CN115479615A - Random error online estimation algorithm for vehicle-mounted inertial device - Google Patents

Random error online estimation algorithm for vehicle-mounted inertial device Download PDF

Info

Publication number
CN115479615A
CN115479615A CN202211117015.7A CN202211117015A CN115479615A CN 115479615 A CN115479615 A CN 115479615A CN 202211117015 A CN202211117015 A CN 202211117015A CN 115479615 A CN115479615 A CN 115479615A
Authority
CN
China
Prior art keywords
random
random error
parking
data
vehicle
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.)
Pending
Application number
CN202211117015.7A
Other languages
Chinese (zh)
Inventor
赵龙
赵泺棣
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.)
Beihang University
Original Assignee
Beihang 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 Beihang University filed Critical Beihang University
Priority to CN202211117015.7A priority Critical patent/CN115479615A/en
Publication of CN115479615A publication Critical patent/CN115479615A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • GPHYSICS
    • G01MEASURING; TESTING
    • G01CMEASURING DISTANCES, LEVELS OR BEARINGS; SURVEYING; NAVIGATION; GYROSCOPIC INSTRUMENTS; PHOTOGRAMMETRY OR VIDEOGRAMMETRY
    • G01C25/00Manufacturing, calibrating, cleaning, or repairing instruments or devices referred to in the other groups of this subclass
    • G01C25/005Manufacturing, calibrating, cleaning, or repairing instruments or devices referred to in the other groups of this subclass initial alignment, calibration or starting-up of inertial devices

Landscapes

  • Engineering & Computer Science (AREA)
  • Manufacturing & Machinery (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Radar, Positioning & Navigation (AREA)
  • Remote Sensing (AREA)
  • Navigation (AREA)

Abstract

The invention discloses an on-line estimation algorithm for random errors of a vehicle-mounted inertial device, which is used for parking detection according to data output by a vehicle-mounted inertial sensor, accumulating static inertial data when parking, removing IMU zero offset, and then performing on-line modeling and parameter estimation of random errors by adopting generalized wavelet moment estimation; when the vehicle runs, an INS/GNSS integrated navigation system is used for navigation positioning, IMU zero offset is estimated, a random noise model built during parking is used for compensating inertial output data, and a compensation result is sent to the navigation system for resolving; as the number of stops increases, the accuracy of the random noise model created by the algorithm increases. The method can overcome the limitation of the conventional inertial sensor calibration method, realize the online modeling of random noise and improve the positioning accuracy of the INS/GNSS integrated navigation system.

