The present invention relates to satellite radio navigation, and more particularly to methods of obtaining pseudo-range and pseudo-velocity measurements at multipath conditions in navigating receivers of Global Navigating Satellite System (GNSS).
Several GNSSs exist or are created in the world. Those are live systems: Global Positioning System (GPS, USA), Global Navigating Satellite System (GLONASS, Russian Federation), Galileo (European Space Agency, ESA) which is under creation, and BeiDou or Compass (Peoples Republic of China). The present invention can be used in navigation receivers of all of these or other GNSSs.
Using navigation receivers in urban canyon conditions results in frequent missing a line-of-sight of satellites. Power of received signals thus becomes low or vanishes. In these conditions, traditional autonomous signal tracking in a close-loop system by means of delay lock loop (DLL) and a frequency lock loop (FLL) is often interrupted.
Tracking satellite signals in an open-loop system (“Open-Loop Tracking”, [U.S. Pat. No. 6,633,255]) is realized by means of calculating, for each satellite, predictions of delay and Doppler frequency measurements based on navigation solution, forming two-dimensional (delay and Doppler frequency) power grid with the center defined by the forecast of measurements, obtaining corrections for the prediction, and forming measurements as the sums of corrections and the prediction. Such tracking ensures operation for low signal/noise ratios.
The signal loss from a part of satellites in the system realizing tracking in an open loop is not critical while navigation solution exists, i.e., measurements remain based on signals from at least 3-4 satellites.
Presence of signals coming by direct and reflected paths results in occurrence both of delay and Doppler frequency errors.
Methods are known which provide reduction of pseudorange measurement errors resulting from presence of reflected paths. They are generally: a narrow correlator [Theory and Performance of Narrow Correlator Spacing in a GPS receiver. A. J. van Dierendonck, Pat Fenton, Tom Ford. Journal of The Institute of Navigation, Vol. 39, No. 3, 1992.], strobe methods for replica forming for a pseudo-random sequence (PRS) of a signal by way of calculation of autocorrelation function (ACF) [U.S. Pat. No. 7,436,356, filed 24 Mar. 2006], methods based on the usage of ACF leading edge, calculation of pieces crossing on leading and trailing edges and ACF samples weighting [U.S. Pat. No. 7,436,356, filed 24 Mar. 2006]. These methods are easily realized in inexpensive mass produced receivers and reduce a pseudorange error, but their accuracy is not high.
Also known are multipath mitigation methods based on optimum processing of the multipath signal including measurement of signal parameters for direct and reflected paths with applying a method of maximum likelihood. Suppression of the reflected signals effect is realized by optimum estimating and exception of reflected signals from observations. These methods are used in high cost geodetic receivers and demand considerable expensive hardware for its realization.
Usage of adaptive antenna array for suppression of signals coming by reflected paths requires several receive antennas and additional processing of signals received by them, which also complicates consumer equipment considerably.
The closest to the claimed solution is the method described in the U.S. Pat. No. 6,031,881, published Feb. 2, 2000. Disadvantages of this method consist in necessity of saving input signal samples, quantized in a high frequency (in the patent the sampling rate is indicated as equal 20 MHz for GPS system), as well as in processing these samples in real-time. For realization of such method a great volume of memory and a high rate of processing information are required.
Other disadvantage is limited accumulation time. The reason of limiting accumulation time is modulation of a received signal by unknown data bits of the navigation message in a transmitted signal (every 20 ms for GPS and GLONASS), which leads to limiting the received ratio of carrier signal power C to noise power N0 in the band of 1 Hz C/N0>30 dB Hz.
An object of the proposed invention is to eliminate multipath influence on measurements of delays and Doppler frequencies, and to create a method, which is efficient for weak signals and does not require great computing capabilities for its realization. Distinguished technical effect consists in providing bit-synchronous forming an accumulated power grid and suppressing multipath errors by way of construction of a likelihood function with exception of nuisance parameters.
The task in view is solved as follows.
