gnss/5g hybridization for urban navigation
TRANSCRIPT
The French Civil Aviation University
La rรฉfรฉrence aรฉronautique
Ecole Nationale de lโAviation Civile
www.enac.fr
Rรฉf: Version: Date:
1/21
1
GNSS/5G Hybridization for Urban Navigation
Ph.D. Dissertation Defense by Anne-Marie TOBIE
Henk Wymeersch Chalmers University of technology President
Gonzalo Seco-Granados Autรณnoma de Barcelona Reviewer
Liang Chen Wuhan University Reviewer
Didier Belot CEA-LETI Member
Corinne Mailhes INP ENSEEIHT Member
Axel Garcia-Pena ENAC Ph.D. thesis director
The French Civil Aviation University
La rรฉfรฉrence aรฉronautique
Ecole Nationale de lโAviation Civile
www.enac.fr
Rรฉf: Version: Date:
2/21
Introduction โ Context (1/2)
โข The need for positioning in constrained environment is in constant growth.
โข GNSS positioning solution in harsh environment is degraded and challenging for urban applications due to multipath and lack of Line-of-Sight (LOS) satellite visibility
โข Due to GNSS limitations, several alternatives are already developed:
โ Hybridization with additional sensors
โ Usage of Signal of Opportunity: mobile communication signals such as 4G LTE or 5G,โฆ
2
The French Civil Aviation University
La rรฉfรฉrence aรฉronautique
Ecole Nationale de lโAviation Civile
www.enac.fr
Rรฉf: Version: Date:
3/21
Introduction โ Context (1/2)
โข The need for positioning in constrained environment is in constant growth.
โข GNSS positioning solution in harsh environment is degraded and challenging for urban applications due to multipath and lack of Line-of-Sight (LOS) satellite visibility
โข Due to GNSS limitations, several alternatives are already developed:
โ Hybridization with additional sensors
โ Usage of Signal of Opportunity: mobile communication signals such as 4G LTE or 5G,โฆ
โข 5G systems use Orthogonal Frequency Division Multiplexing (OFDM) signals, OFDM signal-type ranging modules are already developed in the literature
โข These modules were derived by assuming a constant propagation channel over the duration of an OFDM symbol. An analysis conducted on QuaDRiGa has shown that these models must be refined.
โข Global objective: the design of hybrid navigation modules using both GNSS and 5G measurements in a realistic environment.
3
The French Civil Aviation University
La rรฉfรฉrence aรฉronautique
Ecole Nationale de lโAviation Civile
www.enac.fr
Rรฉf: Version: Date:
4/21
4
Introduction โ Context (2/2)
Global objective: the design of hybrid navigation modules using both GNSS and 5G measurements in a realistic environment.
The French Civil Aviation University
La rรฉfรฉrence aรฉronautique
Ecole Nationale de lโAviation Civile
www.enac.fr
Rรฉf: Version: Date:
5/21
โข Detailed objective 1: The description the 5G physical layer
5
Introduction โ Context (2/2)
Global objective: the design of hybrid navigation modules using both GNSS and 5G measurements in a realistic environment.
The French Civil Aviation University
La rรฉfรฉrence aรฉronautique
Ecole Nationale de lโAviation Civile
www.enac.fr
Rรฉf: Version: Date:
6/21
โข Detailed objective 2: The selection GNSS and 5G compliant propagation channels
6
Introduction โ Context (2/2)
Global objective: the design of hybrid navigation modules using both GNSS and 5G measurements in a realistic environment.
The French Civil Aviation University
La rรฉfรฉrence aรฉronautique
Ecole Nationale de lโAviation Civile
www.enac.fr
Rรฉf: Version: Date:
7/21
โข Detailed objective 3: The development of highly realistic 5G signals correlator outputs for ranging based positioning considering a realistic urban propagation channel
7
Introduction โ Context (2/2)
Global objective: the design of hybrid navigation modules using both GNSS and 5G measurements in a realistic environment.
The French Civil Aviation University
La rรฉfรฉrence aรฉronautique
Ecole Nationale de lโAviation Civile
www.enac.fr
Rรฉf: Version: Date:
8/21
โข Detailed objective 4: The development of 5G ranging modules to estimate the time-of-arrival (TOA) of the signal.
8
Introduction โ Context (2/2)
Global objective: the design of hybrid navigation modules using both GNSS and 5G measurements in a realistic environment.
The French Civil Aviation University
La rรฉfรฉrence aรฉronautique
Ecole Nationale de lโAviation Civile
www.enac.fr
Rรฉf: Version: Date:
9/21
โข Detailed objective 5: The precise Derivation of GNSS and 5G pseudo range errors due to the propagation channels
9
Introduction โ Context (2/2)
Global objective: the design of hybrid navigation modules using both GNSS and 5G measurements in a realistic environment.
The French Civil Aviation University
La rรฉfรฉrence aรฉronautique
Ecole Nationale de lโAviation Civile
www.enac.fr
Rรฉf: Version: Date:
10/21
โข Detailed objective 6: The development of hybrid navigation modules exploiting/adapted to the derived pseudo range measurements mathematical models.
10
Introduction โ Context (2/2)
Global objective: the design of hybrid navigation modules using both GNSS and 5G measurements in a realistic environment.
The French Civil Aviation University
La rรฉfรฉrence aรฉronautique
Ecole Nationale de lโAviation Civile
www.enac.fr
Rรฉf: Version: Date:
11/21
Outline
1. 5G signal presentation
2. Propagation channel
3. Correlator output mathematical models
4. 5G ranging module design
5. Derivation of pseudo range error distributions
6. Navigation modules and results
7. Conclusion
11
The French Civil Aviation University
La rรฉfรฉrence aรฉronautique
Ecole Nationale de lโAviation Civile
www.enac.fr
Rรฉf: Version: Date:
12/21
Outline
1. 5G signal presentation
2. Propagation channel
3. Correlator output mathematical models
4. 5G ranging module design
5. Derivation of pseudo range error distributions
6. Navigation modules and results
7. Conclusion
12
The French Civil Aviation University
La rรฉfรฉrence aรฉronautique
Ecole Nationale de lโAviation Civile
www.enac.fr
Rรฉf: Version: Date:
13/21
โข Objective 1: The description the 5G physical layer 1.1. The interest of 5G for positioning
1.2. 5G transmitted signal
1.3. 5G synchronization signal
13
1. 5G signal presentation (1/4)
The French Civil Aviation University
La rรฉfรฉrence aรฉronautique
Ecole Nationale de lโAviation Civile
www.enac.fr
Rรฉf: Version: Date:
14/21
1. 5G signal presentation (2/4)
1.1. The interest of 5G for positioning โ Objective:
โข To present the interest of 5G for positioning
โ Approaches to address 5G requirements:
โข Denser Network
Many more emitters
โข Millimetre Waves
More bandwidth
โข Massive MIMO antenna
Angle of Arrival measurements
14
5G requirements โ source [15]
Trilateration interest
The French Civil Aviation University
La rรฉfรฉrence aรฉronautique
Ecole Nationale de lโAviation Civile
www.enac.fr
Rรฉf: Version: Date:
15/21
1. 5G signal presentation (2/4)
1.1. The interest of 5G for positioning โ Objective:
โข To present the interest of 5G for positioning
โ Approaches to address 5G requirements:
โข Denser Network
Many more emitters
โข Millimetre Waves
More bandwidth
โข Massive MIMO antenna
Angle of Arrival measurements
15
5G requirements โ source [15]
Trilateration interest
The French Civil Aviation University
La rรฉfรฉrence aรฉronautique
Ecole Nationale de lโAviation Civile
www.enac.fr
Rรฉf: Version: Date:
16/21
1. 5G signal presentation (3/4)
1.2. 5G transmitted signal โ 5G uses Orthogonal Frequency Division Multiplexing (OFDM)
โ Interest:
โข Simple:
โ modulation using an iFFT
โ demodulation using a FFT
โข Robust in multipath environment
โ Frequency selectivity
โ Time dispersity (cycle prefix)
โ OFDM with a scalable numerology
โ Two frequency ranges are identified:
โข FR1: 410 MHz โ 7125 MHz (main focus)
โข FR2: 24250 MHz โ 52600 MHz mmW (partly analyzed)
16
OFDM modulation
๐ โ๐ [๐๐ฏ๐] Cyclic prefix
0 15 Normal
1 30 Normal
2 60 Normal, Extended
3 120 Normal
4 240 Normal OFDM scalable numerology
The French Civil Aviation University
La rรฉfรฉrence aรฉronautique
Ecole Nationale de lโAviation Civile
www.enac.fr
Rรฉf: Version: Date:
17/21
1. 5G signal presentation (4/4)
1.3. Synchronization signals โ The definition of a 5G signal is made through the time/frequency allocation of resources in a resource grid:
โข A resource grid is composed of time-consecutive OFDM symbols
โข An OFDM symbol is composed of frequency-consecutive subcarriers carrying data or pilot symbols.
