CN110231636B - Self-adaptive unscented Kalman filtering method of GPS and BDS dual-mode satellite navigation system - Google Patents

Self-adaptive unscented Kalman filtering method of GPS and BDS dual-mode satellite navigation system Download PDF

Info

Publication number
CN110231636B
CN110231636B CN201910357048.0A CN201910357048A CN110231636B CN 110231636 B CN110231636 B CN 110231636B CN 201910357048 A CN201910357048 A CN 201910357048A CN 110231636 B CN110231636 B CN 110231636B
Authority
CN
China
Prior art keywords
noise
gps
bds
hdop
state
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
CN201910357048.0A
Other languages
Chinese (zh)
Other versions
CN110231636A (en
Inventor
罗楚江
王洽和
陈靖
王铠帆
谢达裕
王景森
徐虎
白明
雷杰
曾清林
许祖浩
陈志朋
黄钰平
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.)
Shenzhen Union Security Industry Co ltd
Original Assignee
Shenzhen Union Security Industry Co ltd
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 Shenzhen Union Security Industry Co ltd filed Critical Shenzhen Union Security Industry Co ltd
Priority to CN201910357048.0A priority Critical patent/CN110231636B/en
Publication of CN110231636A publication Critical patent/CN110231636A/en
Application granted granted Critical
Publication of CN110231636B publication Critical patent/CN110231636B/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
    • G01S19/00—Satellite radio beacon positioning systems; Determining position, velocity or attitude using signals transmitted by such systems
    • G01S19/01—Satellite radio beacon positioning systems transmitting time-stamped messages, e.g. GPS [Global Positioning System], GLONASS [Global Orbiting Navigation Satellite System] or GALILEO
    • G01S19/13—Receivers
    • G01S19/24—Acquisition or tracking or demodulation of signals transmitted by the system

Landscapes

  • Engineering & Computer Science (AREA)
  • Radar, Positioning & Navigation (AREA)
  • Remote Sensing (AREA)
  • Computer Networks & Wireless Communication (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Position Fixing By Use Of Radio Waves (AREA)

Abstract

The invention discloses a self-adaptive unscented Kalman filtering method of a GPS and BDS dual-mode satellite navigation system, which comprises the following steps: the method comprises the following steps: modeling a nonlinear system; step two: performing Newton motion law dispersion, and performing equivalent substitution on the Newton motion law dispersion to enable the Newton motion law dispersion to meet a model of a nonlinear system; step three: the HDOP self-adaptive optimization algorithm selects a GNSS working system with better horizontal accuracy to provide GNSS data, and substitutes the GNSS data into an equivalent substitution nonlinear system model; the method comprises the steps of processing nonlinear transmission of mean values and covariance by using an unscented transformation method, sampling Sigma points by using a UKF algorithm to solve uncertainty of noise, estimating and correcting the noise in real time so as to solve nonlinear characteristics of a combined navigation system, and deducing a noise statistical estimator to estimate and correct unknown or inaccurate noise statistics in real time based on a maximum likelihood criterion. The invention can obviously improve the precision of the GPS and BDS dual-mode satellite navigation system and has good effect when the model nonlinearity and the noise statistical characteristic of the processing system are uncertain.