The method of suppressing multipath errors in a satellite navigation receiver includes steps of searching signals of navigation satellites, receiving and processing complex signals from each satellite, tracking the found signals with a following-up open loop and determining coordinates, velocity of the receiver and an exact time based on measurements of a signal delay and Doppler frequency for direct and reflected paths according the following:
based on navigation parameters including coordinates, receiver velocity and exact time, and also information about movement of the satellite, which information obtained on the basis of measured time and received ephemerides, delay and Doppler frequency predictions are calculated for orientation of a two-dimensional accumulated power grid by delay and Doppler frequency, which grid is formed on the basis of received and processed complex signals. At the same time, in the course of forming grid elements, summation of a complex signal is carried out within an interval of coherent integration synchronously with data bits of the given satellite, wherein the time of coherent integration is matched with a bit length. On the basis of processing the grid of accumulated powers, which grid includes its centering, corrections are calculated for delay and Doppler frequency predictions, and delay and Doppler frequency measurements are calculated as sums of predictions and corresponding corrections. After that, extrapolation of received measurements for the preset moment of time is carried out, and calculation of pseudorange and pseudorate measurements is carried out, and on the basis of measurements received from all satellites, coordinates and receiver velocity are defined. Calculation of delay and Doppler frequency corrections comprises:
calculating corrections of a single-path signal using a centered accumulated power grid,
revealing whether multipath exists, and
in case of positive result of said revealing, carrying out corrections improvement using weights,
at the same time, corrections improvement is carried out by forming a likelihood function on the basis of centered accumulated powers vector, and correction is carried out according to the found global maximum of a likelihood function.
Forming the likelihood function on the basis of centered accumulated powers vector can be carried out with various methods, for example, with two methods described below.
According to the first method, forming the likelihood function on the basis of centered accumulated powers vector comprises the following substeps:
According to the second method, forming the likelihood function on the basis of centered accumulated powers vector comprises the following substeps:
Forming the likelihood function for reduction of Doppler frequency measurement errors is carried out similarly.
In the first case forming the likelihood function comprises the following substeps:
In the second case:
The invention is illustrated by drawings, on which:
a and 1b illustrate a diagram of multipath signal propagation leading to distortions of, respectively,
a) delay only,
b) Doppler shift and delay.
Disclosed below is an example of implementation of the invention.
A block scheme of a typical GNSS receiver with tracking in an open loop system is shown in
Signal processing in a typical navigation receiver is divided on analogue (unit 1) and digital (units 2, 3) parts. A part of digital signal processing requiring high rate of computation, and not requiring complex processing algorithms, is performed by a digital very large scale integrated (VLSI) circuit 2. Other part requiring complex processing algorithms at low rate is performed, as a rule, by programmed processor 3.
Radio frequency electromagnetic signals received by an antenna (2.1) are converted in electric signals and amplified. Then, analogue radio frequency transformation is carried out in step 2.2, namely, transferring from radio frequency to intermediate frequency is carried out, and also analogue signal discrete sampling by time and its quantization by level are carried out. The reference frequency obtained in step 2.3 by means of a crystal oscillator is used for transferring and discrete sampling. Then, a digital signal is digitally transferred with frequency downconversion in step 2.4. For this purpose, the digital signal is transferred to nearest to zero frequency, limited by spectrum in a preset frequency band and downconverted according to Nyquist frequency. This signal X(ti) is a digital complex signal with limited digit capacity in view of memory economy. It comprises an additive mix of received from satellites signals and noise. Then, the obtained complex signal X(ti) is used in step 2.5 to forming and centering a grid Z of accumulated powers. Also to forming a grid of accumulated powers used are delay and Doppler frequency predictions: P=(τp,fp)T, where P—vector of predictions, τp—delay prediction, fp—Doppler frequency prediction,
and offset values of elements of accumulated power grid relatively prediction of delay {Δτn} and Doppler frequency {Δft}.
Then, in step 2.6, on the basis of the centered grid Z of accumulated powers, calculation of delays and Doppler frequencies corrections ΔΦ (relatively predictions) is carried out:
ΔΦ=(Δτp,Δfp)T,
where: Δτp—delay correction,
Δfp—Doppler frequency correction.
In step 2.7, obtained corrections are added with delay and Doppler frequency predictions to obtain delay and Doppler frequency measurements:
Φ=P=ΔΦ,
where: Φ=(τ,f)T—vector of measurements of delay τ and Doppler frequency f.