โข The data symbols correspond to the useful information transmitted to the 5G user โ a-priori unknown.
โข The pilot symbols are known symbols in terms of time/frequency localization and value pilot symbols are used to compute the correlation
โ 3GPP standards has shown that the only set of pilots having a fixed position in the resource grid is the SS/PBCH block.
โ A simplified 5G frame is considered, each symbol of the frame corresponds to the 2nd SS/PBCH symbol
17
SS/PBCH block Frequency domain allocation
โ๐
[๐๐ฏ๐]
๐
[๐๐ฏ๐]
15 3.6
30 7.2
60 14.4
120 28.8
240 57.6
The French Civil Aviation University
La rรฉfรฉrence aรฉronautique
Ecole Nationale de lโAviation Civile
www.enac.fr
Rรฉf: Version: Date:
18/21
Outline
1. 5G signal presentation
2. Propagation channel
3. Correlator output mathematical models
4. 5G ranging module design
5. Derivation of pseudo range error distributions
6. Navigation modules and results
7. Conclusion
18
The French Civil Aviation University
La rรฉfรฉrence aรฉronautique
Ecole Nationale de lโAviation Civile
www.enac.fr
Rรฉf: Version: Date:
19/21
โข Objective 2: The selection GNSS and 5G compliant propagation channels โ Getting a realistic and precise model of the propagation channel is mandatory to develop signal
processing techniques
โ The selection of an appropriate propagation channel is a trade off between complexity and accuracy
โ A deep study of the literature has been performed to find the best compromises
19
2. Propagation channel (1/3)
The French Civil Aviation University
La rรฉfรฉrence aรฉronautique
Ecole Nationale de lโAviation Civile
www.enac.fr
Rรฉf: Version: Date:
20/21
2. Propagation channel (2/3)
โข GNSS propagation channel: SCHUN [12] โ SCHUN: Simplified Channel for Urban Navigation; it
is a hybrid physical-statistical Land Mobile Satellite propagation channel
โ SCHUN:
โข generates the environment by using a virtual city approach (statistic aspect).
โข models the interactions between impinging signals and the environment by using simple Electromagnetic (EM) interaction models (deterministic aspect)
20
SCHUN macro architecture โ source [12]
The French Civil Aviation University
La rรฉfรฉrence aรฉronautique
Ecole Nationale de lโAviation Civile
www.enac.fr
Rรฉf: Version: Date:
21/21
2. Propagation channel (3/3)
โข 5G propagation channel: QuaDRiGa [7] โ New innovations require an evolved propagation channel model:
โข Technologies: massive Multiple Input Multiple Output antenna, denser network, or millimetre waves
โข Case studies: wide variety of case study envisioned for 5G
โ QuaDRiGa: QUAsi Deterministic RadIo channel GenerAtor
โข Developed by the Fraunhofer Heinrich Hertz Institute
โข Enable the modeling of MIMO radio channels for specific network configurations: indoor, satellite or heterogeneous configurations.
โ The QuaDRiGa approach is a โstatistical ray-tracing modelโ. It does not use an exact geometric representation of the environment but distributes the positions of the scattering clusters (the sources of indirect signals such as buildings or trees) randomly.
21
Simplified overview of the modelling approach in QuaDRiGa โ source [7]
Distribution of random scattering clusters example
The French Civil Aviation University
La rรฉfรฉrence aรฉronautique
Ecole Nationale de lโAviation Civile
www.enac.fr
Rรฉf: Version: Date:
22/21
Outline
1. 5G signal presentation
2. Propagation channel
3. Correlator output mathematical models
4. 5G ranging module design
5. Derivation of pseudo range error distributions
6. Navigation modules and results
7. Conclusion
22
The French Civil Aviation University
La rรฉfรฉrence aรฉronautique
Ecole Nationale de lโAviation Civile
www.enac.fr
Rรฉf: Version: Date:
23/21
โข Objective 3: The development of highly realistic 5G signals correlator outputs model for ranging based positioning considering a realistic propagation channel
3.1. To correctly characterize the CIR propagation channel
โข To define the CIR sampling rate
โข To identify the significant time-varying channel parameters with respect to symbol duration
โข To select the appropriate CIR sampling rate
3.2. To derive the correlator output mathematical model of a 5G signal for ranging purposes
23
3. Correlator output models
The French Civil Aviation University
La rรฉfรฉrence aรฉronautique
Ecole Nationale de lโAviation Civile
www.enac.fr
Rรฉf: Version: Date:
24/21
3. Correlator output models
โข 3.1. Impact of a time-evolving channel (1/2) โ High level Objective:
โข To correctly model the propagation channel and its parameters
โ Observation:
โข The CIR sampling rate is the number of CIR samples directly generated by the discrete propagation channel model over the duration of an OFDM symbol
โข 100 CIR per OFDM symbol is equivalent to a time continuous propagation channel reference
โข The CIR sampling rate has consequences on the entire processing chain
โข The path delays evolution over an OFDM symbol is negligible but the propagation channel complex amplitude evolution is significant
โ Detailed objective
โข To select the appropriate CIR sampling
rate to model the evolving propagation channel
24
Channel coefficients modulus over the OFDM symbol
The French Civil Aviation University
La rรฉfรฉrence aรฉronautique
Ecole Nationale de lโAviation Civile
www.enac.fr
Rรฉf: Version: Date:
25/21
3. Correlator output models
3.1. Impact of a time-evolving channel (2/2)
โ Objective: โข To select the appropriate CIR sampling rate to model the evolving propagation channel
โข Appropriate = low enough to obtain a simple mathematical model while still able to accurately represent a true continuous CIR propagation channel
โ Method:
25
๐ด๐ธ๐,๐ = ๐ท100 โ ๐ท๐ ๐
5G symbol generation (SSPBCH)
Processing steps
1 CIR per OFDM
10 CIR per OFDM
100 CIR per OFDM (ref)
20 CIR per OFDM
Demodulation
Correlation
DLL Discriminator
output
Criteria
C2: ๐ด๐ธ๐,๐ <50 ๐๐ 95% ๐๐ ๐ก๐๐๐1 ๐ 99% ๐๐ ๐ก๐๐๐
C1: ๐๐ด๐ธ๐,๐
๐๐ด๐ธ๐,๐<
10 ๐๐ ๐ฟ๐๐ ๐ ๐๐๐๐๐๐๐20 ๐๐ ๐๐ฟ๐๐ ๐ ๐๐๐๐๐๐๐
๐ท1
๐ท10
๐ท100
๐ท20
Propagation channel
application
The French Civil Aviation University
La rรฉfรฉrence aรฉronautique
Ecole Nationale de lโAviation Civile
www.enac.fr
Rรฉf: Version: Date:
26/21
3. Correlator output models
3.1. Impact of a time-evolving channel (2/2)
โ Objective: โข To select the appropriate CIR sampling rate to model the evolving propagation channel
โข Appropriate = low enough to obtain a simple mathematical model while still able to accurately represent a true continuous CIR propagation channel
โ Method:
26
10 CIR per OFDM symbol is the appropriate CIR sampling rate for the majority of analyzed scenario
The French Civil Aviation University
La rรฉfรฉrence aรฉronautique
Ecole Nationale de lโAviation Civile
www.enac.fr
Rรฉf: Version: Date:
27/21
3. Correlator output models
3.2. Correlator output mathematical model (1/3) โ The general correlation formula is
๐ ํ๐ =1
๐๐ ๐ ๐โฒ โ ๐๐โฒ๐
โ2๐๐๐โฒ
๐๐น๐น๐๐
๐โฒโ๐
โ Where
โข ๐๐โฒ the locally generated pilot symbol
localized in the ๐โฒ๐กโ
subcarrier
โข ๐๐ the number of pilot symbols,
โข ๐ ๐โฒ the demodulated symbols
โข ๐ set of pilots in the OFDM symbol
27
OFDM symbol demodulation
Correlator
Local replica
๐บ๐๐
๐๐โฒ =
๐ ๐โฒ๐๐๐๐๐๐
+๐ ๐โฒ๐๐๐๐๐๐๐ ๐๐๐
+๐ ๐โฒ๐๐๐๐๐๐๐๐๐๐๐๐
+๐๐๐๐๐
๐น = ๐น๐๐๐๐๐๐
+๐น๐๐๐๐๐๐๐ ๐๐๐
+๐น๐๐๐๐๐๐๐๐๐๐๐๐
+๐น๐๐๐๐๐
The French Civil Aviation University
La rรฉfรฉrence aรฉronautique
Ecole Nationale de lโAviation Civile
www.enac.fr
Rรฉf: Version: Date:
28/21
3. Correlator output models
3.2. Correlator output mathematical model (2/3) โ The OFDM signal-type correlator output mathematical model
๐น ๐บ๐ = ๐น๐๐๐๐๐๐ ๐บ๐ + ๐น๐๐๐๐๐๐๐ ๐๐๐๐บ๐ + ๐น๐๐๐๐๐๐๐๐๐๐๐
๐บ๐ + ๐น๐๐๐๐๐ ๐บ๐
28
Correlation operation for 5G signal
The French Civil Aviation University
La rรฉfรฉrence aรฉronautique
Ecole Nationale de lโAviation Civile
www.enac.fr
Rรฉf: Version: Date:
29/21
3. Correlator output models
3.2. Correlator output mathematical model (2/3) โ The OFDM signal-type correlator output mathematical model
๐น ๐บ๐ = ๐น๐๐๐๐๐๐ ๐บ๐ + ๐น๐๐๐๐๐๐๐ ๐๐๐๐บ๐ + ๐น๐๐๐๐๐๐๐๐๐๐๐
๐บ๐ + ๐น๐๐๐๐๐ ๐บ๐
29
Correlation operation for 5G signal โ Observation:
โข The resulting complete model is still quite complex
โข To derive a simplified model could be beneficial for theoretical purposes and for practical purposes
The French Civil Aviation University
La rรฉfรฉrence aรฉronautique
Ecole Nationale de lโAviation Civile
www.enac.fr
Rรฉf: Version: Date:
30/21
3. Correlator output models
3.2. Correlator output mathematical model (2/3) โ The OFDM signal-type correlator output mathematical model
๐น ๐บ๐ = ๐น๐๐๐๐๐๐ ๐บ๐ + ๐น๐๐๐๐๐๐๐ ๐๐๐๐บ๐ + ๐น๐๐๐๐๐๐๐๐๐๐๐
๐บ๐ + ๐น๐๐๐๐๐ ๐บ๐
โ Observation:
โข This simplification would allow the generation of only 1 CIR sample per OFDM symbol instead of 10 CIR samples.