Description

Self-adaptive unscented Kalman filtering method of GPS and BDS dual-mode satellite navigation system
Technical Field
The invention relates to a filtering method of a satellite navigation system, in particular to a self-adaptive unscented Kalman filtering method of a GPS and BDS dual-mode satellite navigation system.
Background
With the social development and the increase of vehicles, the urban traffic road conditions are increasingly complex, and people put higher requirements on the precision of positioning navigation instruments. However, all the existing systems for positioning navigation instruments have limitations, so that the instruments cannot fully meet the requirements of people. In order to improve urban traffic, meet the requirements of people, realize high-precision, all-weather and global positioning, various positioning navigation systems can be combined, the advantages are raised, the disadvantages are avoided, and a multi-element organic combined navigation system is formed, so that the precision and the reliability of the positioning navigation system are improved.
In the prior art, please refer to chinese patent application No. 201710724015.6, publication No. 109425876a, entitled "an improved kalman filtering method for improving positioning accuracy", which has a drawback in that, for the non-linear problem, Extended Kalman Filtering (EKF) linearizes a non-linear model using taylor decomposition, which is difficult to calculate and has a linear error, and compared with Unscented Kalman Filtering (UKF), the accuracy of the model is reduced by introducing the linear error.
Compared with EKF, UKF has the following advantages: firstly, the nonlinear system does not need to be linearized, and the calculation error introduced in the linearization process of the nonlinear system is avoided; secondly, the unscented transformation adopted by the UKF can completely reproduce the mean value and the covariance of the state quantity of the nonlinear system; thirdly, the filtering precision of any nonlinear system can reach more than three orders of a Taylor series expansion, and the EKF can only reach the precision of one order; in addition, the UKF does not need to calculate a Jacobian matrix, so that the calculation amount of the UKF is greatly simplified, and calculation errors are not introduced due to the fact that the Jacobian matrix does not need to be calculated.
Disclosure of Invention
The invention aims to solve the technical problem of providing a self-adaptive unscented Kalman filtering method for a GPS and BDS dual-mode satellite navigation system, which can improve the precision of vehicle integrated navigation and aims at overcoming the defects of the prior art.
In order to solve the technical problems, the invention adopts the following technical scheme.
An adaptive unscented kalman filtering method for a GPS and BDS dual-mode satellite navigation system, the method comprising:
and step S1, establishing a nonlinear system model. The nonlinear system was modeled as:
Figure BDA0002045739550000021
in the formula, xkIs the state vector at time k, zkIs the observation vector at time k, fk(. and h)k(. respectively) State function and Observation function of the nonlinear System, wkAnd vkThe two white Gaussian noises are uncorrelated and the mean value and the covariance matrix of the two white Gaussian noises respectively satisfy the following conditions:
Figure BDA0002045739550000022
in the formula, q is a state noise mean value, and r is an observation noise mean value;
and step S2, the Newton motion law is discrete and equivalently substituted for the nonlinear system model. The state vectors are respectively taken as xe＝[ve，ae]T，xn＝[vn，an]T，xeAnd xnEast and north speed, respectively, aeAnd anThe two state vectors are written together as:
x＝[v，a]T；
in the following formula, the sampling interval is set to T, vk-1To observe the instantaneous speed, v, at the starting momentkInstantaneous speed after T time, ak-1The acceleration in T time is obtained by discretizing the Newton's law of motion:
Figure BDA0002045739550000031
vk＝vk-1+ak-1T；
the output sentences of the GPS dual-mode receiver and the BDS dual-mode receiver are used for respectively providing the position and the speed of the east direction and the north direction, and the observation direction is measured as z ═ x, v]TWherein the initial state quantity x of the system0And stateThe noise q and the observation noise r are not related to each other, and the initial state follows Gaussian normal distribution, which
Figure BDA0002045739550000034
A priori mean sum P0The covariance matrices are:
Figure BDA0002045739550000032
let mu letk＝Wk-q，ηk＝vkR, which is substituted into the nonlinear system model to obtain:
Figure BDA0002045739550000033
in the formula, muk-1And ηkAre all white gaussian noise with a mean of zero and variance of Q and R, respectively.
And step S3, carrying out HDOP self-adaptive optimization model algorithm. And extracting a horizontal component precision factor (HDOP) in an output statement of the GPS and BDS dual-mode receiver, wherein the HDOP determines the reliability of the positioning precision and is used as an input value of an HDOP self-adaptive optimization algorithm.
First, an HDOP prediction model is built:
YH(i+1)＝k(i+1)+b
[k，b]T＝r(I，H)
wherein, YH(i+1)The predicted value of the HDOP difference value of the BDS and the GPS at the moment of i +1 is obtained; r (·,) is a least squares function; vector I ═ I (I, I-1,.., I-3)TI is more than or equal to 1, representing the number vector of four continuous measurements; vector H ═ YH(i)，YH(i-1)，...，YH(i-3))TAnd i is more than or equal to 1, which represents the HDOP difference vector of the BDS and the GPS measured four times. To avoid interference of drifting GNSS signals with algorithm spoofing, when YH(i)Taking 0.4 when the content is more than or equal to 0.4; y isH(i)When the content is less than or equal to 0, 0 is selected.
For YH(i)The method comprises the following steps:
YH(i)＝ΔHDOPB(i)-ΔHDOPG(i)；
ΔHDOPG(i)＝λG(i)-λo；
ΔHDOPB(i)＝λo-λB(i)；
in the formula, λG(k)HDOP for GPS signal at time k; lambda [ alpha ]B(k)HDOP which is the BDS signal at time k; lambda [ alpha ]OThe HDOP, O (optimal) of the currently preferred satellite navigation system is optimal.
And finally, establishing an HDOP self-adaptive optimization model:
Figure BDA0002045739550000041
Figure BDA0002045739550000042
Figure BDA0002045739550000043
wherein S isiThe signal is actually selected for the ith time; w is a1Is the weight of the GPS signal, w2Is the weight of the BDS signal; argmin (E) is the minimum of vector E; vector E ═ H, YH(i+1))T。
The HDOP adaptive optimization model algorithm decides to use the GPS or BDS working system and substitute its GNSS data into the equivalent surrogate nonlinear system model of step S2.
Preferably, the method further comprises: by utilizing the nonlinear transmission of the UT transformation processing mean value and the covariance, the UKF filtering algorithm of the system after the UT transformation comprises the following processes: sigma point sampling, time updating, observation updating and noise statistic estimation.