Obtained delay and Doppler frequency measurements Φ serve together with data on a satellite trajectory and velocity (ephemerides) for determination of receiver position, velocity and time in step 2.8 (at that, calculation of pseudorange and pseudovelocity on the basis of delay and Doppler frequency measurements is carried out using a standard method). In step 2.9, on the basis of obtained coordinates, velocity and time, and also data on a trajectory and velocity of a satellite, calculation of delay and Doppler frequency measurements predictions P is performed.
Steps 2.2 and 2.3 are carried out in the analogue part 1, steps 2.4 and 2.5 are carried out in the VLSI circuit 2, steps 2.6-2.9 are carried out in the programmed processor 3.
The system with open loop tracking ensures safe operation for weak signals with carrier power to noise power ratio in the band of 1 Hz: C/N0>10 dB Hz, where C —carrier power, N0—noise power in the band of 1 Hz.
A flow chart of forming one bin of the accumulated power grid with index n,k (n—delay number, k —Doppler frequency number) is represented on
A code mixer (step 3.1) carries out multiplication of complex signal samples by values of real replica of a code:
where:
Xn(1)(ti)—a complex signal from an output of n-th code mixer,
X(ti)—sample of a complex signal from an output of step 2.4,
i—number of a complex signal sample,
ti=δ*i—time of i-th signal sample,
δ—sampling interval,
Cp(τ)—value of the signal code replica at the moment of time t,
τnmix=τp+Δτn—a replica delay of the signal for n-th code mixer,
τp—a delay prediction,
Δτn—an offset value for n-th code mixer,
fp—Doppler frequency prediction,
F—radio frequency of carrier.
A carrier mixer (step 3.2) performs multiplication of a complex signal at the output of code mixer 3.1 by a complex exponent:
Xnl(2)(ti)=Xn(1)(ti)*e−j*2π*f
where:
Xnl(2)(ti)—a complex signal at the output of n-th carrier mixer and n-th code mixer,
j—imaginary unit,
f1mix=fp+Δf1—Doppler frequency value for l-th carrier mixer,
fp—Doppler frequency prediction,
Δf1—a frequency offset value for l-th carrier mixer.
A coherent accumulator (step 3.3) carries out summation of a complex signal in the interval of coherent integration Tcoh=Ncoh*δ synchronously with data bits of the given satellite, wherein the time of the coherent integration is matched with a bit length (20 ms for GPS and GLONASS systems):
where:
Xnl(3)(τp)—a complex signal at the output of a coherent integrator in p-th coherent accumulation interval,
τm=δ*Ncoh*p—time of p-th coherent accumulation interval,
Ncoh—sample quantity of a complex signal in a coherent accumulation interval.
A power calculation unit (step 3.4) calculates a square of a complex signal module from an output of the coherent integrator:
Znl(4)(τp)=Re2[Xnl(3)(τp)]+Im2[Xnl(3)(τp)], (4)
where:
Znl(4)(τp)—real signal power in p-th interval of coherent accumulation,
Re ( ), Im ( )—real and imaginary parts of a complex signal.
As the complex signal at the output of the coherent integrator X(3)nl(τp) is synchronized with data bits, power of this signal does not depend on change of a sign of data in the signal.
An incoherent accumulator (step 3.5) adds powers in the interval of incoherent accumulation Tinc=Tcoh*Ninc:
where:
Znl-signal power at the output of the incoherent integrator in one interval of incoherent accumulation,
Ninc-quantity of samples of real signal powers in an interval of incoherent accumulation.
Similar steps are performed for calculation of each element of accumulated powers with other delay Δτn and Doppler frequency Δf1 offset values relatively predictions. The example of an accumulated power grid resulted from receiving a multipath signal, which differs from a reflected signal in a delay only (see
Calculated accumulated powers are the sum of signal and noise powers. Unknown noise power is a nuisance parameter. For elimination of accumulated noise power, a part of elements of an accumulated power grid, which does not comprise a signal, is used for compensation of this noise power, or centering.
A flow chart of centering is shown in
In the beginning, a statistical model of observations of the accumulated powers vector differing in signal replica delays and corresponding to one peak signal frequency, is considered.