โ Method: Simplification of the propagation channel contribution on the useful term:
๐ด๐๐ 0
๐จ๐๐ ๐ โ
๐ถ๐๐ ๐ + ๐ถ๐
๐+๐ ๐
๐๐๐๐ฝ๐๐๐
๐๐ ๐น๐๐๐โ ๐ต๐ญ๐ญ๐ปโ๐
๐๐๐ ๐ ๐น๐๐๐๐ต๐ญ๐ญ๐ป
๐๐๐ ๐ ๐น๐๐๐
โ Conclusion: The simplified model can be used instead of the initial one for the noiseless useful term for all scenarios since the simplified model is statistically equivalent to the initial model using 10 CIR per OFDM symbol.
30
Correlation operation for 5G signal
The French Civil Aviation University
La rรฉfรฉrence aรฉronautique
Ecole Nationale de lโAviation Civile
www.enac.fr
Rรฉf: Version: Date:
31/21
3. Correlator output models
3.2. Correlator output mathematical model (2/3) โ The OFDM signal-type correlator output mathematical model
๐น ๐บ๐ = ๐น๐๐๐๐๐๐ ๐บ๐ + ๐น๐๐๐๐๐๐๐ ๐๐๐๐บ๐ + ๐น๐๐๐๐๐๐๐๐๐๐๐
๐บ๐ + ๐น๐๐๐๐๐ ๐บ๐
31
Correlation operation for 5G signal
โ The ICI term can be seen as a degradation of the available ๐๐๐ or ๐/๐ต๐
The French Civil Aviation University
La rรฉfรฉrence aรฉronautique
Ecole Nationale de lโAviation Civile
www.enac.fr
Rรฉf: Version: Date:
32/21
3. Correlator output models
3.2. Correlator output mathematical model (2/3) โ The OFDM signal-type correlator output mathematical model
๐น ๐บ๐ = ๐น๐๐๐๐๐๐ ๐บ๐ + ๐น๐๐๐๐๐๐๐ ๐๐๐๐บ๐ + ๐น๐๐๐๐๐๐๐๐๐๐๐
๐บ๐ + ๐น๐๐๐๐๐ ๐บ๐
32
โ Model:
โข The ICI can be, due to the Central Limit Theorem, modelled as a Gaussian random variable which statistics are dependent on the propagation channel generated and thus are trajectory- and scenario-dependent.
โ Validation:
โข The Skewness and Kurtosis statistics have been computed for 10000 Monte Carlo and over the Early, Prompt and Late correlator outputs.
โข For a Gaussian distribution, the Skewness is equal to 0 and the Kurtosis to 3.
Skewness and Kurtosis measurements
Trajectory 1 3
scenario LOS LOS LOS NLOS
Case 1 1 2 1
Skewness measures
R(ฮตฯ)
Early 0.04 -0.03 -0.01 -0.01
Prompt 0.02 -0.01 -0.04 -0.02
Late -0.01 -0.00 -0.01 -0.02
Kurtosis measures
R(ฮตฯ)
Early 2.87 2.93 3.02 3.04
Prompt 3.01 2.96 3.04 3.04
Late 3.21 3.02 2.98 3.10
The French Civil Aviation University
La rรฉfรฉrence aรฉronautique
Ecole Nationale de lโAviation Civile
www.enac.fr
Rรฉf: Version: Date:
33/21
3. Correlator output models
3.2. Correlator output mathematical model (2/3) โ The OFDM signal-type correlator output mathematical model
๐น ๐บ๐ = ๐น๐๐๐๐๐๐ ๐บ๐ + ๐น๐๐๐๐๐๐๐ ๐๐๐๐บ๐ + ๐น๐๐๐๐๐๐๐๐๐๐๐
๐บ๐ + ๐น๐๐๐๐๐ ๐บ๐
33
Correlation operation for 5G signal
The French Civil Aviation University
La rรฉfรฉrence aรฉronautique
Ecole Nationale de lโAviation Civile
www.enac.fr
Rรฉf: Version: Date:
34/21
3. Correlator output models
3.2. Correlator output mathematical model (3/3) โ Another model is therefore possible
๐น ๐บ๐ = ๐น๐๐๐๐๐๐ ๐บ๐ + ๐น๐๐๐๐๐+๐๐๐๐๐๐ ๐บ๐
where
๐น๐๐๐๐๐+๐๐๐๐๐๐ ๐บ๐ ~๐ต ๐น๐๐๐๐๐๐๐๐๐๐๐๐บ๐ , ๐ฝ๐๐ ๐น๐๐๐๐๐ ๐ + ๐ฝ๐๐ ๐น๐๐๐๐๐๐๐ ๐๐๐
๐บ๐
34
Correlation operation for 5G signal
The French Civil Aviation University
La rรฉfรฉrence aรฉronautique
Ecole Nationale de lโAviation Civile
www.enac.fr
Rรฉf: Version: Date:
35/21
3. Correlator output models
โข Conclusion โ CIR characterization
โข The propagation channel complex amplitude significantly varying in time
โข The CIR sampling rate selection is constrained by the trade-off between the generation computing time and the realism of the modelling.
โข Using 10 CIR samples per OFDM is suitable to model the propagation channel variation
โข The time-evolution of the propagation channel creates a spectral broadening effect. This spectral broadening implies the loss of the orthogonality among the OFDM signal subcarriers.
โ Correlator output mathematical model
๐น ๐บ๐ = ๐น๐๐๐๐๐๐ ๐บ๐ + ๐น๐๐๐๐๐+๐๐๐๐๐๐ ๐บ๐
35
Noiseless useful part
Simplified model available
Additive Gaussian variable: ~ enhanced noise power
The French Civil Aviation University
La rรฉfรฉrence aรฉronautique
Ecole Nationale de lโAviation Civile
www.enac.fr
Rรฉf: Version: Date:
36/21
Outline
1. 5G signal presentation
2. Propagation channel
3. Correlator output mathematical models
4. 5G ranging module design
5. Derivation of pseudo range error distributions
6. Navigation modules and results
7. Conclusion
36
The French Civil Aviation University
La rรฉfรฉrence aรฉronautique
Ecole Nationale de lโAviation Civile
www.enac.fr
Rรฉf: Version: Date:
37/21
โข Objective 4: The development of 5G ranging modules to estimate the time-of-arrival (TOA) of the signal.
