielding of an obstacle.
4) Incident signal strength (RSS). According to the method, a photoelectric detector calculates the direct current gain of a direct-projection channel according to received RSS, and a common model is a classical Lambert model. The method relies on a channel propagation model of visible light signals, which can be derived theoretically or can be fit according to a large amount of experimental data.
Compared with other methods, the RSS method does not need to add extra hardware, has lower cost, is slightly influenced by multipath effect, has simple and easy algorithm and can work in various complex environments. The method has the defects that an accurate RSS and distance conversion model is lacked, so that the distance measurement precision is low, and the positioning performance is influenced. Based on such considerations, the present invention employs RSS to estimate the distance between the LED luminaire and the photodetector, proposing an improved visible light localization scheme. According to the scheme, the LED lamps are reasonably distributed, an accurate distance calculation model based on an intelligent algorithm is established, and then the positioning is realized by adopting a filtering algorithm.
In order to realize accurate and reliable estimation of the vehicle position when the GPS cannot be positioned due to shielding in a tunnel environment, the invention provides a vehicle positioning method based on the combination of a least square support vector machine and extended Kalman filtering. The method adopts a least square support vector machine, utilizes the intensity of a received signal as an input quantity, can accurately estimate the distance between the LED lamp and the photoelectric detector, and can effectively overcome the defect that the distance information is difficult to estimate by adopting an accurate mathematical model in an actual environment; and then an extended Kalman filter is used as a positioning algorithm, the distance between the LED lamp and the photoelectric detector, which is acquired by a least square support vector machine, is used as an observed quantity, noise can be effectively filtered, the defects of poor accuracy and dynamic performance of a multipoint positioning algorithm are overcome, accurate and continuous output of positioning information is guaranteed, and the method has the remarkable advantages of high precision, low cost, good real-time performance and the like.
For the vehicle-mounted environment, the present invention adopts the layout scheme as shown in fig. 2. The photoelectric detector is arranged on the roof of a running vehicle, the LED lamps are arranged at equal intervals on two sides of a road in an environment without shielding, the positions of the lamps are known, the interval between every two adjacent lamps is d, the width of the road is l, and no shielding object exists between the LED lamps and the photoelectric detector, so that the photoelectric detector can form a circular identification area on a plane, and the effective identification distance of the circular identification area is R. Since the photoelectric detector is fixed on the roof of the vehicle right above the mass center of the vehicle and moves along with the vehicle, the coordinates of the photoelectric detector can be regarded as the coordinates of the vehicle.
In order to ensure the positioning accuracy, the photoelectric detector needs to simultaneously read effective LED optical signals sent by at least 4 LED lamps at any moment, but if the LED lamps are distributed too densely, the hardware cost is correspondingly increased, and the information quantity at the same moment is large, so that the conflict is easy to occur, and the practical application is not facilitated; if the LED lamp spacing is too large, it cannot be guaranteed that the photodetector reads more than 4 effective optical signals at any time, and therefore, a suitable LED lamp layout spacing d needs to be selected according to the difference between the application scene and the positioning accuracy requirement, under the general condition, the width l of the road is a fixed value, u is the minimum spacing of the LED lamp layout, and d satisfies:
Figure BDA0002568944100000111
in the invention, d is 5-10m, and each LED lamp sends an LED optical signal; the method comprises the steps that a geographical coordinate system xoy is established by taking the east as ox direction, the north as oy direction and the origin as o, the origin o is selected on a fixed point on the earth surface, the coordinate position of each LED lamp under the geographical coordinate system can be obtained in advance, a photoelectric detector is installed on the roof of a running vehicle, the photoelectric detector can receive the incident intensity value of an LED optical signal and the position coordinate of the LED lamp and moves along with the vehicle, the position coordinate of the photoelectric detector under the geographical coordinate system can be regarded as the position coordinate of the vehicle under the geographical coordinate system, and the photoelectric detector and the LED lamp can be approximately regarded as being approximately positioned on the same plane;