Description

Random error online estimation algorithm for vehicle-mounted inertial device
Technical Field
The invention relates to the technical field of inertial device calibration, in particular to an online modeling and parameter estimation algorithm for random errors of an inertial device.
Background
Estimating the position, velocity and attitude of moving objects such as vehicles, airplanes and ships in space in an accurate, continuous and reliable manner is an important issue for practical applications such as automatic driving, unmanned aerial vehicle operation and fine agriculture, and currently, implementing Navigation and positioning by a Global Navigation Satellite System (GNSS) is the most commonly used solution in urban environments. However, when moving objects are in a complex urban environment, the performance of such systems is severely reduced by the partial or complete unavailability of satellite signals due to overpass, tunnels or tree shadows, and for some specific requirements, the GNSS has a low satellite receiver bandwidth (typically below 10 Hz) and cannot fully rely on the navigation parameters provided by the GNSS. Furthermore, GNSS is unable to provide information about attitude.
Based on the above problems, one of the most common and widely used methods is to combine the GNSS and the Inertial Navigation System (INS), and because the GNSS and the INS have advantage complementarity, the combination of the GNSS and the INS can significantly improve the Navigation and positioning performance. When the GNSS signal is available, the combination of the INS and the GNSS is generally implemented by a bayesian technique, wherein common algorithms include standard kalman filtering, deformation of kalman filtering, robust adaptive filtering, and the like; when the GNSS signal quality is poor or completely fails, the INS performs solution in a recursive mode, i.e., the navigation state can be estimated completely independently of the GNSS. In both modes, the overall navigation performance depends to a large extent on the accuracy of the inertial signal, more precisely on the error of the inertial signal. These errors are integrated into the INS for integral calculations, the effect of which increases dramatically over time. In summary, accurate modeling and estimation of the error of the inertial signal is crucial to correct estimation and improve navigation performance quality.
At present, the traditional methods for modeling the error signals of the inertial device, such as the Allan variance method, the PSD analysis and other classical methods, have the problem that the errors can not be identified and separated in the spectral domain, and the specific analysis is as follows:
(1) The AV method is only applicable to noisy processes that can be clearly identified and separated in the spectral domain and are not affected by spectral ambiguity. Also, the AV method does not allow the parameters of the GM process to be read directly, since a larger value makes the process similar to WN, while a smaller value makes the process similar to RW. Therefore, the conventional AV method is limited to a model consisting of a process characterized by a linear region in a wavelet variance log diagram, and in most cases, the wavelet variance log diagram of the sensor cannot show the characteristic of a typical linear region due to interference of many influencing factors in data acquisition, so that it is difficult to actually use coefficient reading using the AV method in most cases.
(2) The calculation of empirical WV is simpler than the PSD analysis without parameters. For example, periodograms are inconsistent estimates of power spectral density functions, and even large sample volumes may be severely biased by frequency leakage effects. Therefore, more complex PSD estimators or smoothing techniques, such as pre-whitening or tapering, need to be employed to approach the consistency that the generalized wavelet moment estimates can provide;
(3) The problem of optimizing the difference between an empirical PSD-based and a model-based PSD is made more difficult to solve when the PSD has large variability over a very narrow frequency band.
Therefore, in combination with the above background, how to overcome the limitations of the conventional method and provide an inertial device random error online estimation algorithm is a problem that needs to be solved urgently by those in the navigation positioning field.
Disclosure of Invention
In view of the above, the invention provides an on-line estimation algorithm for random errors of a vehicle-mounted inertial device, which is used for solving the problems of on-line modeling and parameter estimation of the random errors of the inertial device in the driving process of a specific vehicle-mounted object.
In order to achieve the purpose, the invention adopts the following technical scheme:
an on-line estimation algorithm for random errors of a vehicle-mounted inertial device comprises the following steps:
s1, judging whether the vehicle is in a parking state or not according to IMU observation data output by a vehicle-mounted object,
if not, using the INS/GNSS integrated navigation system to perform navigation positioning, estimating the zero offset of the accelerometer of the IMU, compensating random noise, and sending the compensation result to the navigation system for resolving;
if yes, executing step S2;
s2, obtaining parking lot IMU observation data and accelerometer zero offset, accumulating the parking lot IMU observation data, and subtracting the accelerometer zero offset to obtain a random error component;