4.1. Description of the TOA estimator
4.2. Necessity analysis of the FLL to improve the tracking performances
4.3. Derivation of a ๐ถ/๐0 estimator to state the lock condition of the tracking loops and to fill the filters measurement covariance matrix
37
4. 5G ranging module design
Step 1: DLL
General parameter
Sensitivity
Step 2: FLL
General parameter
Impact of the CFO
AWGN Validation
Step 3: ๐ถ/๐0 estimator
Candidate
Limits
Solution proposed
Ranging module
Complete channel analyses
The French Civil Aviation University
La rรฉfรฉrence aรฉronautique
Ecole Nationale de lโAviation Civile
www.enac.fr
Rรฉf: Version: Date:
38/21
4. 5G ranging module design
4.1.1. DLL Presentation (1/2) โ Objective:
โข To present the delay estimator
โ The proposed DLL is the result of an optimal derivation when assuming a constant AWGN propagation channel
โ It can be re-used for the correlator output derived before with the same optimal solution status.
โ The joint modelling of the noise and ICI terms as a Gaussian variable allows to select the proposed DLL
โ 38
Delay Lock Loop architecture
Update Rate the OFDM symbol duration,
๐๐ ๐ฆ๐๐๐๐ =๐๐น๐น๐+๐๐ถ๐
ฮ๐๐๐น๐น๐
Discriminator Early Minus Late Power
Order 2nd order
One-sided noise bandwidth ๐ต๐ = 10 ๐ป๐ง and ๐ต๐ = 1 ๐ป๐ง
The French Civil Aviation University
La rรฉfรฉrence aรฉronautique
Ecole Nationale de lโAviation Civile
www.enac.fr
Rรฉf: Version: Date:
39/21
4. 5G ranging module design
4.1.2. DLL Presentation (2/2)
โ Sensitivity: โข The SNR at the RF front end output value below for which the
loop can be considered to have lost tracking.
โข The tracking loss criterion specifies that the DLL has lost its lock if the tracking error falls outside the linearity zone of the discriminator
39
DLL threshold (๐บ๐ต๐น๐ ๐๐๐๐ ๐๐๐๐๐๐๐๐๐)
โข Usual criterion:
๐ ๐๐๐[๐บ๐] < ๐ซ๐๐
โข Where
โ ๐ฃ๐๐ ํ๐ variance of the tracking error estimate
โ ๐ท๐กโ one-sided linearity zone of the discriminator
โข ๐บ๐ต๐น๐ ๐๐๐๐ ๐๐๐๐๐๐๐๐ =๐๐๐
๐๐๐ฉ๐๐ป๐๐ฒ๐
๐ต๐๐ + ๐ +
๐โ๐๐๐ฒ๐
๐๐๐๐ฉ๐๐ป๐๐ฒ๐
The French Civil Aviation University
La rรฉfรฉrence aรฉronautique
Ecole Nationale de lโAviation Civile
www.enac.fr
Rรฉf: Version: Date:
40/21
4. 5G ranging module design
4.2. Necessity to implement a carrier tracking loop โ Observation:
โข The DLL study assumed a perfect estimation of the 5G signals carrier
frequency.
โ Problematic
โข The baseband received signal is affected by a residual carrier frequency offset which impact must be analyzed, and corrected if necessary
๐ฟ๐ = ๐๐๐ + ๐ฟ๐๐
Which is the impact of the fractional CFO on the DLL tracking performance when applying the correlation integration for 1 OFDM symbol?
40
Integer part Fractional part ๐ฟ๐๐ โ [โ0.5, 0.5]
๐๐๐ ๐น๐๐
Impact Shift on the demodulated symbol Attenuation + phase rotation on the demodulated
symbol and adds an additional ICI term
Correction Easily estimated/corrected after the demodulation Cannot be corrected after the demodulation
Study Impact assumed corrected and effect neglected Requires to be estimated and corrected
The French Civil Aviation University
La rรฉfรฉrence aรฉronautique
Ecole Nationale de lโAviation Civile
www.enac.fr
Rรฉf: Version: Date:
41/21
4. 5G ranging module design
4.2.2. Theoretical degradation due to CFO โ Objective:
โข To quantify the theoretical degradation of the SNR due to the CFO in
order to infer the necessity of estimating the incoming signal carrier frequency
โ Theoretical degradation of the SNR due to the CFO
โข ๐ท =๐๐๐
๐๐ผ๐๐ =
๐
๐๐ด2๐
๐+๐๐ผ๐ถ๐ผ
=1
๐ด2 1 +7 ๐๐ฟ๐
2
2๐๐๐๐๐ where: ๐ด =
1
๐๐น๐น๐
๐ ๐๐ ๐๐ฟ๐๐๐๐น๐น๐
๐ ๐๐ ๐๐ฟ๐๐
โข Realistic SNR
โ A lower bound for the SNR can be set using the DLL sensitivity
โข Realistic doppler
โ A worst ๐ฟ๐ is derived when the receiver goes away from a base
Station at high speed (50 km/h)
โ ๐ฟ๐,๐ค๐๐๐ ๐ก = 2๐๐๐๐ ๐๐๐๐๐๐๐ =2๐๐
ฮ๐๐๐น๐น๐=
2๐๐โ๐ฃ
๐๐น๐น๐โฮ๐โ๐ (๐๐ฆ๐๐๐/๐ ๐๐๐๐๐)
41
Case Speed ๐ฃ Carrier
frequency ๐๐
Subcarrier
spacing ฮ๐
1
50 km/h
2 GHz 15 kHz
2 30 GHz 120 kHz
3 60 GHz 120 kHz
Determination of a realistic value of frequency error
The French Civil Aviation University
La rรฉfรฉrence aรฉronautique
Ecole Nationale de lโAviation Civile
www.enac.fr
Rรฉf: Version: Date:
42/21
4. 5G ranging module design
4.2.2. Theoretical degradation due to CFO โ Maximum BS emitted power provided in 3GPP standard: ๐๐๐๐๐ฅ โ โ6 ๐๐ต๐.
โ Assumptions: ๐๐ โ โ30 ๐๐ต๐ โ ๐๐๐ ๐๐๐๐๐๐๐๐ก๐๐๐ = 96 ๐๐ต no saturation
โ A carrier frequency estimator is theoretically not necessary in a AWGN channel for the considered conditions
42
Degradation due to a fractional CFO
Impact of the Doppler on the ๐บ๐ต๐น๐๐๐๐๐๐๐๐๐๐๐
๐๐๐ ๐๐๐๐๐๐๐๐ก๐๐๐ = 120 ๐๐ต โ
Base station emitted power
๐๐ โ โ15 ๐๐ต๐
The French Civil Aviation University
La rรฉfรฉrence aรฉronautique
Ecole Nationale de lโAviation Civile
www.enac.fr
Rรฉf: Version: Date:
43/21
4. 5G ranging module design
4.2.1. FLL presentation โ Objective: To present the FLL candidate to lead the analyses
43
Update Rate At half the symbol rate, or every 2 โ ๐๐ ๐ฆ๐๐๐๐
Discriminator computed at half symbol rate: ๐ท๐น๐ฟ๐ฟ=
๐๐๐๐๐ ๐ ๐ ๐ โ๐ ๐ ๐โ1 โ
2๐๐๐น๐น๐
Order 2nd order
Loop bandwidth ๐ต๐๐๐๐ = 10 ๐ป๐ง
Local replica phase ๐๐๐๐, updated at symbol rate:
๐๐๐๐ ๐ = ๐๐๐๐ ๐ โ 1 + 2๐๐๐ ๐
VCO ๐๐ ๐ , updated every two symbols
โ The FLL outputs are used to correct the phase and Doppler of the incoming signal
๐๐๐๐๐,๐๐๐๐๐๐๐ก๐๐ = ๐๐
๐๐๐๐
โ๐ ๐ฝ๐๐๐ ๐โ๐ +2๐๐ฝ๐ช ๐ ๐
โ The contribution channel of the correlator output mathematical model is slightly modified as
๐ด๐๐ 0 โ
๐ผ๐๐ 0 + ๐ผ๐
๐+1 0
2๐๐ ๐0๐
๐ โ๐ฝ๐๐๐ ๐๐๐๐ ๐ฟ๐๐
๐โ๐ฝ๐ ๐ โ ๐๐น๐น๐โ1๐ ๐๐ ๐ ๐ฟ๐๐
๐โ๐ฝ๐ ๐ ๐๐น๐น๐
๐ ๐๐ ๐ ๐ฟ๐๐๐โ๐ฝ๐ ๐
The French Civil Aviation University
La rรฉfรฉrence aรฉronautique
Ecole Nationale de lโAviation Civile
www.enac.fr
Rรฉf: Version: Date:
44/21
4. 5G ranging module design
4.2.3. AWGN Validation โ Objective:
โข Validation of the theoretical results AWGN
โ Presentation of the tested configurations
44
Standard deviations of the tracking error estimate
โ Observations
โข Error approximately equal to 1.5 cm for SNR higher than 40 dB, it decreases to 3 mm for 0 dB
โข Error lower when a FLL is used
โข The differences at high SNR are mostly due to numerical issues.