And selecting the Sigma points, and adjusting the sampling frequency of the Sigma points according to the complexity of the traffic road section within a certain time range. For example, when a vehicle passes through a traffic node such as a crossroad, a bus station, a sidewalk and the like, Sigma points are sampled doubly.
The invention discloses a self-adaptive Unscented Kalman filtering method of a GPS and BDS dual-mode satellite navigation system, which combines an HDOP self-adaptive optimization algorithm to select GNSS system data with better positioning precision, substitutes the GNSS system data into a nonlinear system model, processes the nonlinear system by using an Unscented transformation method and adopts an Unscented Kalman Filter (UKF). Due to the introduction of the HDOP self-adaptive optimization algorithm, the switching frequency of a GNSS working system is reduced, the robustness, stability and reliability of the GPS and BDS dual-mode satellite navigation system are improved, the deception interference of drifting GNSS signals to the system is avoided, and meanwhile, the error rate of the GPS and BDS dual-mode satellite navigation system is reduced. The self-adaptive unscented Kalman filtering method of the GPS and BDS dual-mode satellite navigation system can obviously improve the precision of vehicle integrated navigation and has good effect when the model nonlinearity and the noise statistical characteristic of the system are uncertain.
Drawings
FIG. 1 is a flow chart of an adaptive unscented Kalman filtering method for a GPS and BDS dual-mode satellite navigation system;
FIG. 2 is a graph of a GPS/BDS vehicle terminal filtering test experiment;
FIG. 3 is a graph of an AKF east and north position test experiment;
FIG. 4 is a graph of AKF east and north speed error testing experiments;
FIG. 5 is a UKF east and north position error test experiment graph;
FIG. 6 is a UKF east and north speed error test experiment graph;
FIG. 7 is a graph of an experiment for the error test of the east and north positions of AUKF;
FIG. 8 is a graph of AUKF east and north velocity error test experiments;
FIG. 9 is a graph of a position error test experiment before and after AUKF filtering;
FIG. 10 is a graph of an east position comparison experiment for each filter;
FIG. 11 is a graph of north position comparison experiments for various filters;
FIG. 12 is a graph of east speed comparison experiments for each filter;
fig. 13 is a graph of north speed comparison experiment of each filter.
Detailed Description
The invention is described in more detail below with reference to the figures and examples.
The invention discloses a self-adaptive unscented Kalman filtering method of a GPS and BDS dual-mode satellite navigation system, please refer to FIG. 1, the method comprises the following steps:
and step S1, establishing a nonlinear system model. The nonlinear system was modeled as:
Figure BDA0002045739550000061
in the formula, xkIs the state vector at time k, zkIs the observation vector at time k, fk(. and h)k(. respectively) State function and Observation function of the nonlinear System, wkAnd vkThe two white Gaussian noises are uncorrelated and the mean value and the covariance matrix of the two white Gaussian noises respectively satisfy the following conditions:
Figure BDA0002045739550000062
in the formula, q is a state noise mean value, and r is an observation noise mean value;
since a land vehicle can be approximated as a two-dimensional plane motion carrier, motion in a two-dimensional plane can be decomposed into two mutually perpendicular one-dimensional motions, such as motion in the east-west direction and motion in the north-south direction. The motion in two directions has no correlation, so that the one-dimensional motion in two directions can be respectively subjected to Kalman filtering;
and step S2, the Newton motion law is discrete and equivalently substituted for the nonlinear system model. The state vectors are respectively taken as xe＝[ve，ae]T，xn＝[vn，an]T，xeAnd xnAre respectively eastHeading and northbound speeds, aeAnd anThe two state vectors are written together as:
x＝[v，a]T；
in the following formula, the sampling interval is set to T, vk-1To observe the instantaneous speed, v, at the starting momentkInstantaneous speed after T time, ak-1The acceleration in T time is obtained by discretizing the Newton's law of motion:
Figure BDA0002045739550000071
vk＝vk-1+ak-1T；
providing east and north positions and velocities by using output sentences of a GPS and BDS dual-mode receiver respectively, and taking z as [ x, v ] as an observation vector]TWherein the initial state quantity x of the system0Is independent of both state noise and observation noise, and the initial state follows a Gaussian normal distribution which
Figure BDA0002045739550000073
A priori mean sum P0The covariance matrices are:
Figure BDA0002045739550000072
let mu letk＝wk-q，ηk＝vkR, which is substituted into the nonlinear system model to obtain:
Figure BDA0002045739550000081
in the formula, muk-1And ηkAre all white gaussian noise with a mean of zero and variance of Q and R, respectively.
And step S3, carrying out HDOP self-adaptive optimization model algorithm. And extracting a horizontal component precision factor (HDOP) in an output statement of the GPS and BDS dual-mode receiver, wherein the HDOP determines the reliability of the positioning precision and is used as an input value of an HDOP self-adaptive optimization algorithm.
First, an HDOP prediction model is built:
YH(i+1)＝k(i+1)+b
[k，b]T＝r(I，H)
wherein, YH(i+1)The predicted value of the HDOP difference value of the BDS and the GPS at the moment of i +1 is obtained; r (·,) is a least squares function; vector I ═ I (I, I-1,.., I-3)TI is more than or equal to 1, representing the number vector of four continuous measurements; vector H ═ YH(i)，YH(i-1)，...，YH(i-3))TAnd i is more than or equal to 1, which represents the HDOP difference vector of the BDS and the GPS measured four times. To avoid interference of drifting GNSS signals with algorithm spoofing, when YH(i)Taking 0.4 when the content is more than or equal to 0.4; y isH(i)When the content is less than or equal to 0, 0 is selected.
For YH(i)The method comprises the following steps:
YH(i)＝ΔHDOPB(i)-ΔHDOPG(i)；
ΔHDOPG(i)＝λG(i)-λo；
ΔHDOPB(i)＝λo-λB(i)；
in the formula, λG(k)HDOP for GPS signal at time k; lambda [ alpha ]B(k)HDOP which is the BDS signal at time k; lambda [ alpha ]OThe HDOP, O (optimal) of the currently preferred satellite navigation system is optimal.
And finally, establishing an HDOP self-adaptive optimization model:
Figure BDA0002045739550000082
Figure BDA0002045739550000091
Figure BDA0002045739550000092
wherein S isiThe signal is actually selected for the ith time; w is a1Is the weight of the GPS signal, w2Is the weight of the BDS signal; argmin (E) is the minimum of vector E; vector E ═ H, YH(i+1))T。