s3, performing parameter estimation on the random error component by utilizing generalized wavelet moment estimation, compensating random noise according to the parameter estimation, and sending a compensation result to a navigation system for resolving;
preferably, parking detection criteria are established based on static output data of the IMU,
the parking inspection criterion is as follows:
Figure BDA0003845718770000031
in the formula, A i For data output by the accelerometer at time i, U i Is the average value of the data in the fixed time window at the moment i, the number of the data in the fixed time window is N, T i The standard deviation of the data at the moment i and lambda is a detection threshold;
preferably, the verification threshold λ is dynamically adjusted by the following formula,
Figure BDA0003845718770000032
wherein k is the number of times of detecting a parking spot, and the initial threshold value is empirically selected as T i =0.01;
Preferably, the random error components are linearly combined by an independent random process, and the independent random process includes: white gaussian noise (WN), random Walk (RW), random Ramp (RR), quantization Noise (QN), and first-order autoregressive process (AR);
preferably, v (τ) for each of said independent stochastic processes is performed before step S3 is performed j )~τ j Carrying out log-log logarithmic processing on the curve to obtain a double-log curve, constructing a plurality of random error component candidate models according to the slope characteristic of the double-log curve, and carrying out parameter estimation on the plurality of random error component candidate models by utilizing generalized wavelet moment estimation;
preferably, a preference criterion is constructed, and optimal parameter estimation of the plurality of random error component candidate models is determined, where the preference criterion is:
Figure BDA0003845718770000033
in the formula, theta is a parameter to be estimated,
Figure BDA0003845718770000034
for the wavelet variance obtained from the observation sequence,
Figure BDA0003845718770000035
omega is a positive definite weight matrix which makes the formula convex, and is the wavelet variance calculated according to the model;
selecting a model with the minimum GOF value as an optimal model, wherein the parameter estimation corresponding to the optimal model is the optimal parameter estimation;
preferably, a plurality of candidate models of the random error components are constructed for the first section of parking data, and the parameter estimation is directly carried out by using the optimal model for the nth section of parking data, wherein n is more than or equal to 1;
preferably, the expression for performing parameter estimation on the random error component by using generalized wavelet moment estimation is as follows:
Figure BDA0003845718770000036
in the formula, theta is a parameter to be estimated,
Figure BDA0003845718770000041
phi (theta) is the wavelet variance obtained from the observation sequence, and omega is a positive weighting matrix which makes the formula convex;
in the formula, Ω is a positive weighting matrix making the formula convex.
According to the technical scheme, compared with the prior art, the invention discloses an on-line estimation algorithm for random errors of a vehicle-mounted inertial device, and the on-line parameter estimation algorithm adopts wavelet variance and a least square method to carry out on-line parameter estimation on the random errors of the inertial device so as to improve the real-time navigation positioning accuracy of a vehicle-mounted object;
another purpose of the invention is to construct a random error candidate model and a model preference criterion, and realize automatic shaping and parameter estimation of random noise so as to further improve the precision of parameter estimation;
still another objective of the present invention is to accumulate static data of parking status when obtaining random error components, and as the number of parking times increases, the amount of IMU static data increases, and modeling IMU random noise becomes more and more accurate;
the invention also aims to construct a candidate model only for the first section of parking data and traverse the candidate model, and when static data accumulation is carried out in a detected parking state, the candidate model with the optimal random error component can be directly used for parameter estimation, so that the estimation efficiency is improved.
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 schematic flow chart of an on-line estimation algorithm for random errors of a vehicle-mounted inertial device,
FIG. 2 is a graph showing v (. Tau.) of a typical stochastic process method j )~τ j A slope characteristic diagram presented by a log-log graph;
fig. 3 is a flow chart of an algorithm for automatically selecting an optimal model based on GMWM.
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.
The random error component of the low-cost inertial sensor is quite complex in structure, and the common random error analysis methods such as Allan variance method, PSD analysis and other classical methods have the problem that errors cannot be identified and separated in a spectral domain.