โ Conclusion:
โข Confirmation of the theoretical derivations: the implementation of a carrier frequency estimator is not mandatory for an AWGN propagation channel.
CFE Case 1 Case 4
Fine frequency acquisition process Not conducted Conducted
Carrier frequency tracking process Not implemented FLL implemented
The French Civil Aviation University
La rรฉfรฉrence aรฉronautique
Ecole Nationale de lโAviation Civile
www.enac.fr
Rรฉf: Version: Date:
45/21
4. 5G ranging module design
4.2.4. Analyses for a complete (QuaDRiGa) channel
45
Propagation channel path power for the radial trajectory
Cases 10 Hz-DLL 1 Hz-DLL
ฮผ [m] ฯ [m] RMSE [m] ฮผ [m] ฯ [m] RMSE [m]
CFE C1 0,63 4,74 4,78 -0,99 4,17 4,28
CFE C4 0,57 4,71 4,75 0,21 2,76 2,77 Pseudo range estimation error statistics
โ Hypothesis:
โข Due to the degradation linked to the multipath, a carrier frequency estimator could improve the tracking performances
โ Test case:
โข 4 CFE cases assuming a 10 Hz and a 1 Hz DLL loop bandwidth and for the first 300 m of the radial trajectory
โ Observations:
โข FLL not required for a 10 Hz DLL loop bandwidth
โข the tracking performances improved for a 1 Hz DLL loop bandwidth when using a FLL
โ Conclusion:
โข The FLL is not required but can improve the performances of the ranging module according to the DLL loop bandwidth chosen
The French Civil Aviation University
La rรฉfรฉrence aรฉronautique
Ecole Nationale de lโAviation Civile
www.enac.fr
Rรฉf: Version: Date:
46/21
4. 5G ranging module design
4.3.1. ๐ช/๐ต๐ estimator implementation (1/3) โ ๐ถ/๐0 estimator candidate: Narrow-to-Wideband-Power-Ratio
โข The method is based on the comparison of total signal-
plus-noise power in two different bandwidths
โข After simulations: ๐พ = 50 and ๐ = 10 is the best found set of values
โ Limits identification:
โข In AWGN, a saturation is observed
46
AWGN case ๐ช/๐ต๐ estimator study
๐ถ/๐0 = 10 โ log10
1
๐
๐ ๐๐ โ 1
๐ โ ๐ ๐๐ ๐ ๐๐ =
1
๐พ
๐๐ต๐๐๐๐ต๐๐
๐พ
๐=1
๐๐ต๐๐ = ๐ผ๐2 + ๐๐
2
๐
๐=1๐
๐ต๐ฉ๐ท๐ = ๐ฐ๐
๐ด
๐=๐๐
๐
+ ๐ธ๐
๐ด
๐=๐๐
๐
phase sensitive!
The French Civil Aviation University
La rรฉfรฉrence aรฉronautique
Ecole Nationale de lโAviation Civile
www.enac.fr
Rรฉf: Version: Date:
47/21
4. 5G ranging module design
4.3.2. ๐ช/๐ต๐ estimator implementation (2/3) โ Limits identification:
โข For a complete QuaDRiGa channel
โข The NWPR- ๐ถ/๐0 estimator works only for the circular trajectory in the LOS case where the LOS signal phase is constant the ๐ช/๐ต๐ estimator is sensitive to phase variation
โข The evolution of the multipath phases for the radial trajectory is twice the evolution of the multipath phases for the circular trajectory the saturation threshold is linked to the phase variation amplitude
47
Multipath case ๐ช/๐ต๐ estimator study โ Radial trajectory ๐ช/๐ต๐ estimator behavior in multipath case for a circular trajectory
Saturation around ๐ถ/๐0 =60 ๐๐ต๐ป๐ง
Saturation around ๐ถ/๐0 =80 ๐๐ต๐ป๐ง
The French Civil Aviation University
La rรฉfรฉrence aรฉronautique
Ecole Nationale de lโAviation Civile
www.enac.fr
Rรฉf: Version: Date:
48/21
4. 5G ranging module design
4.3.3. ๐ช/๐ต๐ estimator implementation (3/3) โ Presentation of potential solutions:
48
๐ช/๐ต๐ estimator Interest Behaviour Conclusion
Moment Method estimator
It is insensitive to the correlator outputs phase evolution
The estimation is working properly for ๐ถ/๐0 < 60 ๐๐ต๐ป๐ง.
The estimation is not behaving as desired for higher ๐ถ/๐0.
arctangent-NWPR C/N0 phase-corrected
It corrects the phase in the initial NWPR ๐ถ/๐0 estimator
๐ ๐k๐๐๐๐๐๐๐ก๐๐
ํ๐ = ๐ ๐k ํ๐ ๐โ๐โ ๐
[๐]
๐ ๐ = atan๐ผ๐๐๐ ๐ ๐
k๐
๐ ๐๐๐ ๐ ๐k
๐
The estimation works well for ๐ถ/๐0 < 90/95 ๐๐ต๐ป๐ง but it is still erroneous for higher ๐ถ/๐0 values
Both NWPR ๐ถ/๐0 phase-corrected estimators are better than the NWPR- ๐ถ/๐0 estimator. For ๐ถ/๐0 > 90/95 ๐๐ต๐ป๐ง, the estimator is still not working optimally.
FLL-driven NWPR ๐ถ/๐0 phase-corrected
๐ ๐ = ๐ ๐ โ 1 + 2๐๐๐,๐น๐ฟ๐ฟ ๐
The French Civil Aviation University
La rรฉfรฉrence aรฉronautique
Ecole Nationale de lโAviation Civile
www.enac.fr
Rรฉf: Version: Date:
49/21
4. 5G ranging module design
โข Conclusion: โ ToA estimation:
โข Since the correlator output can be model as a Gaussian variable, the optimal DLL structure for a constant propagation channel can be reused for the evolving propagation channel
โ Carrier Frequency estimation:
โข In AWGN a carrier tracking loop is not necessary
โข For a complete propagation channel, the FLL is not required but can improve the performances of the ranging module according to the DLL loop bandwidth chosen
โ ๐ถ/๐0 estimation:
โข ๐ถ/๐0 is expected to be used as lock detector and to fill the measurement covariance matrix of the filter
โข The NWPR- ๐ถ/๐0 estimator cannot correctly estimate high ๐ถ/๐0 values since it saturates due to the phase evolution between symbols in the channel contribution term
โข The arctangent and FLL driven ๐ถ/๐0 phase corrected NWPR estimator can be used instead but are still limited for high ๐ถ/๐0
49
The French Civil Aviation University
La rรฉfรฉrence aรฉronautique
Ecole Nationale de lโAviation Civile
www.enac.fr
Rรฉf: Version: Date:
50/21
Outline
1. 5G signal presentation
2. Propagation channel
3. Correlator output mathematical models
4. 5G ranging module design
5. Derivation of pseudo range error distributions
6. Navigation modules and results
7. Conclusion
50
The French Civil Aviation University
La rรฉfรฉrence aรฉronautique
Ecole Nationale de lโAviation Civile
www.enac.fr
Rรฉf: Version: Date:
51/21
โข Objective 5: The precise Derivation of GNSS and 5G pseudo range errors due to the propagation channels
5.1. Pseudo range definition and time frame considerations
5.2. Problematic and method derivations:
5.3. Presentation of the characterization step
5.4. Presentation of the approximation step
5.5. QuaDRiGa approximation
5.6. SCHUN approximation
51
5. Pseudorange derivation (1/9)
The French Civil Aviation University
La rรฉfรฉrence aรฉronautique
Ecole Nationale de lโAviation Civile
www.enac.fr
Rรฉf: Version: Date:
52/21
5. Pseudorange derivation (2/9)
5.1. Pseudo range definition and time frame considerations
52
The French Civil Aviation University
La rรฉfรฉrence aรฉronautique
Ecole Nationale de lโAviation Civile
www.enac.fr
Rรฉf: Version: Date:
53/21
5. Pseudorange derivation (3/9)
5.2. Problematic and method derivations: โ The propagation channels SCHUN and QuaDRiGa and the thermal noise introduce errors on the pseudo range
measurements.