according to the layout, at each discrete time k (k is 1,2,3, … in the invention), the photodetector can stably read the information of s LED lamps nearest to the photodetector, s is more than or equal to 4 in the invention for ensuring the positioning precision, and the information read by the photodetector comprises the incident signal intensity value RSS of the LED light emitted by the LED lampsi(k) (i ═ 1,2,. s), the position coordinates (X) of the LED luminaire in a geographical coordinate systemi(k),Yi(k))(i＝1,2,...s)；
In order to determine the relationship between the light transmission distance of the LED and the intensity value of the incident signal, a corresponding mathematical model needs to be established, and the accuracy of the model directly influences the positioning precision. The lambertian propagation model is the most commonly used method for determining the direct current gain of an optical channel of an LED light direct path, and in an ideal environment without shielding, the model can better describe the characteristics of LED light propagation, and the relation between RSS and transmission distance can be deduced according to the model. However, in an actual traffic environment, the model usually interferes with the received signal strength due to reflection, scattering or shadowing phenomena of the signal, and these phenomena are usually environment-dependent, and it is difficult to use an accurate mathematical model to characterize the effect of the model on the received signal strength. For example, in a tunnel environment, due to wall reflection and diffraction effects, LED light signals are often affected by severe multipath effects, thereby affecting the effectiveness of the lambertian model. Based on such consideration, the present section proposes a mathematical model for distance estimation based on a Least Squares Support Vector Machine (LSSVM) method with respect to uncertainty of signal propagation.
A Support Vector Machine (SVM) is a modeling method based on artificial intelligence, is a statistical learning method established on the structural Risk minimization principle (SRM), has a complete statistical learning theoretical basis by establishing a classification hyperplane as a decision surface, and has theoretical guarantee on the popularization error bound. The SVM has the advantages of simple calculation, high learning speed, strong generalization capability, no local minimum, strong robustness and the like, so that the SVM is widely applied to solving the problems of nonlinear regression, fitting, classification and the like. Although the SVM has good generalization capability and can always obtain a global optimal solution during training, the training of the SVM is a constrained quadratic programming problem, and the constraint condition is equal to the capacity of a training sample. Therefore, the modeling problem of the SVM for large-capacity training samples is that it will result in too long training time. For this problem, Suykens proposes a Least Squares Support Vector Machine (LSSVM) with a loss function as a quadratic function and a constraint condition in the form of an equation. The training of the LSSVM is a solving problem of a linear equation set, and compared with solving of a quadratic programming problem of SVM training, the method has the advantages that the calculation amount is greatly reduced, the convergence speed is high, and the precision is high. Compared with the traditional artificial intelligence modeling method (such as a neural network), the fitting of the neural network is based on the calculation of error square sum reduction to solve the optimal solution, the problem that the optimal solution is easy to fall into a local minimum point and cannot be obtained globally is solved, the topological structure of the LSSVM is determined by the support vector, so that the congenital problems of the neural network such as high dimensional number, small local sample and the like are solved well, the operation complexity is low, the real-time performance is good, the accuracy is high, and the generalization capability is good so as to adapt to different application environments.