The model of centered vector Z of accumulated power grid, which model is corresponding to peak signal frequency f1max, can be written as:
Z=P(τ)θ(a,Δφ)+ξ, (6)
where:
Z=(zl1, zl2 . . . , zlN)T—vector of accumulated powers after a centering, corresponding to all delays and one peak signal frequency,
N—quantity of powers corresponding to one frequency,
τ=(τ0, τ1 . . . , τK)T—vector of delays of direct and reflected signals, index k=0 relates to a direct signal,
K+1—total number of paths,
a=(a0, a1, . . . , aK)T—vector of amplitudes of signals coming by direct and reflected paths,
Δφ=(Δφ1, Δφ2 . . . , ΔφK)T—vector of phase difference between signals coming by direct and reflected paths (ΔφK=(φK−φ0),
P (τ)=(R00, R01, . . . , R0K, R11, R12, . . . , R1K, RKK)—a matrix with dimension (N×(K+1)*(K+2)/2), depending on delay τ0 of a signal coming by a direct path and delays τ1, τ2, . . . , τK of signals coming by reflected paths.
Rkm—vector of products of ACF signals coming by k-th and m-th paths on elements of accumulated power grid, defined as:
Rkm=(r(τ1mix−τk)r(τ1mix−τm), r(τ2mix−τk)r(τ2mix−τm), . . . , r(τNmix−τk)r(τNmix−τm)T
τk, τm—propagation delays of k-th and m-th paths.
r(τ)—ACF,
τnmix—signal delay in n-th code mixer and n-th element of accumulated power grid (n=1, . . . N),
θ(a, Δφ)—vector depending on amplitudes a and phase difference Δφ of signals, coming by direct and reflected paths, defined as:
θ(a, Δφ)=(a02, 2a0a1 cos(Δφ1), . . . , 2a0aK cos(ΔφK, a12, 2a1a2 cos(Δφ2−Δφ1), . . . , 2a1aK cos(ΔφK−Δφ1), . . . , aK2)T
ξ=(ξ1 . . . , ξN)T—random vector of noise fluctuations of a centered accumulated power grid with mathematical expectation equal to zero and covariance matrix Q=<ξξT>.
As incoherent accumulation time usually is more or equal 1 second, the quantity of incoherent accumulations of power grid Ninc is equal 50 and more. In these conditions, a noise distribution approaches normal distribution with zero mathematical expectation and is entirely defined by known covariance matrix Q.
In the conditions of a real city canyon, the quantity of reflected paths can be large. However, the greatest effect on measurement errors is exerted as a rule by one strongest reflected path. Thus, in case of two paths (direct and one reflected, K=1) number of unknown parameters is equal to 5. They are:
A system of likelihood equations comprises 5 unknown nonlinear parameters. Solving such system is a complex computation problem.
In case of two paths (direct and reflected), matrix P dimension is N×3, vector θ dimension is 3×1, where N —number of accumulated power grid elements for one frequency.
Matrix P and vector θ for two signal paths can be written as:
P(τ)={P0(τ0),P1(τ1),P2(τ0,τ1)} (7)
θ=(a02,a12,2a0a1 cos(Δφ)T, (8)
where:
P0(τ0), P1 (τ1), P2 (τ0, τ1)—vectors with dimension (N×1), depending on delays of direct and reflected paths, with elements:
P0n(τ0)=r2(τnmix−τ0),
P1n(τ1)=r2(τnmix−τ1),
P2n(τ0, τ1)=r(τnmix−τ0) r(τnmix−τ1),
n=1, . . . , N,
r(τ)—ACF,
a0, a1—amplitudes of signals coming by direct and reflected paths,
Δφ=(φ1−φ0)—phase difference of signals coming by direct and reflected paths.
If amplitudes and phase differences of signals are changed during accumulation time of accumulated power grid, it is necessary to replace a linear parameters vector (8) with a vector θ averaged in the same interval of accumulation. Changes of matrix P in the interval of accumulation of grid of accumulated powers may be ignored.
Pseudorandom navigation signals are changed in radio frequency and digital transformations steps 2.2, 2.4. These changes have a determined nature and are completely defined by known transfer functions of used filters. These signals are calculated and used for precalculating autocorrelation functions (ACF) r(τ) by way of multiplication by signal and accumulation replica. ACF r(τ) depends on type of signal replica, receiver bandwidth and amplitude-phase characteristics of analogue and digital filters used in a receiver.
ACF is precalculated and tabulated for simplification of calculations. In the sequel, ACF r(τ) is considered as known function depending on difference between delays of received signal and replica. The form of function r(τ) also depends on the number of GPS pseudo-random sequence (PRS) and number of GLONASS frequency channel.