In the method, the GNSS system data with better positioning precision is selected by combining with the HDOP self-adaptive optimization algorithm and is substituted into the equivalent substitution nonlinear system model. Due to the introduction of the HDOP self-adaptive optimization algorithm, the switching frequency of a GNSS working system is reduced, the robustness, stability and reliability of the GPS and BDS dual-mode satellite navigation system are improved, the deception interference of drifting GNSS signals to the system is avoided, and meanwhile, the error rate of the GPS and BDS dual-mode satellite navigation system is reduced.
For the adaptive unscented kalman filter algorithm, in this embodiment, the adaptive kalman filter is used under the condition that the system mathematical model is uncertain or the statistical characteristics of the system state noise and the observation noise are uncertain, and performs filtering by adjusting the system mathematical model parameters and the noise statistical characteristic parameters in real time in the filtering process; the UKF mainly carries out filtering on a nonlinear system, but the conventional UKF algorithm has limitations, namely the prior statistical characteristics of system state noise and observation noise need to be accurately known, so that the traditional UKF cannot achieve the expected filtering effect on the nonlinear system with uncertain system models and uncertain system noise statistical characteristics, the system is unstable or even diverged, the adaptive Kalman filter and the UKF are combined to form the adaptive unscented Kalman filter, the noise statistical estimator is used for carrying out real-time estimation and correction on unknown or inaccurate noise statistics while carrying out filtering calculation, the UKF has the adaptive capacity of coping with noise change, and the UKF still has excellent stability and convergence under the condition of unknown or inaccurate noise statistics; and the noise statistics estimator is unbiased in its estimation of the noise statistics.
For the adaptive unscented kalman filter algorithm, in this embodiment, when the noise mean is non-zero, that is, under the condition of non-gaussian white noise, the statistical characteristics of the state noise and observation noise of the system are studied, the UKF algorithm is used to perform Sigma point sampling, and the state noise w is sampledk-1And measuring noiseSound vkThe proposed processing adopts AKF algorithm with noise change coping capability to output observation information to estimate and correct noise statistics in real time, namely W, due to uncertain statistical characteristics of state noise and observation noisek-1And VkThe mean and covariance matrix, and an adaptive unscented Kalman filtering algorithm.
In this embodiment, the method further includes: by utilizing the nonlinear transmission of the UT transformation processing mean value and the covariance, the UKF filtering algorithm of the system after the UT transformation comprises the following processes: sigma point sampling, time updating, observation updating and noise statistic estimation.
Furthermore, in the Sigma point sampling process, a proportional correction sampling strategy is adopted to ensure the semi-positive property (more than or equal to 0) of the covariance of the output variable. And selecting the Sigma points, and adjusting the sampling frequency of the Sigma points according to the complexity of the traffic road section within a certain time range. For example, when a vehicle passes through a traffic node such as a crossroad, a bus station, a sidewalk and the like, Sigma points are sampled doubly.
In the time updating step, the estimated value of the k-1 time is used
Figure BDA0002045739550000101
Covariance matrix P of errors at time k-1k-1/k-1Calculate Sigma Point ξi，k-1/k-1(i ═ 0, 1.., L), Sigma points are passed through a nonlinear state function fk(.) + q propagation becomes gammai，k/k-1Q is the system state noise used to calculate the one-step state prediction mean
Figure BDA0002045739550000102
Sum error covariance matrix Pk/k-1Namely:
γi，k/k-1＝fk-1(ξi，k/k-1)+q，i＝0，1，...，L；
Figure BDA0002045739550000103
Figure BDA0002045739550000111
wherein, Wi mAnd Wi cTo correspond to xiiWeight of (i ═ 0, 1., L), Wi mAnd Wi cRespectively calculating weight coefficients of first-order and second-order statistical characteristics; q is the covariance of the system state noise Q;
by using
Figure BDA0002045739550000112
And Pk/k-1And a proportional correction sampling strategy to calculate the Sigma point, xii，k/k-1(i ═ 0, 1.., L) by a non-linear observation function hk(. r) + r propagation as xi，k/k-1R is the covariance of the observed noise R, from which the output prediction mean is obtained
Figure BDA0002045739550000113
And covariance matrix
Figure BDA0002045739550000114
And
Figure BDA0002045739550000115
namely:
xi，k/k-1＝hk(ξi，k/k-1)+r，i＝0，1，...，L；
Figure BDA0002045739550000116
Figure BDA0002045739550000117
Figure BDA0002045739550000118
in the observation updating step, after a new observed quantity is obtained, actual filtering updating is carried out by using a conventional Kalman filter recursion formula, and a basic equation is as follows:
Figure BDA0002045739550000119
Figure BDA00020457395500001110
Figure BDA00020457395500001111
in the formula, KkIn order to filter the gain matrix of the filter,
Figure BDA0002045739550000121
the noise statistics estimating step is implemented based on a noise statistics estimator:
setting W to all obey normal distributionk-1And VkAre independent of each other, derive the sub-optimal maximum prior noise statistics estimator applied to UKF according to the maximum likelihood probability estimation criterion, when Q, Q, R, R are unknown, it and the state quantity x0，x1，...，xkMaximum likelihood probability estimation of
Figure BDA0002045739550000122
And
Figure BDA0002045739550000123
obtaining by using the maximization density;
the recursion formula of the secondary unbiased maximum prior noise statistic estimator is as follows:
the suboptimal maximum prior unbiased estimation of the system state noise q is as follows:
Figure BDA0002045739550000124
the suboptimal maximum prior unbiased estimation of the system state noise covariance Q is:
Figure BDA0002045739550000125
the suboptimal maximum prior unbiased estimation of the observation noise r is as follows:
Figure BDA0002045739550000126
the suboptimal maximum prior unbiased estimation of the observed noise covariance R is:
Figure BDA0002045739550000127
after the state noise statistical characteristic and the observation noise statistical characteristic of the UKF in the filtering process are determined, the unscented transformation is carried out on the mathematical model of the system by using the UKF algorithm, and the recursive process of the filtering algorithm of the self-adaptive unscented Kalman filter based on the noise statistical estimator is obtained:
the prior prediction process is as follows:
Figure BDA0002045739550000131