LSSVMs are capable of modeling and fitting nonlinear functions by off-line training of large amounts of experimental data collected in different environments. The method determines the distance between the photoelectric detector and the LED lamp by establishing a Least Square Support Vector Machine (LSSVM), wherein the RSS is the incident signal intensity of the LED visible light emitted by the LED lamp i (i is 1,2.. s) at a discrete time ki(k) The input quantity of the LSSVM is discrete moment k, the output quantity is the distance between the LED lamp i and the photoelectric detector, and training data are acquired from different environments through tests; the training process is completely off-line, and the trained LSSVM model is used for estimating the distance between the LED lamp and the photoelectric detector in real time; in order to ensure the effectiveness of the LSSVM model, the same group of LED lamps and photoelectric detectors are adopted for data acquisition and real-time application, and the richness and typicality of training data are ensured; the steps of establishing the relation between the RSS and the distance by using the LSSVM are as follows:
for a given training data set { x }m,ymN is the total number of samples, where x is 1,2mTo input an RSS vector, ymIs the output distance vector, xm,ym∈R1,R1For a one-dimensional vector space, the LSSVM model can be represented in the feature space as follows:
Figure BDA0002568944100000121
in the formula (1), the reaction mixture is,
Figure BDA0002568944100000122
is made byThe adjusted hyperplane weight vector is then used,
Figure BDA0002568944100000123
representing a high-dimensional feature space (n)h> 1), superscript' denotes transposing the matrix, b is a scalar threshold, nonlinear mapping
Figure BDA0002568944100000124
Representing a mapping from the low-dimensional input vector to a high-dimensional feature space, such that a linear regression in the high-dimensional feature space corresponds to a non-linear regression in the low-dimensional input space; the objective of the regression problem is to determine the optimal function
Figure BDA0002568944100000125
So that f (x) the unknown input vector can be correctly regressed with as high a probability as possible; the weight ω and the threshold b are quantities to be solved, and the unknown quantity can be determined by the following optimization problem:
Figure BDA0002568944100000131
the constraint conditions are as follows:
Figure BDA0002568944100000132
in the formula, emRepresenting error variable, gamma is more than or equal to 0 and is a regularization constant, and in order to solve the optimization problem, a Lagrangian function L is definedLS-SVM：
Figure BDA0002568944100000133
Wherein alpha ismLagrangian function L for Lagrangian factors according to Karush Kuhn Tucker (KKT) optimization conditionsLS-SVMShould satisfy:
Figure BDA0002568944100000134
in the formula (5), the reaction mixture is,
Figure BDA0002568944100000135
represents LLS-SVMThe derivation of omega is carried out on the basis of,
Figure BDA0002568944100000136
represents LLS-SVMThe result of the derivation of b is obtained,
Figure BDA0002568944100000137
represents LLS-SVMTo emThe derivation is carried out by the derivation,
Figure BDA0002568944100000138
represents LLS-SVMFor alphamDerivation is carried out; the solved optimization problem is thus converted into solving a linear equation:
Figure BDA0002568944100000139
wherein Y ═ Y1,…,yN]′,1N＝[1,…,1]′，α＝[α1,…,αN]′，INFor an N × N unit array, Ω is the kernel function:
Figure BDA00025689441000001310
commonly used kernel functions include linear kernel functions, polynomial kernel functions, gaussian kernel functions, multi-layered perceptron functions, etc., and the present invention employs Radial Basis Functions (RBFs) as kernel functions K (·, ·):
Figure BDA00025689441000001311
wherein
Figure BDA0002568944100000141
Being the square of the euclidean distance between the two space vectors, σ being the width of the RBF, the LSSVM model is approximated as:
Figure BDA0002568944100000142
in the formula, alphamAnd B is a solution of formula (6) such that B ═ Ω + INA/gamma is then alpham＝B-1(Y-b1N)，
Figure BDA0002568944100000143
The values of gamma and sigma can be determined through a large amount of training, and through off-line training, the formula (9) can be used for estimating the distance between the photoelectric detector and the LED lamp i at the discrete moment k on line;
because the positions of the LED lamps are determined in advance, after the relation between RSS and the distance is determined by using an LSSVM algorithm, the absolute position coordinates of the photoelectric detector can be calculated according to the acquired distance between the photoelectric detector and each LED lamp, and the vehicle can be positioned.