As time of incoherent accumulation is big enough in comparison with time of the coherent accumulation (for example 1s and 20 ms), noise distribution is normalized. Therefore, further, Gaussian noise model is used at synthesis of the method.
For Gaussian noise distribution in the statistical model of observations described by equation (6), an asymptotically optimum method of conjoint measuring parameters of direct and reflected path signals is the maximum likelihood method (MLM). A likelihood function (LF) for model (6) relating to vector of centered accumulated powers Z can be written as:
y(τ0,τ1,θ)=ln(p(θ,τ|Z)=const−0.5*log(det(2πQ))−0.5*(Z−Pθ)TQ−1(Z−Pθ)−>max (9)
where:
p(Z)—density function of probability distribution of centered vector of accumulated powers Z, which is defined as:
p(Z)={det(2πQ)}−1/2exp{−0.5*(Z−P(τ)θ)Q−1(Z−P(τ)θ)}.
This LF is nonlinear function depending on 5 nonlinear parameters. It is necessary to find LF maximum for all parameters of equation (9). Solution of system of 5 nonlinear likelihood equations is a complex computation problem.
To find the maximum of LF (9) for parameters of delays τ0, τ1 at first used are estimations of maximum likelihood θml related to linear part of parameters θ, which can be obtained in explicit form:
θml=[P(τ)TQ−1P(τ)]−1P(τ)TQ−1Z (10)
After that, obtained estimations θml are substituted in (9) for determining LF maximum y(τ0, τ1) relatively parameters τ0, τ1:
y(τ0,τ1)=−0.5*(ZTQ−1Z−ZTQ−1P(τ)P(τ)TQ−1P(τ))−1P(τ)TQ−1Z)−>max (11)
After involving additional designations:
S=Q−1/2P(τ)—signal vector after noise whitening, depending on signal delays of direct and return paths,
X=Q−1/2Z —centered vector of accumulated powers after noise whitening.
After substituting S and X in expression (11), LF is obtained as function of delays τ0, τ1 in new form:
y(τ0,τ1)=−0.5*XT[E−S(STS)−1ST]X=−0.5*XTX+0.5*XTS(STS)−1STX−>max (12)
The first summand of equation (12) does not depend on parameters τ0, τ1, so it is necessary to calculate the second summand maximum only. A method of correction measurement for parameters of delay Δτ0, Δτ1 is defined, thus, by calculation of the second summand in equation (12).
Flow chart of calculation of measurements with elimination of multipath effect for open loop tracking is shown in
If multipath is detected, improvement of corrections ΔΦ in step 7.3 (described below in details) and calculation of delay and Doppler frequency measurements in step 7.6 are performed with these corrections. The combination of possible offset positions of delays Δτi=τ−τ0 of signals coming by a direct path is defined from the error analysis of used method for determination of single-path signal corrections with presence of reflected path. Delay Δτj of reflected path signal is selected from the band, where a reflected signal has significant effect on measurement accuracy of direct path delay.
During improvement of corrections in step 7.1, weights which can be obtained, for example, from bank of weights in step 7.5, are used. For this purpose, the band of direct and reflected signal delays is divided into intervals, for each of which precalculation of weights required for calculating likelihood function is performed in step 7.4 (probably out of real-time system), and writing these precalculated coefficients in a bank of weights is performed.
A flow chart of measuring set according to equation (12) is shown in
In step 8.1, noise whitening (multiplication of vector of centered accumulated powers Z by matrix Q−1/2) is performed:
Matrix Q−1/2 can be calculated, for example, by way of decomposition into two triangular matrices according to Cholesky scheme and subsequent triangular matrix inversion [Trefethen L. N., Bau D. Numerical Linear Algebra, SIAM, 1997, p. 172]. It is necessary to notice that noise whitening operation only reduces measurement noises and does not take an effect on offset errors caused by multipath. Modeling has shown that without whitening (which corresponds to Q=E, where E is a unitary matrix), direct signal delay measurements with presence of multipath remain unshifted (
Further, in step 8.2, three sets of weights corresponding to vectors S1ij, S2ij, S3ij are obtained for each value of direct and reflected path delays τi0, τj1, accordingly, and, in step 8.3, calculation of three weighed sums of vector X components with these weights is performed:
where:
q=1, 2, 3—number of column-vector of Sij matrix, corresponding to a pair of delays of direct and reflected paths,
i, j—numbers of delays of direct and reflected paths on a grid of LF construction.