in the formula (I), the compound is shown in the specification,
Figure BDA0002045739550000132
is a priori estimated value, Wi mIs sigma point xii，k-1/k-1Mean value weighting parameters;
using a priori estimates
Figure BDA0002045739550000133
Sum error covariance matrix Pk-1|k-1To calculate the sigma point xii，k-1/k-1(i ═ 0, 1.. times, L), the sigma point is passed through a nonlinear state function fk(.) + q becomes gammai，k/k-1Then there is γi，k/k-1＝fk-1(ξi，k-1/k-1) + q and the error covariance matrix is obtained:
Figure BDA0002045739550000134
the posterior estimation process is as follows:
Figure BDA0002045739550000135
in the formula (I), the compound is shown in the specification,
Figure BDA0002045739550000136
for a posteriori estimate, zkIn order to be a weighted measurement variable,
Figure BDA0002045739550000137
to predict value, zkAnd
Figure BDA0002045739550000138
the difference of (a) is the residue of the measurement process, KkA gain matrix that is a residue;
the gain calculation process is as follows:
Figure BDA0002045739550000139
in the formula, residual gain KkFor minimizing the a posteriori estimation error covariance,
Figure BDA0002045739550000141
for a priori estimation
Figure BDA0002045739550000142
Covariance of error Pk/k-1，
Figure BDA0002045739550000143
For a posteriori estimation
Figure BDA0002045739550000144
Covariance of error Pk/k；
The state covariance matrix updating process comprises the following steps:
Figure BDA0002045739550000145
in the formula, Pk/kFor a posteriori estimation of error covariance matrix, Pk/k-1For the prior estimation of the error covariance matrix,
Figure BDA0002045739550000146
is the residual error covariance.
Based on the above formulas, combining with an HDOP adaptive optimization algorithm, GNSS system data with better positioning accuracy is selected and substituted into a nonlinear system model to be used as an input source of an AUKF (adaptive unknown Kalman Filter) filter. Simultaneously, the nonlinear system is processed by an Unscented transformation method, and an Unscented Kalman Filter (Unscented Kalman Filter, Unscented kf) is adopted. The method can solve the nonlinear characteristic of a GPS and BDS dual-mode satellite navigation system, meanwhile, a noise statistics estimator is used for estimating and correcting unknown or inaccurate noise statistics in real time, compared with single Adaptive Kalman Filtering (AKF) and UKF, the filtering algorithm has the Adaptive capacity of coping with noise change, the filtering algorithm still has excellent stability and convergence under the condition that the noise statistics is unknown or inaccurate, and the noise statistics estimator estimates the noise statistics unbiased. Due to the introduction of the HDOP self-adaptive optimization algorithm, the switching frequency of a GNSS working system is reduced, the robustness, stability and reliability of the GPS and BDS dual-mode satellite navigation system are improved, the deception interference of drifting GNSS signals to the system is avoided, and meanwhile, the error rate of the GPS and BDS dual-mode satellite navigation system is reduced. In the UKF process, the nonlinear system does not need to be linearized, and the linear error introduced in the linearization process of the nonlinear system is avoided; the unscented transformation adopted by the UKF can completely reproduce the mean value and covariance of the state quantity of the nonlinear system; the filtering precision of any nonlinear system can reach more than three orders of Taylor series expansion; the calculation of the Jacobian matrix is not needed, calculation errors are not introduced, and the calculation amount of the UKF is greatly simplified.
In the actual test process, the present embodiment performs test research on the adaptive unscented kalman filter algorithm, assuming that the vehicle makes an accelerated motion at an initial speed of 10m/s, the initial parameter is selected to be the observation point N of 40, the sampling period T of 1s, and the covariance matrix of the satellite system observation noise is [20 [2，202]The comparison result of the filtering test experiment of the GPS and BDS dual-mode terminal is shown in fig. 2.
The simulated carrier in the experiment of the embodiment is a car, the running environment of the car is selected in the section with dense population vehicles, high buildings and tunnels, and in addition, the car does not encounter an emergency in the running process; the GPS and BDS dual-mode vehicle-mounted terminal of the embodiment is applied to carry out experiments, and the dual-mode satellite navigation system respectively selects an information fusion algorithm of AKF, UKF and Adaptive Unscented Kalman (AUKF) to carry out performance comparison. The mathematical model for AKF is known and is a linear system model, whereas the mathematical models for UKF and AUKF are typical non-linear systems. Due to the influence of factors such as the device and the external environment, state noise and observation noise in the system model are unknown in the running process of the vehicle, the statistical parameters of the state noise and the observation noise of the system need to be adjusted and estimated in real time, and the obtained statistical characteristics of the noise and the system mathematical model are utilized to filter the combined navigation system. The experiment for testing the position error (east error, north error, and because the automobile runs on the earth surface, it can be approximately considered that the automobile does not move in the vertical direction and neglects the elevation error) and the speed error by using the AKF, the UKF and the AUKF is shown in fig. 3 to fig. 9.
From the experimental results of fig. 3 to 9, it can be known that the position error and the speed error of the vehicle in the process of traveling after the UKF filtering processing are both smaller than those after the conventional AKF filtering processing, and the phenomenon of data divergence exists when the AKF filtering processing reaches a certain time. In contrast, comparative experiments of the position error and the velocity error of AKF, UKF and AUKF are shown in fig. 10 to 13. From the experimental results of fig. 10 to fig. 13, it can be known that, in the vehicle running process, the AUKF is significantly better than the AKF, and even if the system model is nonlinear, the final position error and speed error are smaller than the errors of the AKF and the UKF, which indicates that the AUKF algorithm can greatly improve the precision of the vehicle combination navigation and has a good effect when processing the system model nonlinearity and the uncertain noise statistical characteristics.
The above description is only for the purpose of illustrating the preferred embodiments of the present invention and is not to be construed as limiting the invention, and any modifications, equivalents or improvements made within the technical scope of the present invention should be included in the scope of the present invention.