In the field of indoor positioning, the most widely applied positioning method based on position is a multipoint positioning method, assuming that at discrete time k, a photodetector can simultaneously read effective visible light signals emitted by i (i > ═ 3) LED lamps, the coordinates of the LED lamps are (xi, yi), the LSSVM method can determine the distance ri from the photodetector to each LED lamp, and then the unknown coordinates (x, y) of the photodetector can be obtained by the following formula:
Figure BDA0002568944100000144
the coordinate of the photoelectric detector, namely the vehicle coordinate, can be obtained by solving the formula (10), wherein the formula (10) is an over-determined equation set (namely the number of equations is greater than the number of unknowns), the redundant equations are beneficial to improving the positioning precision, and common solving methods comprise recursive least squares, a maximum likelihood method and the like. However, for vehicle positioning, due to the complex and variable driving environment and the limitation of the prior art, it cannot be guaranteed that visible light transmission within the effective identification range of the photoelectric detector is stable and reliable, and in addition, due to the influence of scattering, diffraction and multipath effects, the estimated distance value still contains large errors, so that the accuracy of primary positioning is influenced. In order to inhibit the errors and improve the accuracy of primary positioning, the invention does not adopt a multipoint positioning method, but adopts a filtering calculation method to carry out primary positioning.
Kalman filtering and extended kalman filtering are the most common methods for filter positioning. Because the positioning problem in this chapter is a nonlinear problem, the initial positioning needs to be realized by extended kalman filtering. The extended Kalman filter is an optimal state estimation filter taking minimum mean square error as a criterion, the estimation of real-time signals can be realized only by carrying out recursion calculation by a computer according to a current observation value and an estimation value at the previous moment without storing a past measurement value, and the extended Kalman filter has the characteristics of small data storage amount and simple and convenient algorithm.
At the discrete time k, the photoelectric detector can measure the incident intensity values of the LED lights emitted by s (s is more than or equal to 4) LED lamps, and the distance r between the photoelectric detector and the LED lamps can be estimated in real time according to the step 31(k),r2(k)...rs(k) Then, the real-time vehicle position can be obtained by establishing an extended Kalman filter; establishing a discrete state equation and a measurement equation as follows:
X(k)＝AX(k-1)+W(k-1)
Z(k)＝h(X(k))+V(k) (11)
where k denotes a discrete time, x (k) ═ pe(k) pn(k)]'is a state vector, superscript' denotes transposing the matrix, pe(k) And pn(k) Respectively representing the east and north positions of the photoelectric detector in the geographical coordinate system set in the step 1) and state transition vectors
Figure BDA0002568944100000151
W and V are respectively system state noise and measurement noise vectors which meet the assumption of Gaussian white noise, and the covariance matrixes of the W and V are respectively Q and R; z ═ r1(k) r2(k) ... rs(k)]' is the observation vector, h is the corresponding observation function:
Figure BDA0002568944100000152
(Xi, Yi) are the coordinates of the known ith LED luminaire,
Figure BDA0002568944100000156
representing a corresponding observed noise vector;
the standard extended kalman filtering process includes time update and measurement update:
and (3) time updating:
Figure BDA0002568944100000153
P(k,k-1)＝A(k,k-1)P(k-1)A′(k,k-1)+Q(k-1)
and (3) measurement updating:
K(k)＝P(k,k-1)H′1(k)·[H(k)P(k,k-1)H′(k)+R(k)]-1
Figure BDA0002568944100000155
P(k)＝[I-K(k)·H(k)]P(k,k-1)
where I represents the identity matrix and H is the Jacobian matrix of H (-) over X;
Figure BDA0002568944100000154
Figure BDA0002568944100000161
through filtering calculation, the vehicle position at each moment can be calculated, and vehicle positioning under the tunnel environment is realized.
EXAMPLES example 2
In order to test the actual effect of the vehicle positioning method based on the distance measurement of the support vector machine under the tunnel environment, the vehicle positioning method is verified through a real vehicle test. The test vehicle is a Buick Sail SRV, the test vehicle is provided with a photoelectric detector, the receiving frequency of the photoelectric detector is 1Hz, and a high-precision Differential GPS (DGPS) measured value is used as a reference true value to verify the algorithm positioning effect. The RSS range of the visible light of the LED is normalized to 0 to 255, and a large number of previous experiments show that the RSS signal is attenuated very much beyond 6m, and therefore, in order to ensure the reliability of the signal, the distance between the LED lamps is set to 6m when the LED lamps of the present embodiment are arranged.