The Generalized Wavelet moment estimation Method (GMWM) associates Wavelet variance with PSD, represents a sequence to be estimated as a model formed by the sum of White gaussian Noise (WN), random Walk (RW), random Ramp (RR), quantization Noise (QN) and finite first-order Auto-regression (AR), the GMWM estimator formed by the model has progressive consistency, and then corresponds the Wavelet variance of the sequence to be estimated to the Wavelet variance implied by the hypothesis model, and estimates the parameter of the latter by using the Generalized least square Method to minimize the difference between the two. The method can effectively avoid the defects of the traditional method and has strong practical application value.
Specifically, the embodiment of the invention discloses an on-line estimation algorithm for random errors of a vehicle-mounted inertial device, which comprises the following steps as shown in fig. 1:
s1, judging whether the vehicle is in a parking state or not according to IMU observation data output by a vehicle-mounted object,
if not, using the INS/GNSS integrated navigation system to perform navigation positioning, estimating the zero offset of the accelerometer of the IMU, compensating random noise, and sending the compensation result to the navigation system for resolving;
if yes, executing step S2;
s2, obtaining parking lot IMU observation data and accelerometer zero offset, accumulating the parking lot IMU observation data, and subtracting the accelerometer zero offset to obtain a random error component;
and S3, performing parameter estimation on the random error component by using generalized wavelet moment estimation, compensating random noise according to the parameter estimation, and sending a compensation result to a navigation system for resolving.
For the vehicle-mounted navigation system, a vehicle carrier can contain some specific constraint information under different motion states, and particularly when the vehicle is in a parking state, the ground speed is zero and the posture is not changed, and the zero speed and the zero angular velocity can be used as constraint conditions. The information can provide additional observation information for the navigation system, which is beneficial to improving the precision and stability of the integrated navigation system, and especially plays an important role in inhibiting inertial navigation accumulated errors when GNSS signals are unavailable.
Firstly, according to the potential relation between the output and the motion state of an Inertial device under different motion states, the invention is beneficial to the statistical characteristics of accelerometer data of an Inertial Measurement Unit (IMU) under a static state, and establishes a parking detection criterion, specifically, when a vehicle is in a parking state, because the Inertial device is not influenced by the maneuvering of the vehicle, the output data is relatively stable, so that the parking state can be judged by performing stability analysis on an acceleration output data sequence, a common method adopts the standard difference of the accelerometer data in a fixed time window as a test statistic, and the judgment criterion is as follows:
Figure BDA0003845718770000061
in the formula, A i For data output by the accelerometer at time i, U i Is the average value of the data in the fixed time window at the moment i, the number of the data in the fixed time window is N, T i The standard deviation of the data at the moment i and lambda is a detection threshold;
in one embodiment, 1s is selected as the time window length, and the accelerometer sampling frequency is 100Hz, then N =100.
Because the specifications of different types of inertial sensors are different, the standard deviation of output data in a static state is different, the phenomenon of misjudgment is easy to occur by adopting an empirical threshold, and for the condition researched by the invention, the IMU data for GMWM estimation modeling is required to be ensured to be in the static state, and the influence of misdetection on the accuracy of modeling estimation is larger than that of missing detection, therefore, the detection threshold lambda is dynamically adjusted by the following formula,
Figure BDA0003845718770000062
wherein k is the number of times of detecting a parking spot, and the initial threshold value is empirically selected as T i =0.01。
The physical meaning of the first term is: when the reoccurrence data standard deviation is smaller than the inspection threshold, the vehicle is more likely to be considered to be in a continuous parking state, and then the detection threshold is increased, so that the detection as the parking state is easier;
the physical meaning of the second term is: if the standard deviation of the data at the current moment and the data at the previous moment is too large, the motion state of the vehicle is changed, the vehicle is more likely to be not in the parking state, and then the detection threshold value is reduced, so that the parking state is more difficult to detect.
With the above criteria, it is determined whether the vehicle-mounted object is in a stopped state,
if the vehicle-mounted object is not in a parking state, namely the vehicle-mounted object is in driving, providing navigation positioning service by using an INS/GNSS combined navigation system, and estimating the zero offset of an accelerometer of the IMU;
if the vehicle-mounted object is in a parking state, acquiring parking section IMU observation data and accelerometer zero offset, accumulating the parking section IMU observation data, and subtracting the accelerometer zero offset to obtain a random error component,