Claims (7)

1. An adaptive unscented Kalman filtering method for a GPS and BDS dual-mode satellite navigation system is characterized by comprising the following steps:
step S1, establishing a nonlinear system model, and setting the nonlinear system model as:
Figure FDA0002942222110000011
in the formula, xkIs the state vector at time k, zkIs the observation vector at time k, fk(. and h)k(. respectively) State function and Observation function of the nonlinear System, wkAnd vkThe two white Gaussian noises are uncorrelated and the mean value and the covariance matrix of the two white Gaussian noises respectively satisfy the following conditions:
Figure FDA0002942222110000012
in the formula, q is a state noise mean value, and r is an observation noise mean value;
s2, carrying out discretization equivalent substitution on a nonlinear system model by using a Newton motion law; the state vectors are respectively taken as xe＝[Ve，ae]T，xn＝[Vn，an]T，xeAnd xnEast and north speed, respectively, aeAnd anThe two state vectors are written together as:
x＝[v，a]T；
in the following formula, the sampling interval is set to T, vk-1To observe the instantaneous speed, v, at the starting momentkInstantaneous speed after T time, ak-1The acceleration in T time is obtained by discretizing the Newton's law of motion:
Figure FDA0002942222110000013
vk＝vk-1+ak-1T；
the output sentences of the GPS dual-mode receiver and the BDS dual-mode receiver are used for respectively providing the position and the speed of the east direction and the north direction, and the observation direction is measured as z ═ x, v]TWherein the initial state quantity x of the system0Is independent of both state noise and observation noise, and the initial state follows a Gaussian normal distribution which
Figure FDA0002942222110000014
A priori mean sum P0The covariance matrices are:
Figure FDA0002942222110000021
let mu letk＝wk-q，ηk＝vkR, which is substituted into the nonlinear system model to obtain:
Figure FDA0002942222110000022
in the formula, muk-1And ηkThe white Gaussian noises are all white Gaussian noises with the mean value of zero and the variance of Q and R respectively;
step S3, extracting a horizontal component precision factor HDOP in an output statement of the GPS and BDS dual-mode receiver by the HDOP self-adaptive optimization model algorithm, wherein the HDOP determines the reliability of the positioning precision and is used as an input value of the HDOP self-adaptive optimization algorithm;
first, an HDOP prediction model is built:
YH(i+1)＝k(i+1)+b
[k，b]T＝r(I，H)
wherein, YH(i+1)The predicted value of the HDOP difference value of the BDS and the GPS at the moment of i +1 is obtained; r (·,) is a least squares function; vector I ═ I (I, I + 1., I +3)TI is more than or equal to 1, representing the number vector of four continuous measurements; vector H ═ YH(i)，YH(i+1)，...，YH(i+3))TWhen i is more than or equal to 1, the HDOP difference vector of BDS and GPS is measured for four times, when Y isH(i)Taking 0.4 when the content is more than or equal to 0.4; y isH(i)When the content is less than or equal to 0, taking 0;
for YH(i)The method comprises the following steps:
YH(i)＝ΔHDOPB(i)-ΔHDOPG(i)；
ΔHDOPG(i)＝λG(i)-λo；
ΔHDOPB(i)＝λo-λB(i)；
in the formula, λG(k)HDOP for GPS signal at time k; lambda [ alpha ]B(k)HDOP which is the BDS signal at time k; lambda [ alpha ]OThe HDOP, O (optimal) of the currently preferred satellite navigation system is optimal;
and finally, establishing an HDOP self-adaptive optimization model:
Figure FDA0002942222110000031
Figure FDA0002942222110000032
Figure FDA0002942222110000033
wherein S isiThe signal is actually selected for the ith time; w is a1Being GPS signalsWeight, w2Is the weight of the BDS signal; argmin (E) is the minimum of vector E; vector E ═ H, YH(i+1))T；
The HDOP adaptive optimization model algorithm decides to use the GPS or BDS working system and substitute its GNSS data into the equivalent surrogate nonlinear system model of step S2.
2. The adaptive unscented kalman filter method of GPS and BDS dual-mode satellite navigation system of claim 1, wherein the HDOP adaptive optimization model algorithm selects GNSS system data with better positioning accuracy and substitutes it into an equivalent surrogate nonlinear system model.
3. The adaptive unscented kalman filter method of a dual mode GPS and BDS satellite navigation system of claim 1, wherein the method further comprises: by utilizing the nonlinear transmission of the UT transformation processing mean value and the covariance, the UKF filtering algorithm of the system after the UT transformation comprises the following processes: sigma point sampling, time updating, observation updating and noise statistic estimation.
4. The adaptive unscented kalman filter method of GPS and BDS dual mode satellite navigation system of claim 3, wherein in the Sigma point sampling process, a proportional modified sampling strategy is employed to ensure the semi-positivity of the covariance of the output variable, the Sigma point is selected, and the sampling frequency of the Sigma point is adjusted according to the complexity of the traffic segment within a certain time range.
5. The adaptive unscented kalman filter method for dual mode GPS and BDS satellite navigation system as claimed in claim 3, wherein the time update step uses the estimated value of time k-1
Figure FDA0002942222110000034
Covariance matrix P of errors at time k-1k-1/k-1Calculate Sigma Point ξi，k-1/k-1(i ═ 0, 1.., L), Sigma points pass through a nonlinear state functionNumber fk(.) + q propagation becomes gammai，k/k-1Q is the system state noise used to calculate the one-step state prediction mean
Figure FDA0002942222110000041
Sum error covariance matrix Pk/k-1Namely:
γi，k/k-1＝fk-1(ξi，k/k-1)+q，i＝0，1，...，L；
Figure FDA0002942222110000042
Figure FDA0002942222110000043
wherein, Wi mAnd Wi cTo correspond to xiiWeight of (i ═ O, 1., L), Wi mAnd Wi cRespectively calculating weight coefficients of first-order and second-order statistical characteristics; q is the covariance of the system state noise Q;
by using
Figure FDA0002942222110000044
And Pk/k-1And a proportional correction sampling strategy to calculate the Sigma point, xii，k/k-1(i ═ 0, 1.., L) by a non-linear observation function hk(. r) + r propagation as xi，k/k-1R is the covariance of the observed noise R, from which the output prediction mean is obtained
Figure FDA0002942222110000045
And covariance matrix
Figure FDA0002942222110000046
And
Figure FDA0002942222110000047
namely:
xi，k/k-1＝hk(ξi，k/k-1)+r，i＝0，1，...，L；
Figure FDA0002942222110000048
Figure FDA0002942222110000049
Figure FDA00029422221100000410
6. the adaptive unscented kalman filter method of a GPS and BDS dual-mode satellite navigation system as claimed in claim 3, wherein in the observation update step, after obtaining a new observation quantity, the actual filtering update is performed using a conventional kalman filter recursion formula, and the basic equation is as follows:
Figure FDA0002942222110000051
Figure FDA0002942222110000052
Figure FDA0002942222110000053
in the formula, KkIn order to filter the gain matrix of the filter,
Figure FDA0002942222110000054
7. the adaptive unscented kalman filter method of a dual mode GPS and BDS satellite navigation system of claim 3, wherein the noise statistics estimating step is based on a noise statistics estimator:
setting W to all obey normal distributionk-1And VkAre independent of each other, derive the sub-optimal maximum prior noise statistics estimator applied to UKF according to the maximum likelihood probability estimation criterion, when Q, Q, R, R are unknown, it and the state quantity x0，x1，...，xkMaximum likelihood probability estimation of
Figure FDA0002942222110000055
And
Figure FDA0002942222110000056
obtaining by using the maximization density;
the recursion formula of the secondary unbiased maximum prior noise statistic estimator is as follows:
the suboptimal maximum prior unbiased estimation of the system state noise q is as follows:
Figure FDA0002942222110000057
the suboptimal maximum prior unbiased estimation of the system state noise covariance Q is:
Figure FDA0002942222110000058
the suboptimal maximum prior unbiased estimation of the observation noise r is as follows:
Figure FDA0002942222110000059
the suboptimal maximum prior unbiased estimation of the observed noise covariance R is:
Figure FDA00029422221100000510
after the state noise statistical characteristic and the observation noise statistical characteristic of the UKF in the filtering process are determined, the unscented transformation is carried out on the mathematical model of the system by using the UKF algorithm, and the recursive process of the filtering algorithm of the self-adaptive unscented Kalman filter based on the noise statistical estimator is obtained:
the prior prediction process is as follows:
Figure FDA0002942222110000061
in the formula (I), the compound is shown in the specification,
Figure FDA0002942222110000062
is a priori estimated value, Wi mIs sigma point xii，k-1/k-1The weight used for mean weighting;
using a priori estimates
Figure FDA0002942222110000063
Sum error covariance matrix Pk-1|k-1To calculate the sigma point xii，k-1/k-1(i ═ 0, 1.. times, L), the sigma point is passed through a nonlinear state function fk(.) + q becomes gammai，k/k-1Then there is γi，k/k-1＝fk-1(ξi，k-1/k-1) + q and the error covariance matrix is obtained:
Figure FDA0002942222110000064
the posterior estimation process is as follows:
Figure FDA0002942222110000065
in the formula (I), the compound is shown in the specification,
Figure FDA00029422221100000616
for a posteriori estimate, zkIn order to be a weighted measurement variable,
Figure FDA0002942222110000067
to predict value, zkAnd
Figure FDA0002942222110000068
the difference of (a) is the residue of the measurement process, KkA gain matrix that is a residue;
the gain calculation process is as follows:
Figure FDA0002942222110000069
in the formula, residual gain KkFor minimizing the a posteriori estimation error covariance,
Figure FDA00029422221100000610
for a priori estimation
Figure FDA00029422221100000611
Covariance of error Pk/k-1，
Figure FDA00029422221100000612
For a posteriori estimation
Figure FDA00029422221100000613
Covariance of error Pk/k；
The state covariance matrix updating process comprises the following steps:
Figure FDA00029422221100000614
in the formula, Pk/kFor a posteriori estimation of error covariance matrix, Pk/k-1For the prior estimation of the error covariance matrix,
Figure FDA00029422221100000615
is the residual error covariance.
CN201910357048.0A 2019-04-29 2019-04-29 Self-adaptive unscented Kalman filtering method of GPS and BDS dual-mode satellite navigation system Active CN110231636B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201910357048.0A CN110231636B (en) 2019-04-29 2019-04-29 Self-adaptive unscented Kalman filtering method of GPS and BDS dual-mode satellite navigation system