In different scenes such as indoor and outdoor non-shielding environments, tunnels and the like, the relation between the distance from the LED lamp to the photoelectric detector is estimated by adopting an LSSVM algorithm. In comparison, a lambertian propagation model is also used to estimate the distance between the LED fixture and the photodetector. The estimation result in the tunnel environment is shown in fig. 3, and the distance estimation error is shown in fig. 4. In fig. 4, the error value is obtained by comparing the fitting value with the reference true value. Table 1 lists the Mean (Mean) and Standard deviation (STD) of the fitting errors of the LSSVM method to the lambertian model under different circumstances.
TABLE 1 mean and standard deviation of fitting errors
Figure BDA0002568944100000162
As can be seen from fig. 3 and 4 and table 1, the LSSVM algorithm achieves better effects than the lambertian propagation model, and can provide more accurate distance information, thereby improving the performance of the subsequent initial positioning algorithm. Particularly in a tunnel environment, due to the influence of multipath effect and the like, the estimation performance of the Lambert model is obviously reduced, and the LSSVM algorithm is obviously improved, because the LSSVM algorithm has the learning capability in a specific environment, the LSSVM algorithm has strong adaptability to different environments.
In order to verify the effect of the positioning algorithm based on the extended kalman filter proposed by the present invention, a multipoint positioning (multilateration) method is also applied as a comparison. The east and north errors of the position estimates for both algorithms are shown in fig. 5. Table 2 shows the Euclidean distance error statistics when the vehicle is fully driven in the tunnel, the statistics including the maximum error Max and the root mean square error RMS.
TABLE 2 Tunnel inner positioning error statistics (unit: m)
Figure BDA0002568944100000171
From a comparison of table 2 and fig. 5, it can be seen that the method of the present invention has a high positioning accuracy.
In conclusion, the vehicle visible light positioning method based on the support vector machine and the extended Kalman filtering under the tunnel environment can accurately estimate the vehicle position information under the tunnel environment, only visible light communication information is needed, and the method has the advantages of high precision, low cost, good real-time performance and reliability and the like.
The method for positioning the visible light of the vehicle in the tunnel environment based on the support vector machine and the extended kalman filter is described in detail, and a specific example is applied to explain the principle and the implementation mode of the method, and the description of the embodiment is only used for helping to understand the method and the core idea of the method; meanwhile, for a person skilled in the art, according to the idea of the present invention, there may be variations in the specific embodiments and the application scope, and in summary, the content of the present specification should not be construed as a limitation to the present invention.
It should be noted that, in the present specification, the embodiments are all described in a progressive manner, each embodiment focuses on differences from other embodiments, and the same and similar parts among the embodiments may be referred to each other. The device disclosed by the embodiment corresponds to the method disclosed by the embodiment, so that the description is simple, and the relevant points can be referred to the method part for description.
It is further noted that, herein, relational terms such as first and second, and the like may be used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions. Also, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include or include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising an … …" does not exclude the presence of other identical elements in a process, method, article, or apparatus that comprises the element.
The previous description of the disclosed embodiments is provided to enable any person skilled in the art to make or use the present invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the generic principles defined herein may be applied to other embodiments without departing from the spirit or scope of the invention. Thus, the present invention is not intended to be limited to the embodiments shown herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims (1)