โ The derivation process objective is to find the Gaussian distribution which bests approximates a true distribution (minimization of a cost function), that will allow to maximize the performances of the navigation filters
โ In order to derive Gaussian distributions for the pseudo range measurements, a two-steps method has been developed:
53
Description of the method
1. To accurately characterize in terms of mean, variance and probability density functions (PDF) the pseudo range measurement errors.
2. To approximate these PDFs by Gaussian distributions. Two algorithms have been developed in this work: an over-bounding method and a fitting method.
The French Civil Aviation University
La rรฉfรฉrence aรฉronautique
Ecole Nationale de lโAviation Civile
www.enac.fr
Rรฉf: Version: Date:
54/21
โ SCHUN scenario design:
โข The trajectory of the receiver
โ Ensure constant relative azimuth
โข The satellites location in azimuth and elevation
โ Uniform distribution
โข The virtual city characterized by the building width, height, etc
โ QuaDRiGa scenario design:
โข high-level scenario
โ typical terrestrial pico-base stations deployed below rooftop in densely populated urban areas
โข signal/receiver characteristics
โ 15 kHz subcarrier spacing
โ 2 GHz carrier frequency.
โข receiverโs trajectories
5. Pseudorange derivation (4/9)
5.3. Presentation of the characterization step
54
From these scenarios, a PDF characterization is made for each value of true ๐ช/๐ต๐ (2 dB step) Step 2 is then applied to each PDF characterization
The French Civil Aviation University
La rรฉfรฉrence aรฉronautique
Ecole Nationale de lโAviation Civile
www.enac.fr
Rรฉf: Version: Date:
55/21
5. Pseudorange derivation (5/9)
5.4. Presentation of the approximation step โ Objective:
โข To select the Gaussian distribution which minimizes a given cost function, where the cost function depends on the Gaussian approximation method.
โ Approximation parameters
55
CDF of the error, the fitting and overbounding distributions
Confidence bound 95% 99% 99.9%
Mean ๐ 0, ๐ ๐ ๐๐๐๐, ๐ ๐ ๐๐๐ฅ, ๐
โ Fitting
โข To find the variance such that the norm of the error over the abscises axis between the fitting CDF and the CDF of the error to be fitted is minimized.
โข Cost function:
๐ด๐๐
๐๐ช๐ซ๐ญ๐๐๐๐๐ โ ๐ช๐ซ๐ญ๐๐๐๐๐๐๐ ๐
๐๐ ๐๐๐ ๐๐๐๐๐๐ ๐๐๐๐ ๐๐๐๐๐ ๐๐๐๐๐๐๐๐
โ Overbounding
โข To determine the minimal variance of the distribution that will permit to envelop the true distribution over the wanted confidence bound.
โข Cost function:
๐๐๐๐๐๐๐๐ ๐๐๐๐๐ ๐๐ ๐ช๐ซ๐ญ๐๐๐๐๐๐๐๐๐ ๐๐๐ ๐ฌ๐๐๐๐ โ ๐ช๐ซ๐ญ๐๐๐๐๐ ๐ฌ๐๐๐๐ > ๐ ๐๐๐ ๐ฌ๐๐๐๐ < ๐
๐ช๐ซ๐ญ๐๐๐๐๐๐๐๐๐ ๐๐๐ ๐ฌ๐๐๐๐ โ ๐ช๐ซ๐ญ๐๐๐๐๐ ๐ฌ๐๐๐๐ < ๐ ๐๐๐ ๐ฌ๐๐๐๐ > ๐
The French Civil Aviation University
La rรฉfรฉrence aรฉronautique
Ecole Nationale de lโAviation Civile
www.enac.fr
Rรฉf: Version: Date:
56/21
5. Pseudorange derivation (6/9)
5.4. Presentation of the approximation step โ Objective:
โข To select the Gaussian distribution which minimizes a given cost function, where the cost function depends on the Gaussian approximation method.
โ Method comparison:
โข The fitting distribution allows a better derivation of small errors while the overbounding distribution will cover the tails of the error characterized
56
PDF of the error, the fitting and overbounding distributions
The French Civil Aviation University
La rรฉfรฉrence aรฉronautique
Ecole Nationale de lโAviation Civile
www.enac.fr
Rรฉf: Version: Date:
57/21
5. Pseudorange derivation (7/9)
5.5. QuaDRiGa approximation: โ Parameters:
โข Characterisation for a LOS picocell scenario
โข Generation of pseudo ranges for a set of trajectories to model different distance between the receiver and the BS
โข Characterization made for the true ๐ถ/๐0
57
QuaDRiGa pseudo range errors characterization
The French Civil Aviation University
La rรฉfรฉrence aรฉronautique
Ecole Nationale de lโAviation Civile
www.enac.fr
Rรฉf: Version: Date:
58/21
5. Pseudorange derivation (8/9)
5.6. SCHUN characterization: โ Parameters:
โข Characterisation based on LOS signal with multipath
โข Generation of pseudo ranges for a uniform set of azimuth and elevation satellites
โข Characterization made with respect to the ๐ถ/๐0
โ Characterization made for GPS and used for Galileo as a higher bound characterization
58
SCHUN pseudo range errors characterization
The French Civil Aviation University
La rรฉfรฉrence aรฉronautique
Ecole Nationale de lโAviation Civile
www.enac.fr
Rรฉf: Version: Date:
59/21
5. Pseudorange derivation (9/9)
5.7. Conclusion โ The derivation of Gaussian distribution for the pseudo range
measurement errors in presence of noise and multipath has been performed for both 5G and GNSS signal
โ The final outputs are look up tables obtained using the true ๐ถ/๐0
59
QuaDRiGa conclusions SCHUN conclusion
- The receiver tracking parameter has an impact.
- The fitting and the over bounding methodologies yield very different results.
- The distributions are not Gaussian, especially for 5G at low ๐ถ/๐0, (positively skewed).