the IMU static data acquired during the parking is accumulated with all static data, the IMU static data volume is increased along with the increase of the parking times, and the GMWM modeling of IMU random noise is more and more accurate.
In one embodiment, prior to parameter estimation of the random error component using generalized wavelet moment estimation, a candidate model of the random error component is constructed,
specifically, the random error component is formed by linearly combining independent random processes, and the independent random processes include: gaussian White Noise (WN), random Walk (RW), random Ramp (RR), quantization Noise (QN), and first-order autoregressive process (AR) and flicker noise, and GMWM estimates based on these basic stochastic processes satisfy a consistent and asymptotic normal distribution.
When using a Haar wavelet filter, the analytical expressions for WN, QN, RW, RR and multiple AR processes are shown in table 1,
TABLE 1 PSD and discrete difference equation of state for the underlying stochastic process
Figure BDA0003845718770000071
According to the relation between variance and power spectral density function
Figure BDA0003845718770000081
Taking base
2 logarithm operation on both ends of the formula at v (tau) j )~τ j The typical stochastic process shown in Table 1 is plotted in a log-log plot, which is shown in FIG. 2 as exhibiting a particular slope characteristic, which is shown in Table 2.
The above process can be referred to IMU noise parameter identification-Allen variance (https:// zhuanlan. Zhihu. Com/p/158927004)
TABLE 2 slope of WV in log-log plot of the underlying stochastic process
Model (model) QN WN AR RW Drift(RR)
WV -2 -1 -1~1 1 2
Further, according to the slope characteristic of the log-log curve, a plurality of random error component candidate models are constructed, and data { y } are output for the IMU k K =1, …, N }, according to which v (τ) j )~τ j The slope characteristics presented by the log-log plots of (a) infer its possible composition, constructing it as a linear combination of the QN, WN, RW, RR and k AR models. The candidate model can automatically select an optimal model through artificial inference, or by constructing a random error overall model and a preferred ordering criterion, and for an inertial device, a model for describing random noise of the inertial device is as follows:
x t+1 =e -βΔt x t +w t+1 +u t+1
y t+1 =x t+1 +v t+1
where Δ t is the time interval of the observed quantity, u t =ωΔt,w t N (0,q) and
Figure BDA0003845718770000082
Figure BDA0003845718770000083
the sequence is composed of a first order gaussian markov model, a drift model and a white noise model, i.e., error = GM + DR + WN.
In one implementation, to cover all possible models to the maximum extent, the present invention defines the IMU random noise ensemble model as:
error=5*AR+DR+WN+QN+RW
wherein, random Ramp (RR) is also called Drift (DR),
when the sizing is carried out, all models formed by combining all basic random processes in the models are considered as candidate models,
then, parameter estimation is carried out on a plurality of random error component candidate models by utilizing generalized wavelet moment estimation.
Since the IMU observation data used in the online estimation are accumulated during the parking period, and the data amount is small, the parameter estimation for all candidate models is quite rapid, and the calculation is completed within seconds.
Further, wavelet Variance (WV) can be understood as the Variance of a random process after passing through an approximate band-pass filter, and is calculated by maximum over-sampling Discrete Wavelet Transform (MODWT), and its Wavelet coefficients are calculated by using Wavelet filtering
Figure BDA0003845718770000091
Construction, for j =1, modwt is satisfied,
Figure BDA0003845718770000092
wherein L < 0,l ≧ L 1 Is provided with
Figure BDA0003845718770000093
L 1 Is that
Figure BDA0003845718770000094
M is a non-zero integer.
Figure BDA0003845718770000095
Transfer function ofComprises the following steps:
Figure BDA0003845718770000096
wavelet filter of j-th level h j,l Length L of } j =(2 j -1)(L 1 -1) +1 can be obtained by inverse Fourier transform of the calculation formula,
Figure BDA0003845718770000097
MODWT is actually a discrete wavelet transform filter h j,l A scaled version of (i), i.e.
Figure BDA0003845718770000098
For a finite sequence y k K =1, …, N } uses
Figure BDA0003845718770000099
The obtained MODWT wavelet coefficients:
Figure BDA00038457187700000910
defining wavelet variance as tau j =2 j-1 Sequence on a scale
Figure BDA00038457187700000911
Variance of (c):
Figure BDA00038457187700000912
wavelet coefficient sequence
Figure BDA00038457187700000913
Is stationary, its Power Spectral Density (PSD) is:
Figure BDA00038457187700000914
where, | - | denotes a modulo operation, F θ The model expression F (θ) for generating data is:
Figure BDA00038457187700000915
equation (7) indicates that the variance of the wavelet coefficient sequence is equal to the integral of its power spectral density, i.e., there is an implicit relationship between the wavelet variance and the parameter θ of the model F (θ). Thus, this implicit relationship can be explored by defining a suitable estimator of θ, i.e., by matching v (τ) j ) And the model-based v (τ) given by the sample estimator and equation j ) Expressions are used to estimate the parameter theta. Thus, there is a mapping:
Figure BDA0003845718770000101