1. A vehicle visible light positioning method in a tunnel environment based on a support vector machine and extended Kalman filtering is characterized in that: in a tunnel environment where a GPS (Global Positioning System) signal is blocked and cannot be located, arranging an LED (Light Emitting Diode) lamp in the tunnel, using a visible Light communication technology, accurately estimating a distance between each LED lamp and a photodetector by using a Least Square Support Vector Machine (LSSVM) according to an incident intensity value of a Light signal emitted by a vehicle-mounted photodetector received by the LED lamp, and further estimating position information of a vehicle in the tunnel environment in real time and reliably by using the distance as an observed quantity and combining known coordinates of each LED lamp through an extended kalman filter; the method comprises the following specific steps:
1) layout of LED lamp and photoelectric detector
Aiming at a vehicle-mounted environment, the LED lamps are arranged at the two sides of the top layer of the tunnel at equal intervals of 5-10m, and each LED lamp sends an LED optical signal; the method comprises the steps that a geographical coordinate system xoy is established by taking the east as ox direction, the north as oy direction and the origin as o, the origin o is selected on a fixed point on the earth surface, the coordinate position of each LED lamp under the geographical coordinate system can be obtained in advance, a photoelectric detector is installed on the roof of a running vehicle, the photoelectric detector receives the incident intensity value of an LED optical signal and the position coordinate of the LED lamp and moves along with the vehicle, and at the moment, the position coordinate of the photoelectric detector under the geographical coordinate system can be regarded as the position coordinate of the vehicle under the geographical coordinate system and can be regarded as that the photoelectric detector and the LED lamp are approximately positioned on the same plane approximately;
2) location required information acquisition
According to the layout in step 1, at each discrete time k, where k is 1,2,3, …, the photodetector can stably receive the light information emitted by s LED lamps nearest to the photodetector, and s is greater than or equal to 4 to ensure the positioning accuracy, through the visible light communication technology, the information that can be read by the photodetector includes the incident intensity value RSS of the light emitted by the LED lampsi(k) (i ═ 1,2.. s) and the position coordinates (X) of the LED luminaire in a geographical coordinate systemi(k),Yi(k))(i＝1,2,...s)；
3) Distance estimation based on LSSVM
Estimating the distance between the photodetector and the LED luminaire by using a Least Square Support Vector Machine (LSSVM) algorithm, wherein the LED luminaire i (i ═ 1,2.. s) emits an incident intensity value RSS of LED visible light at a discrete time ki(k) The input quantity of the LSSVM is the output quantity of the LSSVM, the output quantity is the distance between the k photoelectric detector and the LED lamp i at the discrete moment, and training data are acquired from different environments through tests; the training process is completely off-line, and the trained LSSVM model is used for estimating the distance between the photoelectric detector and the LED lamp in real time; in order to ensure the effectiveness of the LSSVM model, the same set of photoelectric detectors and LED lamps are adopted for data acquisition and real-time application, andensuring the richness and typicality of training data; using LSSVM to establish incident intensity RSSi(k) The steps of the relationship with the distance are as follows:
for a given training data set { x }m,ymN is the total number of samples, where x is 1,2mTo input an RSS vector, ymIs the output distance vector, xm,ym∈R1,R1For a one-dimensional vector space, the LSSVM model can be represented in the feature space as follows:
Figure FDA0003493890170000027
in the formula (1), the reaction mixture is,
Figure FDA0003493890170000026
is an adjustable hyperplane weight vector that,
Figure FDA0003493890170000025
representing a high-dimensional feature space, where nh> 1, superscript' denotes transposing the matrix, b is a scalar threshold, nonlinear mapping