- A slight increase of the variances can be observed as the ๐ถ/๐0 decreases
- The distributions for ๐ถ/๐0 >35 ๐๐ต๐ป๐ง are Gaussian
Look up table for QuaDRiGa
The French Civil Aviation University
La rรฉfรฉrence aรฉronautique
Ecole Nationale de lโAviation Civile
www.enac.fr
Rรฉf: Version: Date:
60/21
Outline
1. 5G signal presentation
2. Propagation channel
3. Correlator output mathematical models
4. 5G ranging module design
5. Derivation of pseudo range error distributions
6. Navigation modules and results
7. Conclusion
60
The French Civil Aviation University
La rรฉfรฉrence aรฉronautique
Ecole Nationale de lโAviation Civile
www.enac.fr
Rรฉf: Version: Date:
61/21
โข Objective 6: The development of hybrid navigation modules exploiting/adapted to the derived pseudo range measurements mathematical models. 6.1. Navigation Filter Presentation
6.2. Configurations
6.3. Results
61
6. Navigation modules and results
The French Civil Aviation University
La rรฉfรฉrence aรฉronautique
Ecole Nationale de lโAviation Civile
www.enac.fr
Rรฉf: Version: Date:
62/21
6. Navigation modules and results
6.1. Navigation filters presentation (1/2) โ State model: ๐ฟ๐ = ๐ฟ ๐ฝ๐ฟ ๐ ๐ฝ๐ ๐ ๐ฝ๐ ๐ โ ๐๐๐ฎ๐ท๐บ/๐๐๐๐ ๐ โ ๐๐ ๐ฎ๐ท๐บ/๐๐๐๐ ๐ โ ๐๐๐ฎ๐ท๐บ/๐ฎ๐๐ ๐ โ ๐๐๐ฎ๐ท๐บ/๐๐ฎ
โ Measurement model: ๐๐ = ๐ ๐ฟ๐ + ๐ฝ๐
โข โ ๐๐ is observation matrix โ ๐๐ =
โ(๐บ๐๐) ๐๐
โ(๐บ๐๐) ๐๐
โ(5๐บ) ๐๐
โ (๐บ๐๐) ๐๐
โ (๐บ๐๐) ๐๐
โข โ(๐) are pseudo range code measurements
โข โ (๐)are pseudo range rate measurements
โข ๐๐ measurement noise vector
โ Navigation filters: EKF and UKF
62
The French Civil Aviation University
La rรฉfรฉrence aรฉronautique
Ecole Nationale de lโAviation Civile
www.enac.fr
Rรฉf: Version: Date:
63/21
6. Navigation modules and results
6.1. Navigation filters presentation (1/2) โ State model: ๐ฟ๐ = ๐ฟ ๐ฝ๐ฟ ๐ ๐ฝ๐ ๐ ๐ฝ๐ ๐ โ ๐๐๐ฎ๐ท๐บ/๐๐๐๐ ๐ โ ๐๐ ๐ฎ๐ท๐บ/๐๐๐๐ ๐ โ ๐๐๐ฎ๐ท๐บ/๐ฎ๐๐ ๐ โ ๐๐๐ฎ๐ท๐บ/๐๐ฎ
โ Measurement model: ๐๐ = ๐ ๐ฟ๐ + ๐ฝ๐
โข โ ๐๐ is observation matrix โ ๐๐ =
โ(๐บ๐๐) ๐๐
โ(๐บ๐๐) ๐๐
โ(5๐บ) ๐๐
โ (๐บ๐๐) ๐๐
โ (๐บ๐๐) ๐๐
โข โ(๐) are pseudo range code measurements
โข โ (๐)are pseudo range rate measurements
โข ๐๐ measurement noise vector
โ Navigation filters: EKF and UKF
63
Hypothesis: Perfectly known
The French Civil Aviation University
La rรฉfรฉrence aรฉronautique
Ecole Nationale de lโAviation Civile
www.enac.fr
Rรฉf: Version: Date:
64/21
6. Navigation modules and results
6.1. Navigation filters presentation (1/2) โ State model: ๐ฟ๐ = ๐ฟ ๐ฝ๐ฟ ๐ ๐ฝ๐ ๐ ๐ฝ๐ ๐ โ ๐๐๐ฎ๐ท๐บ/๐๐๐๐ ๐ โ ๐๐ ๐ฎ๐ท๐บ/๐๐๐๐ ๐ โ ๐๐๐ฎ๐ท๐บ/๐ฎ๐๐ ๐ โ ๐๐๐ฎ๐ท๐บ/๐๐ฎ
โ Measurement model: ๐๐ = ๐ ๐ฟ๐ + ๐ฝ๐
โข โ ๐๐ is observation matrix โ ๐๐ =
โ(๐บ๐๐) ๐๐
โ(๐บ๐๐) ๐๐
โ(5๐บ) ๐๐
โ (๐บ๐๐) ๐๐
โ (๐บ๐๐) ๐๐
โข โ(๐) are pseudo range code measurements
โข โ (๐)are pseudo range rate measurements
โข ๐๐ measurement noise vector
โ Navigation filters: EKF and UKF
64
The French Civil Aviation University
La rรฉfรฉrence aรฉronautique
Ecole Nationale de lโAviation Civile
www.enac.fr
Rรฉf: Version: Date:
65/21
6. Navigation modules and results
6.1. Navigation filters presentation (1/2) โ State model: ๐ฟ๐ = ๐ฟ ๐ฝ๐ฟ ๐ ๐ฝ๐ ๐ ๐ฝ๐ ๐ โ ๐๐๐ฎ๐ท๐บ/๐๐๐๐ ๐ โ ๐๐ ๐ฎ๐ท๐บ/๐๐๐๐ ๐ โ ๐๐๐ฎ๐ท๐บ/๐ฎ๐๐ ๐ โ ๐๐๐ฎ๐ท๐บ/๐๐ฎ
โ Measurement model: ๐๐ = ๐ ๐ฟ๐ + ๐ฝ๐
โข โ ๐๐ is observation matrix โ ๐๐ =
โ(๐บ๐๐) ๐๐
โ(๐บ๐๐) ๐๐
โ(5๐บ) ๐๐
โ (๐บ๐๐) ๐๐
โ (๐บ๐๐) ๐๐
โข โ(๐) are pseudo range code measurements
โข โ (๐)are pseudo range rate measurements
โข ๐๐ measurement noise vector
โ Navigation filters: EKF and UKF
65
Hypothesis: BS perfectly synchronized among each others
The French Civil Aviation University
La rรฉfรฉrence aรฉronautique
Ecole Nationale de lโAviation Civile
www.enac.fr
Rรฉf: Version: Date:
66/21
6. Navigation modules and results
6.1. Navigation filters presentation (1/2) โ State model: ๐ฟ๐ = ๐ฟ ๐ฝ๐ฟ ๐ ๐ฝ๐ ๐ ๐ฝ๐ ๐ โ ๐๐๐ฎ๐ท๐บ/๐๐๐๐ ๐ โ ๐๐ ๐ฎ๐ท๐บ/๐๐๐๐ ๐ โ ๐๐๐ฎ๐ท๐บ/๐ฎ๐๐ ๐ โ ๐๐๐ฎ๐ท๐บ/๐๐ฎ
โ Measurement model: ๐๐ = ๐ ๐ฟ๐ + ๐ฝ๐
โข โ ๐๐ is observation matrix โ ๐๐ =
โ(๐บ๐๐) ๐๐
โ(๐บ๐๐) ๐๐
โ(5๐บ) ๐๐
โ (๐บ๐๐) ๐๐
โ (๐บ๐๐) ๐๐
โข โ(๐) are pseudo range code measurements
โข โ (๐)are pseudo range rate measurements
โข ๐๐ measurement noise vector
โ Navigation filters: EKF and UKF
66
The French Civil Aviation University
La rรฉfรฉrence aรฉronautique
Ecole Nationale de lโAviation Civile
www.enac.fr
Rรฉf: Version: Date:
67/21
6. Navigation modules and results
6.1. Navigation filters presentation (2/2)
67
UKF processing
EKF processing
Linearization techniques Example: observation
equation Illustration
EKF
- Based on an analytical method - Taylor polynomial expansion
about a single point
๐ ๐+1|๐ = ๐ป๐ ๐ ๐
๐ป๐ =๐โ
๐๐๐|๐๐=๐ ๐|๐โ1
UKF
- Based on a statistical method - linear regression between ๐
points drawn from the prior distribution of the random variable.
- The state distribution is represented using a minimal set of carefully chosen sample points called sigma points.
๐ ๐+1|๐
โ ๐ค๐โ ๐๐๐, ๐๐
2๐
๐=0
๐ : ๐ ๐ก๐๐ก๐ ๐ฃ๐๐๐ก๐๐ ฮฃ: ๐๐๐ฃ๐๐๐๐๐๐๐ ๐๐๐ก๐๐๐ฅ ๐: ๐ ๐ก๐๐ก๐ ๐ฃ๐๐๐ก๐๐ ๐๐๐๐๐๐ ๐๐๐
๐ค๐: ๐ค๐๐๐โ๐ก
๐๐๐ : ๐ ๐๐๐๐ ๐๐๐๐๐ก ๐
The French Civil Aviation University
La rรฉfรฉrence aรฉronautique
Ecole Nationale de lโAviation Civile
www.enac.fr
Rรฉf: Version: Date:
68/21
6. Navigation modules and results
6.2. Configurations โ Objective:
โข To determine the positioning performances obtained using the
previously describe navigation filters
โ Scenario
โข SCHUN and QuaDRiGa have been used to generate realistic propagation channel
โข Ultra Dense Network envisioned
โข High mask angle value implying few satellites in view
โข 60 s simulation duration
โ Figure of merit
โข Root Mean Square Error (RMSE) of the positioning solution
โ Along the horizontal plane :
๐ ๐๐๐ธ๐ป ๐ก = ๐ธ ๐๐ก๐๐ข๐ ๐ก โ ๐๐๐ ๐ก๐๐ ๐ก 2 + ๐๐ก๐๐ข๐ ๐ก โ ๐๐๐ ๐ก๐๐ ๐ก 2
โ Along the vertical axis: ๐ ๐๐๐ธ๐ง ๐ก = ๐ธ ๐๐ก๐๐ข๐ ๐ก โ ๐๐๐ ๐ก๐๐ ๐ก 2
68
Skyplot of the simulated scenario
Mask angle 50ยฐ ~4 floors building
Base Station localization of the simulated scenario
The French Civil Aviation University
La rรฉfรฉrence aรฉronautique
Ecole Nationale de lโAviation Civile
www.enac.fr
Rรฉf: Version: Date:
69/21
6. Navigation modules and results
69
6.3. Results (1/4)
โ Objectives: โข To study the performances of the EKF and UKF navigation solutions in order to define the more
appropriate tuning in term of characterization, synchronization module parameterization.
โข To study the impact of the navigation filter, of the measurement covariance matrix and ๐ถ/๐0 estimator over the best solution defined in the previous section.