for finite sequences y k K =1, …, N }, and MODWT has a wavelet variance estimator of
Figure BDA0003845718770000102
In the formula (I), the compound is shown in the specification,
Figure BDA0003845718770000103
and M j =N-L j +1. The estimator is v (τ) j ) Consistent estimates of (c) are made.
The core idea of GMWM is to find the implicit parameter estimates from the empirical WV obtained from the observation sequence, i.e.
Figure BDA0003845718770000104
Where θ (v) is the inverse function of the theoretical wavelet variance v (θ),
Figure BDA0003845718770000105
is the wavelet variance obtained from the observation sequence using the MODWT estimator.
In most cases, the mapping shown in equation (8) is generally implicit, so it is often difficult to find the inverse function θ (v). However, the difference between the empirical WV obtained from the observation sequence and the theoretical WV contained in the model F (θ) should be minimized, and this idea is not in line with the least squares principle. The essence of the least squares estimation is to establish an estimate such that the smaller the difference between the actual and estimated values, the better, and to measure this difference by the square of the difference between the actual and estimated values.
Therefore, the wavelet variance and the generalized least square principle are combined to construct an estimator, WV obtained by the model is recorded as phi (theta), and WV obtained by the observation sequence is recorded as phi (theta)
Figure BDA0003845718770000106
Calculation of the GLS optimization problem shown by equation (11)
Figure BDA0003845718770000107
Figure BDA0003845718770000108
In the formula, theta is a parameter to be estimated,
Figure BDA0003845718770000109
for the wavelet variance obtained from the observation sequence, φ (θ) is the wavelet variance calculated according to the model, Ω is the positive weighting matrix making the equation convex, and equation (11) is the established GMWM estimator.
As described above, there is consistency with respect to WN, QN, RW, RR and multiple AR models and their combined processes, and thus distribution obeys.
Figure BDA0003845718770000111
In the formula (I), the compound is shown in the specification,
Figure BDA0003845718770000112
and B = (D) T ΩD) -1 D T Omega, matrix
Figure BDA0003845718770000113
When the voltage of the power supply is Ω = I,
Figure BDA0003845718770000114
when selecting
Figure BDA0003845718770000115
The simplest estimator expression can be obtained:
Figure BDA0003845718770000116
when using a Haar wavelet filter, since the WV of a composite process in which a plurality of independent processes are linearly combined is equal to the sum of the WV of the individual independent processes constituting the composite process, i.e., equation (7) can be extended
Figure BDA0003845718770000117
Wherein m is the number of independent processes,
Figure BDA0003845718770000118
is PSD, v mj ) Is in the process { (Y) m ) k Scale τ j At WV. When the analytical expression of v (θ) is known, the parameter estimation can be performed using equation (11).
Performing parameter estimation on a plurality of candidate models of the random error components through generalized wavelet moment estimation to obtain a plurality of parameter estimation results, wherein the parameter estimation results need to be optimized, and correspondingly, the objective function given by the formula (11) is used for measuring phi (theta) obtained by model calculation and observation data
Figure BDA0003845718770000119
A measure of the difference between, minimizing this difference makes the model closer to the observed sequence. Based on this physical meaning, a preference criterion (Goodness of Fitness, GOF) is defined as:
Figure BDA00038457187700001110
in the formula, theta is a parameter to be estimated,
Figure BDA00038457187700001111
for the wavelet variance obtained from the observation sequence,
Figure BDA00038457187700001112
omega is a positive weighting matrix that makes the equation convex, for the wavelet variance calculated from the model,
and after the GMWM is used for finishing parameter estimation of all candidate models, sequencing all the candidate models based on a GOF criterion, and selecting the model with the minimum GOF value as an optimal model. The whole flow chart for automatically selecting the optimal model is shown in fig. 3.
In one embodiment, a plurality of candidate models of the random error components are constructed only for the first section of parking data, and for the nth section of parking data, n is larger than or equal to 1, the optimal model is directly utilized for parameter estimation.
For the same inertia device, the static random noise characteristic is stable, the model composition of the random noise is unchanged along with the increase of time, and only the parameter changes along with the change of time, so the operation of traversing all candidate models only needs to be carried out when the first section of parking data is acquired, and the setting is not needed when the static data accumulation is carried out after the parking state is detected, and the GMWM can be directly used for parameter estimation.
The embodiments in the present description are described in a progressive manner, each embodiment focuses on differences from other embodiments, and the same and similar parts among the embodiments are 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.
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 (8)