Figure FDA0003493890170000028
Representing a mapping from a low-dimensional input vector to a high-dimensional feature space; the weight ω and the threshold b are quantities to be solved, and can be determined by the following optimization problem:
Figure FDA0003493890170000021
the constraint conditions are as follows:
Figure FDA0003493890170000022
in the formula, emDefining Lagrange as error variable and gamma ≧ 0 as regularization constant(Lagrangian) function LLS-SVM：
Figure FDA0003493890170000023
Wherein alpha ismLagrangian function L for Lagrangian factors according to Karush Kuhn Tucker (KKT) optimization conditionsLS-SVMShould satisfy:
Figure FDA0003493890170000024
in the formula (5), the reaction mixture is,
Figure FDA0003493890170000031
represents LLS-SVMThe derivation of omega is carried out on the basis of,
Figure FDA0003493890170000032
represents LLS-SVMThe result of the derivation of b is obtained,
Figure FDA0003493890170000033
represents LLS-SVMTo emThe derivation is carried out by the derivation,
Figure FDA0003493890170000034
represents LLS-SVMFor alphamDerivation is carried out; the solved optimization problem is thus converted into solving a linear equation:
Figure FDA0003493890170000035
wherein Y ═ Y1,…,yN]′,1N＝[1,…,1]′，α＝[α1,…,αN]′，INFor an N × N unit array, Ω is the kernel function:
Figure FDA0003493890170000036
taking a Radial Basis Function (RBF) as the kernel function K (·, · for):
Figure FDA0003493890170000037
wherein
Figure FDA0003493890170000038
Being the square of the euclidean distance between the two space vectors, σ being the width of the RBF, the LSSVM model is approximated as:
Figure FDA0003493890170000039
in the formula, alphamAnd B is a solution of formula (6) such that B ═ Ω + INA/gamma is then alpham＝B-1(Y-b1N)，
Figure FDA00034938901700000310
The values of gamma and sigma can be determined through a large amount of training, and through offline training, the formula (9) can be used for estimating the distance between the photoelectric detector and the LED lamp i at the discrete moment k on line;
4) vehicle positioning based on extended Kalman filtering
At the discrete time k, the photoelectric detector can measure the incident intensity values of the LED lights emitted by s (s is more than or equal to 4) LED lamps, and the distance r between the photoelectric detector and the LED lamps can be estimated in real time according to the step 31(k),r2(k)...rs(k) Acquiring the real-time vehicle position by establishing an extended Kalman filter; establishing a discrete state equation and a measurement equation as follows:
Figure FDA00034938901700000311
Z(k)＝h(X(k))+V(k) (10)
where k denotes a discrete time, x (k) ═ pe(k) pn(k)]'is a state vector, superscript' denotes transposing the matrix, pe(k) And pn(k) Respectively representing the east and north positions of the photoelectric detector in the geographical coordinate system set in the step 1) and state transition vectors
Figure FDA0003493890170000041
W and V are respectively system state noise and measurement noise vectors which meet the assumption of Gaussian white noise, and the covariance matrixes of the W and V are respectively Q and R;
z＝[r1(k) r2(k)...rs(k)]' is the observation vector, h is the corresponding observation function:
Figure FDA0003493890170000042
(i 1,2.., s); (Xi, Yi) are the coordinates of the known ith LED luminaire,
Figure FDA0003493890170000043
representing a corresponding observed noise vector;
the standard extended kalman filtering process includes time update and measurement update:
and (3) time updating:
Figure FDA0003493890170000044
P(k，k-1)＝A(k，k-1)P(k-1)A′(k，k-1)+Q(k-1)
and (3) measurement updating:
K(k)＝P(k，k-1)H′1(k)[H(k)P(k，k-1)H′(k)+R(k)]-1
Figure FDA0003493890170000045
P(k)＝[I-K(k)H(k)]P(k，k-1)
where I represents the identity matrix and H is the Jacobian matrix of H (-) over X;
Figure FDA0003493890170000046
Figure FDA0003493890170000047
and calculating the vehicle position at each moment through filtering calculation, and realizing vehicle positioning in the tunnel environment.
CN202010631380.4A 2020-07-03 2020-07-03 Vehicle visible light positioning method in tunnel environment Active CN111751785B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202010631380.4A CN111751785B (en) 2020-07-03 2020-07-03 Vehicle visible light positioning method in tunnel environment