After step 8.3, obtained values of weighed sums h1ij, h2ij and h3ij are multiplied pairwise in step 8.5, and then they are summed with the weights of an inverse matrix Rij=(SijTSij)−1 with dimension 3×3 obtained in step 8.4, for calculating LF:
This operation represents calculation of a quadratic form.
Result of calculations is LF {εij}, saved in memory in step 8.6.
This procedure is repeated for each pair of delays of direct and reflected paths (each pair of delays τi0, τj1 is associated with corresponding values of weight vectors Sij and inverse matrices Rij). After that, in step 8.7, determining and interpolating global maximum for all calculated LF values {εij} for delay parameters of direct and reflected paths τi0, τj1 are performed.
A delay search area for direct and reflected paths was determined from analysis of pseudorange errors calculated for a narrow correlator (
τmax−15
where C—velocity of light,
and reflected path delays τj1 were selected in a band, in which reflected path takes an effect on accuracy of direct path delay estimation, namely:
τmax+15
Modeling has shown that for search of direct path delay the step of 1-2 m/C, and for search of reflected path delay the step of 10-20 m/C is required.
Example of LF for two (direct and 1 reflection) path signal is shown in
Two-dimensional parabolic LF maximum approximation was used for obtaining delay {circumflex over (τ)}0, {circumflex over (τ)}1 estimations of signals coming by direct and reflected paths.
Comparative modeling, which results are submitted on
Bias errors of delay measurement of the signal coming by direct path for the proposed method represented in
Above (
The matrix (STS) is a symmetrical nonnegatively defined matrix with dimension 3×3. This matrix can be submitted in the following form:
(STS)=UΛUT, (16)
where:
U—a square matrix of orthonormal eigenvectors with dimension 3×3,
Λ=diag (λ1, λ2, λ3)—a diagonal matrix of non-negative eigenvalues.
Accordingly, inverse matrix (STS)−1 can be submitted in a form:
(STS)−1=UΛ−1UT=(UΛ−1/2)(UΛ−1/2)T. (17)
Thus, matrix S(STS)−1ST can be submitted in a form:
S(STS)−1ST=(SUΛ−1/2)(SUΛ−1/2)T=VVT, (18)
where V=(V1, V2, V3)=SUΛ−1/2—a matrix with dimension N×3, which columns V1, V2, V3 are 3 orthonormalized eigenvectors of matrix S(STS)−1ST. This matrix depends only on delays (τ0, τ1) of direct and reflected paths. After substitution of equation (18) in (12), calculation of LF can be written as:
y(τ0,τ1=const+0.5*XTV(τ0,τ1)V(τ0,τ1)TX−>max. (19)
Using orthonormal vectors simplifies digital implementation of the proposed method. In this case, operation of inverting matrices is excluded, quadratic form is led to diagonal kind, and integer calculations can be applied.
Another method of similar simplification of calculations is obtained at matrix (STS) decomposition into two triangular matrices according to the Cholesky scheme (See reference above —Trefethen L. N., Bau D. Numerical Linear Algebra, SIAM, 1997, p. 172):
(STS)=ΔTΔ, (20)
where:
Δ—lower triangular matrix,
ΔT—upper triangular matrix.
In this case inverse matrix is equal:
(STS)−1=(ΔTΔ)−1=Δ−1(Δ−1)T. (21)
Operation of triangular matrices conversation becomes essentially simpler. In this case, orthonormalized vectors V are precalculated according to the formula:
V=(V1,V2,V3)=SΔ−1. (22)
Further processing is similar to processing with eigenvectors (19).
Orthonormalized vectors V also can be obtained from vectors S using Gram-Shmidt procedure of orthonormalization [Trefethen L. N., Bau D. Numerical Linear Algebra, SIAM, 1997, p. 56]
Its differences from described above are that, in step 9.3 of calculating weighed sums for each vector, orthonormalized vectors Vij=(Vij1, Vij2, Vij3) are used:
where g=1, 2, 3—eigenvector number.