1. An on-line estimation algorithm for random errors of a vehicle-mounted inertial device is characterized by comprising the following steps:
s1, judging whether the vehicle is in a parking state or not according to IMU observation data output by a vehicle-mounted object,
if not, using the INS/GNSS integrated navigation system to perform navigation positioning, estimating the zero offset of the accelerometer of the IMU, compensating random noise, and sending the compensation result to the navigation system for resolving;
if yes, executing steps S2-S3;
s2, obtaining parking section IMU observation data and accelerometer zero offset, accumulating the parking section IMU observation data, and subtracting the accelerometer zero offset to obtain a random error component;
and S3, performing parameter estimation on the random error component by using generalized wavelet moment estimation, compensating random noise according to the parameter estimation, and sending a compensation result to a navigation system for resolving.
2. The on-board inertial device random error on-line estimation algorithm according to claim 1, wherein a parking detection criterion is established based on the IMU observation data,
the parking inspection criterion is as follows:
Figure FDA0003845718760000011
in the formula, A i For data output by the accelerometer at time i, U i Is the average value of the data in the fixed time window at the moment i, the number of the data in the fixed time window is N, T i The standard deviation of the data at time i and lambda is the check threshold.
3. The on-board inertial device random error on-line estimation algorithm according to claim 2, wherein the verification threshold λ is dynamically adjusted by the following formula,
Figure FDA0003845718760000012
wherein k is the number of times of detecting a parking spot, and the initial threshold value is empirically selected as T i =0.01。
4. The on-board inertial device random error on-line estimation algorithm according to claim 1, wherein the random error components are linearly combined by an independent random process, and the independent random process comprises: gaussian white noise WN, random walk RW, random ramp RR, quantization noise QN and first order autoregressive process AR.
5. The on-board inertial device random error on-line estimation algorithm according to claim 4, wherein v (τ) for each of the independent random processes is performed before step S3 is performed j )~τ j And performing log-log logarithm processing on the curve to obtain a double-logarithm curve, constructing a plurality of random error component candidate models according to the slope characteristic of the double-logarithm curve, and performing parameter estimation on the plurality of random error component candidate models by utilizing generalized wavelet moment estimation.
6. The on-board inertial device random error on-line estimation algorithm according to claim 5, wherein a preference criterion is constructed to determine optimal parameter estimates for a plurality of the random error component candidate models,
the preference criterion is as follows:
Figure FDA0003845718760000021
in the formula, theta is a parameter to be estimated,
Figure FDA0003845718760000022
for the wavelet variance obtained from the observation sequence,
Figure FDA0003845718760000023
omega is a positive weighting matrix that makes the equation convex for the wavelet variance calculated from the model.
And selecting the model with the minimum GOF value as the optimal model, wherein the parameter estimation corresponding to the optimal model is the optimal parameter estimation.
7. The on-line estimation algorithm for the random error of the vehicle-mounted inertial device according to claim 5, characterized in that a plurality of candidate models of the random error component are constructed for the first section of parking data, and the optimal model is directly utilized to carry out parameter estimation for the nth section of parking data, wherein n is larger than or equal to 1.
8. The on-line estimation algorithm for random error of the vehicle-mounted inertial device according to claim 1, wherein the expression for performing parameter estimation on the random error component by using generalized wavelet moment estimation is as follows:
Figure FDA0003845718760000024
in the formula, theta is a parameter to be estimated,
Figure FDA0003845718760000025
for the wavelet variance obtained from the observation sequence, φ (θ) is the wavelet variance calculated from the model, and Ω is the positive weighting matrix that makes the equation convex.
CN202211117015.7A 2022-09-14 2022-09-14 Random error online estimation algorithm for vehicle-mounted inertial device Pending CN115479615A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202211117015.7A CN115479615A (en) 2022-09-14 2022-09-14 Random error online estimation algorithm for vehicle-mounted inertial device