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201910357048.0A CN110231636B (en) 2019-04-29 2019-04-29 Self-adaptive unscented Kalman filtering method of GPS and BDS dual-mode satellite navigation system

Publications (2)

Publication Number Publication Date
CN110231636A CN110231636A (en) 2019-09-13
CN110231636B true CN110231636B (en) 2021-03-26

Family

ID=67860975

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201910357048.0A Active CN110231636B (en) 2019-04-29 2019-04-29 Self-adaptive unscented Kalman filtering method of GPS and BDS dual-mode satellite navigation system

Country Status (1)

Country Link
CN (1) CN110231636B (en)

Families Citing this family (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110972071A (en) * 2019-10-29 2020-04-07 福建星网智慧软件有限公司 Multi-mode positioning method, positioning server, terminal and storage medium
EP3828583A1 (en) 2019-11-27 2021-06-02 Honda Research Institute Europe GmbH Analysis of localization errors in a mobile object
CN111323793B (en) * 2020-03-30 2021-02-05 中国矿业大学 GNSS pseudo-range single-point positioning state domain integrity monitoring method
CN113075712B (en) * 2021-03-17 2023-08-11 北京云恒科技研究院有限公司 Autonomous controllable multi-system high-precision navigation equipment
CN113407909B (en) * 2021-07-15 2024-01-09 东南大学 Calculation method of odorless algorithm for non-analytic complex nonlinear system
CN113709662B (en) * 2021-08-05 2023-12-01 北京理工大学重庆创新中心 Autonomous three-dimensional inversion positioning method based on ultra-wideband
CN113791432A (en) * 2021-08-18 2021-12-14 哈尔滨工程大学 Integrated navigation positioning method for improving AIS positioning accuracy
CN114348002A (en) * 2021-12-29 2022-04-15 江苏大学 System and method for estimating speed of hub motor driven electric automobile

Citations (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20050251328A1 (en) * 2004-04-05 2005-11-10 Merwe Rudolph V D Navigation system applications of sigma-point Kalman filters for nonlinear estimation and sensor fusion
CN102176031A (en) * 2011-01-06 2011-09-07 中国科学院国家授时中心 System time difference based receiver completeness failure detection method in dual-mode navigation system
EP2400267A1 (en) * 2010-06-25 2011-12-28 Thales Navigation filter for navigation system by terrain matching.
US20120162006A1 (en) * 2010-12-28 2012-06-28 Stmicroelectronics S.R.L. Cinematic parameter computing method of a satellite navigation system and receiving apparatus employing the method
CN103411616A (en) * 2013-07-05 2013-11-27 东南大学 Vehicle-mounted integrated navigation method based on simplified inertia measurement assembly
CN103592657A (en) * 2013-09-30 2014-02-19 北京大学 Method for realizing single-mode RAIM (Receiver Autonomous Integrity Monitoring) under small number of visible satellites based on assistance of clock correction
CN104966126A (en) * 2015-05-18 2015-10-07 深圳市联和安业科技有限公司 Smart taxi-taking system
CN105785412A (en) * 2016-03-03 2016-07-20 东南大学 Vehicle rapid optimizing satellite selection positioning method based on GPS and Beidou double constellations
CN106646543A (en) * 2016-12-22 2017-05-10 成都正扬博创电子技术有限公司 High-dynamic satellite navigation signal carrier tracking method based on master-slave AUKF algorithm
CN106772524A (en) * 2016-11-25 2017-05-31 安徽科技学院 A kind of agricultural robot integrated navigation information fusion method based on order filtering