Further, in step 9.4, sum of squares of weighed sums hijg is calculated:
which is LF value corresponding to parameters of delays τi0, τj1 of signals coming by direct and reflected paths.
The method of reduction of measurement errors of delay of a signal coming by a direct path in presence of a reflected path having the same frequency has been considered above.
A multipath signal, observed by moving consumer, has offset in frequency domain caused by Doppler effect. In this case, a direct path signal has not shifted Doppler offset, and a reflected path signal has Doppler offset depending on position of reflecting surface (see
In this case, matrix P, unlike expression (7), can be written in the following form:
P=P(f)=(P0,P1,P2), (25)
where:
f=(f0, f1)T—a vector of unknown Doppler offsets for direct f0 and reflected f1 paths,
P0, P1, P2—vectors of a signal with dimension L×1 depending on Doppler frequencies of direct and reflected signals with elements:
P0l=sin c2[π(flmix−f0)Tcoh],
P1l=sin c[π(flmix−f0)Tcoh] sin c[π(flmix−f1)Tcoh],
P2l=sin c2[π(flmix−f1)Tcoh],
l=1, . . . , L—accumulated power grid frequency number,
flmix—frequency of l-th carrier mixer (and of l-th element of power grid),
Tcoh-coherent accumulation time.
Function of sinc (x) is defined as: sin c(x)=sin(x)/[Ncoh·sin(x/Ncoh)].
Bins of centered power grid are used as vector Z=(Z1n, Z2n, . . . , ZLn)T, which grid corresponds to all peak signal delay frequencies σnmax. Clarified corrections Δf0, Δf1 are obtained by maximization of LF, which LF is set by equation (26) corresponding to usage of signal vectors:
y(f0,f1)=−0.5*XT[E−S(STS)−1ST]X=−0.5*XTX+0.5*XTS(STS)−1STX−>max, (26)
where:
S=Q−1/2P(f0, f1)—matrix of signal vectors depending on Doppler offsets f0, f1 for direct and reflected paths, after noise whitening.
X=Q1/2Z—centered accumulated powers vector corresponding to all frequencies and one delay of peak signal, after noise whitening.
Equation (26) also can be simplified by usage of orthonormalized vectors of a signal:
y(f0,f1)=const+0.5*XTV(f0,f1)V(f0,f1)TX−>max, (27)
where:
V(f0, f1)—orthonormalized signal vector matrix depending on Doppler frequencies.
Implementation of the method is similar to described above concerning delays and represented in
Results of comparative tests are presented for confirmation of efficiency of the claimed method (
In
Pseudorange errors are decreased 4-5 times due to the usage of the claimed method.
Thus, the proposed delays and Doppler frequency measurement method allows to suppress multipath errors for weak signals using power grid data.
Coherent accumulation of complex signals is performed bit-synchronously with coherent accumulation time matched with a data bit length equal 20 ms for GPS and GLONASS systems. It allows to dejam from unknown data bits in a transmitted signal navigation message and to lower an admissible signal/noise ratio below a data decoding threshold.
In case if data bits are known or can be decoded, a corresponding to these bits complex signal carrier phase inversion can be compensated. In this case, coherent accumulation time is selected based on reference frequency oscillator properties and usually equals tenths of second. Multipath errors suppression according to accumulated power grid data, in this case, also is possible for low signal/noise ratios.
The method allows to suppress multipath errors, arising at measurement of both delay (pseudorange) and Doppler frequency (pseudovelocity) after detection of multipath according to power grid data.
The proposed method also allows performing operations necessary for suppression of multipath errors, 1 time per second and can be implemented in a programmed processor.
This application claims the benefit of U.S. Provisional Application No. 61/170,120, filed Apr. 17, 2009, which is incorporated herein by reference in its entirety.
Number | Name | Date | Kind |
---|---|---|---|
5361276 | Subramanian | Nov 1994 | A |
6031881 | Weill et al. | Feb 2000 | A |
6407699 | Yang | Jun 2002 | B1 |
6633255 | Krasner | Oct 2003 | B2 |
7436356 | Bochkovskiy et al. | Oct 2008 | B2 |
7646830 | Weill | Jan 2010 | B1 |
Number | Date | Country | |
---|---|---|---|
20100265133 A1 | Oct 2010 | US |
Number | Date | Country | |
---|---|---|---|
61170120 | Apr 2009 | US |