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202010631380.4A CN111751785B (en) 2020-07-03 2020-07-03 Vehicle visible light positioning method in tunnel environment

Publications (2)

Publication Number Publication Date
CN111751785A CN111751785A (en) 2020-10-09
CN111751785B true CN111751785B (en) 2022-04-12

Family

ID=72678880

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202010631380.4A Active CN111751785B (en) 2020-07-03 2020-07-03 Vehicle visible light positioning method in tunnel environment

Country Status (1)

Country Link
CN (1) CN111751785B (en)

Families Citing this family (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112925000B (en) * 2021-01-25 2022-05-17 东南大学 Vehicle positioning method in tunnel environment based on visible light communication and inertial navigation
CN112965031B (en) * 2021-02-20 2023-11-21 兰州交通大学 Subway train positioning model and positioning method based on VLC-RSSI
CN113132006B (en) * 2021-04-23 2022-08-16 湖南大学 High-precision visible light positioning method for moving vehicle based on image sensor
CN113702906B (en) * 2021-08-31 2023-03-17 苏州大学 Three-dimensional wireless optical positioning method and system
CN117269887B (en) * 2023-11-21 2024-05-14 荣耀终端有限公司 Positioning method, electronic equipment and readable storage medium

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102928816A (en) * 2012-11-07 2013-02-13 东南大学 High-reliably integrated positioning method for vehicles in tunnel environment
CN105467382A (en) * 2015-12-31 2016-04-06 南京信息工程大学 SVM (Support Vector Machine)-based multi-sensor target tracking data fusion algorithm and system thereof
CN109195104A (en) * 2018-08-27 2019-01-11 上海市计量测试技术研究院 A kind of indoor orientation method combined based on support vector regression and Kalman filtering
CN109511095A (en) * 2018-11-30 2019-03-22 长江大学 A kind of visible light localization method and system based on Support vector regression

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102928816A (en) * 2012-11-07 2013-02-13 东南大学 High-reliably integrated positioning method for vehicles in tunnel environment
CN105467382A (en) * 2015-12-31 2016-04-06 南京信息工程大学 SVM (Support Vector Machine)-based multi-sensor target tracking data fusion algorithm and system thereof
CN109195104A (en) * 2018-08-27 2019-01-11 上海市计量测试技术研究院 A kind of indoor orientation method combined based on support vector regression and Kalman filtering
CN109511095A (en) * 2018-11-30 2019-03-22 长江大学 A kind of visible light localization method and system based on Support vector regression

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
基于小波神经网络的多传感器自适应融合算法;原泉等;《北京航空航天大学学报》;20081130;第34卷(第11期);第1331-1334页 *

Also Published As

Publication number Publication date
CN111751785A (en) 2020-10-09

Similar Documents

Publication Publication Date Title
CN111751785B (en) Vehicle visible light positioning method in tunnel environment
CN103199923B (en) A kind of underground moving target light fingerprint location tracking based on visible light communication
US9194933B2 (en) Context and map aiding for self-learning
CN101191832B (en) Wireless sensor network node position finding process based on range measurement
CN102932742B (en) Based on indoor orientation method and the system of inertial sensor and wireless signal feature
CN112533163B (en) Indoor positioning method based on NB-IoT (NB-IoT) improved fusion ultra-wideband and Bluetooth
CN107014375B (en) Indoor positioning system and method with ultra-low deployment
CN113706612B (en) Underground coal mine vehicle positioning method fusing UWB and monocular vision SLAM
CN112925000B (en) Vehicle positioning method in tunnel environment based on visible light communication and inertial navigation
CN110187372B (en) Combined navigation method and system in low-speed unmanned vehicle park
CN112927565B (en) Method, device and system for improving accuracy of comprehensive track monitoring data of apron
CN109188360A (en) A kind of indoor visible light 3-D positioning method based on bat algorithm
CN110673181A (en) GNSS interference source positioning method based on grid energy traversal search
CN103096465A (en) Environment self-adaption multi-target direct locating method
Marquez et al. Understanding LoRa-based localization: Foundations and challenges
Svertoka et al. Evaluation of real-life LoRaWAN localization: Accuracy dependencies analysis based on outdoor measurement datasets
CN114758364B (en) Industrial Internet of things scene fusion positioning method and system based on deep learning
KR20170088732A (en) Indoor Positioning System and Method
CN114521014B (en) Method for improving positioning precision in UWB positioning process
CN113271542A (en) Indoor mobile terminal positioning method based on Bluetooth and visible light
Kuxdorf-Alkirata et al. Improved energy efficiency of indoor positioning systems by adaptive sampling and smart post-processing of sensor data
Lou et al. Indoor localization and map building for autonomous mobile robot
CN116939815B (en) UWB positioning base station selection method based on laser point cloud map
Cao et al. The design of LED array for single anchor-based visible light positioning
US20230123690A1 (en) Systems and methods for locating tagged objects in remote regions

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