Family Cites Families (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US6407701B2 (en) * 2000-03-24 2002-06-18 Clarion Co., Ltd. GPS receiver capable of calculating accurate 2DRMS
CN101976290B (en) * 2010-11-01 2012-09-05 北京航空航天大学 Navigation constellation optimization design platform and method based on decomposition thought and particle swarm fusion method
CN109631913A (en) * 2019-01-30 2019-04-16 西安电子科技大学 X-ray pulsar navigation localization method and system based on nonlinear prediction strong tracking Unscented kalman filtering

Patent Citations (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20050251328A1 (en) * 2004-04-05 2005-11-10 Merwe Rudolph V D Navigation system applications of sigma-point Kalman filters for nonlinear estimation and sensor fusion
EP2400267A1 (en) * 2010-06-25 2011-12-28 Thales Navigation filter for navigation system by terrain matching.
US20120162006A1 (en) * 2010-12-28 2012-06-28 Stmicroelectronics S.R.L. Cinematic parameter computing method of a satellite navigation system and receiving apparatus employing the method
CN102176031A (en) * 2011-01-06 2011-09-07 中国科学院国家授时中心 System time difference based receiver completeness failure detection method in dual-mode navigation system
CN103411616A (en) * 2013-07-05 2013-11-27 东南大学 Vehicle-mounted integrated navigation method based on simplified inertia measurement assembly
CN103592657A (en) * 2013-09-30 2014-02-19 北京大学 Method for realizing single-mode RAIM (Receiver Autonomous Integrity Monitoring) under small number of visible satellites based on assistance of clock correction
CN104966126A (en) * 2015-05-18 2015-10-07 深圳市联和安业科技有限公司 Smart taxi-taking system
CN105785412A (en) * 2016-03-03 2016-07-20 东南大学 Vehicle rapid optimizing satellite selection positioning method based on GPS and Beidou double constellations
CN106772524A (en) * 2016-11-25 2017-05-31 安徽科技学院 A kind of agricultural robot integrated navigation information fusion method based on order filtering
CN106646543A (en) * 2016-12-22 2017-05-10 成都正扬博创电子技术有限公司 High-dynamic satellite navigation signal carrier tracking method based on master-slave AUKF algorithm

Non-Patent Citations (7)

* Cited by examiner, † Cited by third party
Title
A new fast satellite selection algorithm for BDS-GPS receivers;Fanchen Meng 等;《SiPS 2013 Proceedings》;20131202;第371-376页 *
BDS and GPS stand-alone and integrated attitude dilution of precision definition and comparison;Yong Wang 等;《ScienceDirect》;20171202;第2972-2981页 *
一种顾及加权水平精度因子的室内宽带定位算法;尚俊娜 等;《中国惯性技术学报》;20181231;第26卷(第6期);第792-798页 *
基于自适应EKF的BDS_GPS精密单点定位方法;赵琳 等;《系统工程与电子技术》;20160930;第38卷(第9期);第2142-2148页 *
多星座组合导航自适应信息融合滤波算法;吴玲 等;《航天控制》;20101231;第28卷(第6期);第38-42、62页 *
车载组合导航系统自适应无迹卡尔曼滤波算法研究;唐苗苗;《中国优秀硕士学位论文全文数据库 工程科技Ⅱ辑》;20140415;第1-92页 *
陆基导航定位系统中的HDOP模型与计算方法;白敏;《数字通信》;20100425;第51-54页 *

Also Published As

Publication number Publication date
CN110231636A (en) 2019-09-13

Similar Documents

Publication Publication Date Title
CN110231636B (en) Self-adaptive unscented Kalman filtering method of GPS and BDS dual-mode satellite navigation system
CN109459019B (en) Vehicle navigation calculation method based on cascade adaptive robust federal filtering
CN111985093B (en) Adaptive unscented Kalman filtering state estimation method with noise estimator
CN109459040B (en) Multi-AUV (autonomous Underwater vehicle) cooperative positioning method based on RBF (radial basis function) neural network assisted volume Kalman filtering
CN110779518B (en) Underwater vehicle single beacon positioning method with global convergence
Almagbile et al. Evaluating the performances of adaptive Kalman filter methods in GPS/INS integration
Li et al. A novel hybrid fusion algorithm for low-cost GPS/INS integrated navigation system during GPS outages
CN110794409B (en) Underwater single beacon positioning method capable of estimating unknown effective sound velocity
CN109724599A (en) A kind of Robust Kalman Filter SINS/DVL Combinated navigation method of anti-outlier
CN110823217A (en) Integrated navigation fault-tolerant method based on self-adaptive federal strong tracking filtering
CN109059911B (en) Data fusion method of GNSS, INS and barometer
CN110749891B (en) Self-adaptive underwater single beacon positioning method capable of estimating unknown effective sound velocity
CN110779519B (en) Underwater vehicle single beacon positioning method with global convergence
CN108983271B (en) Train combined positioning method based on RTK-GPS/INS
CN108134640B (en) Cooperative positioning system and method based on node motion state constraint
CN105180938A (en) Particle filter-based gravity sampling vector matching positioning method
CN103776453A (en) Combination navigation filtering method of multi-model underwater vehicle
CN110646783B (en) Underwater beacon positioning method of underwater vehicle
CN105157704A (en) Bayesian estimation-based particle filter gravity-assisted inertial navigation matching method
CN110031798A (en) A kind of indoor objects tracking based on simplified Sage-Husa adaptive-filtering
CN103529461A (en) Receiver quick positioning method based on strong tracking filtering and Hermite interpolation method
CN109507706B (en) GPS signal loss prediction positioning method
CN112577496A (en) Multi-source fusion positioning method based on self-adaptive option
CN110989341B (en) Constraint auxiliary particle filtering method and target tracking method
CN113324546A (en) Multi-underwater vehicle collaborative positioning self-adaptive adjustment robust filtering method under compass failure

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