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202211117015.7A CN115479615A (en) 2022-09-14 2022-09-14 Random error online estimation algorithm for vehicle-mounted inertial device

Publications (1)

Publication Number Publication Date
CN115479615A true CN115479615A (en) 2022-12-16

Family

ID=84424045

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202211117015.7A Pending CN115479615A (en) 2022-09-14 2022-09-14 Random error online estimation algorithm for vehicle-mounted inertial device

Country Status (1)

Country Link
CN (1) CN115479615A (en)

Similar Documents

Publication Publication Date Title
CN111780755B (en) Multi-source fusion navigation method based on factor graph and observability analysis
EP1057141B1 (en) Road profile detection
US20100007550A1 (en) Positioning apparatus for a mobile object
US9057615B2 (en) Systems and methods for navigating using corrected yaw bias values
EP2912407A1 (en) System and method for water column aided navigation
CN110572139B (en) Fusion filtering implementation method and device for vehicle state estimation, storage medium and vehicle
CN110954132A (en) Method for carrying out navigation fault identification through GRNN (generalized regression neural network) assisted adaptive Kalman filtering
CN113465628A (en) Inertial measurement unit data compensation method and system
CN110637209B (en) Method, apparatus and computer readable storage medium having instructions for estimating a pose of a motor vehicle
CN112797967B (en) Random drift error compensation method of MEMS gyroscope based on graph optimization
CN110850450A (en) Adaptive estimation method for satellite clock error parameters
CN104407366B (en) A kind of method being smoothed to pseudorange
CN112762944B (en) Zero-speed interval detection and zero-speed updating method
CN115479615A (en) Random error online estimation algorithm for vehicle-mounted inertial device
CN115047505B (en) GNSS positioning method and navigation method based on carrier phase differential assistance
CN114741659B (en) Adaptive model on-line reconstruction robust filtering method, device and system
CN113932815B (en) Robustness optimization Kalman filtering relative navigation method, device, equipment and storage medium
CN114184209A (en) Inertial error suppression method for low-speed detection platform system
CN114932912A (en) Reference vehicle speed detection method, device, equipment and storage medium
CN113959447A (en) Relative navigation high-noise measurement identification method, device, equipment and storage medium
CN111123323B (en) Method for improving positioning precision of portable equipment
US11982534B2 (en) Prediction of a cartographic readjustment parameter by deep learning
CN112346479B (en) Unmanned aircraft state estimation method based on centralized Kalman filtering
CN113034538A (en) Pose tracking method and device of visual inertial navigation equipment and visual inertial navigation equipment
CN113804202B (en) Integrated navigation method, electronic device and storage medium

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