โ Method:
โข Step 1: EKF and UKF tuning based on the study of: - Confidence bound impact - Approximation method impact - Synchronization module impact
โข Step 2: Impact analysis over the best identified solution - Navigation filter - Measurement error characterization - ๐ถ/๐0 estimator
The French Civil Aviation University
La rรฉfรฉrence aรฉronautique
Ecole Nationale de lโAviation Civile
www.enac.fr
Rรฉf: Version: Date:
70/21
6. Navigation modules and results
70
6.3. Step 1: EKF and UKF tuning (2/4) โ Confidence bound impact
โข To study the impact of the confidence bound used for the characterization on the filter performances
99.9% confidence bound recommended
โ Approximation method impact
โข To study the impact of the characterization method chosen on the filter performances
Overbounding method recommended
โ Synchronization module impact
โข To study the impact of the DLL loop bandwidth: 10 Hz or 1 Hz when no FLL is considered
โข To study the impact of the FLL for a 1 Hz DLL loop bandwidth
1 Hz DLL loop bandwidth and a FLL is recommended
โ Best tuning:
โข 99.9% overbounding characterization with a 1 Hz DLL loop bandwidth and a FLL
RMSE (m) X Y Z Xยฒ+Yยฒ
EKF GNSS 0.47 1.10 2.66 1.20
EKF 5G 0.61 0.81 4.64 1.01
UKF 5G 0.67 0.78 17.09 1.03
EKF HYBRID 0.49 0.64 2.96 0.81
UKF HYBRID 0.43 0.38 0.75 0.58
Synchronization module impact โ horizontal RMSE
Optimal solution performances
10 Hz DLL No FLL
1 Hz DLL No FLL
1 Hz DLL FLL
The French Civil Aviation University
La rรฉfรฉrence aรฉronautique
Ecole Nationale de lโAviation Civile
www.enac.fr
Rรฉf: Version: Date:
71/21
6. Navigation modules and results
71
6.3. Step 2: Impact analysis over the best solution (3/4) โ Impact of the navigation filter
โข Observation:
โ For 5G standalone solution, there are no significant differences between the UKF and the EKF
โ The vertical RMSE is much higher than the horizontal RMSE, a predictable behavior since all BSs are at the same height (bad Z axis geometry)
โ The hybrid navigation solution provides better solution than the standalone navigation solutions.
โข Conclusion:
โ The use of the hybrid navigation solution is recommended
โ The use of the UKF is recommended compared to the EKF.
Impact of the navigation filters used over the optimal solution
Number of measures
The French Civil Aviation University
La rรฉfรฉrence aรฉronautique
Ecole Nationale de lโAviation Civile
www.enac.fr
Rรฉf: Version: Date:
72/21
6. Navigation modules and results
72
6.3. Step 2: Impact analysis over the best solution (4/4) โ Impact of the measurement error characterization
โข To determine the improvements brought by an accurate measurement covariance matrix
โ Impact of the ๐ถ/๐0 estimator
โข To determine the impact of an estimation of the ๐ถ/๐0
โ Observations:
โข An accurate measurement covariance matrix improves the filter performances
โข A estimation of the ๐ถ/๐0 degrades the filter performances
โข The improvements brought by the accurate covariance matrix are erased by an inaccurate estimate of the ๐ถ/๐0
Characterization impact RMSE along the horizontal plane
Accurate covariance matrix Constant covariance matrix
๐ช/๐ต๐ estimator impact โ RMSE along the horizontal plane
True ๐ถ/๐0 Arctangent ๐ถ/๐0 FLL driven ๐ถ/๐0
The French Civil Aviation University
La rรฉfรฉrence aรฉronautique
Ecole Nationale de lโAviation Civile
www.enac.fr
Rรฉf: Version: Date:
73/21
6. Navigation modules and results
73
6.4. Conclusions โ Identification of the optimal EKF and UKF tuning
โข 99.9% overbounding characterization with a 1 Hz DLL loop bandwidth and a FLL
โ Impact analysis
โข Impact of the navigation filter
โ The use of the hybrid navigation solution is recommended
โ The use of the UKF is recommended compared to the EKF.
โข Impact of the measurement error characterization
โ An accurate approximation of the pseudo range measurement errors improves the performances of the navigation filters
โข Impact of the ๐ถ/๐0 estimator
โ The UKF hybrid navigation solution are degraded by approximately 40 cm with the use of the NWPR ๐ถ/๐0 phase corrected estimator along the horizontal plane.
The French Civil Aviation University
La rรฉfรฉrence aรฉronautique
Ecole Nationale de lโAviation Civile
www.enac.fr
Rรฉf: Version: Date:
74/21
Outline
1. 5G signal presentation
2. Propagation channel
3. Correlator output mathematical models
4. 5G ranging module design
5. Derivation of pseudo range error distributions
6. Navigation modules and results
7. Conclusion
74
The French Civil Aviation University
La rรฉfรฉrence aรฉronautique
Ecole Nationale de lโAviation Civile
www.enac.fr
Rรฉf: Version: Date:
75/21
7. Conclusions (1/4)
75
โข Main contributions
โ A description of the 5G physical layer, and notably the Synchronization Signal: SS PBCH
โ The selection and mathematical model definition of a 5G compliant propagation channel. Methods have been designed to study the evolving parameters of the propagation channel and to select the required propagation channel sampling rate generation.
โ The development of mathematical models of 5G correlator outputs, taking into account the effects of non-constant propagation channel parameters, noise.
The French Civil Aviation University
La rรฉfรฉrence aรฉronautique
Ecole Nationale de lโAviation Civile
www.enac.fr
Rรฉf: Version: Date:
76/21
7. Conclusions (2/4)
76
โข Main contributions
โ The proposition of a simplified model of 5G correlator outputs, which is valid in a LOS scenario for sub-6 GHz carrier frequency
โ The proposition of a statistical model for the interference term, which can be generate as a Gaussian variable with a mean and a variance derived in the study
โ A synchronization module based on a DLL and on a potential FLL has been derived. The synchronization module also implements a ๐ช/๐ต๐ estimator.
The French Civil Aviation University
La rรฉfรฉrence aรฉronautique
Ecole Nationale de lโAviation Civile
www.enac.fr
Rรฉf: Version: Date:
77/21
7. Conclusions (3/4)
77
โข Main contributions
โ The development of a simulator, coded in C, providing ranging measurements from GNSS and 5G tracking loops, taking into account simulated multipath propagation channels
โ The development of methods to characterize GNSS and 5G ranging measurement errors in urban environment.
โ The proposition of Gaussian modeling of ranging measurements as a function of ๐ช/๐ต๐
โ The study of simulated positioning performances using different navigation filters (EKF, UKF) and various parameters.
The French Civil Aviation University
La rรฉfรฉrence aรฉronautique
Ecole Nationale de lโAviation Civile
www.enac.fr
Rรฉf: Version: Date:
78/21
7. Conclusions (4/4)
78
โข Future work โ Real data collect
โ Correlator output mathematical model
โข The interference term of the correlator output mathematical model for NLOS scenario must be completed; therefore, the pseudo range measurements quality will be degraded and consequently, the positioning solution obtained will be less accurate.
โ Pseudo range error characterization
โข To derive the characterization for the NLOS scenario
โข To derive the characterization for mmW
โ Navigation solution
โข The development of Particle Filters could be implemented. The main advantage of PF is that they do not rely on a Gaussian PDF approximation contrary to Kalman Filters; an a-priori information on the distributions must be provided, and such information has been derived in this study.
โข The GNSS and 5G pseudo range measurements models used in the navigation filter could be refined by adding the time shifts between the transmitters and the reference time (GNSS or 5G) which are considered as perfectly corrected in this PhD
The French Civil Aviation University
La rรฉfรฉrence aรฉronautique
Ecole Nationale de lโAviation Civile
www.enac.fr
Rรฉf: Version: Date:
79/21
The French Civil Aviation University
La rรฉfรฉrence aรฉronautique
Ecole Nationale de lโAviation Civile
www.enac.fr
79
79
Thank you! Any questions?
The French Civil Aviation University
La rรฉfรฉrence aรฉronautique
Ecole Nationale de lโAviation Civile
www.enac.fr
Rรฉf: Version: Date:
80/21
References [1] K.Shamaei, J.Khalife, Z.M. Kassas, โExploiting LTE Signals for Navigation: Theory to Implementationโ, 2018
[2] C.Mensing, S.Plass, A.Dammann, โSynchronization algorithms for positioning with OFDM communciations signalsโ, 2007
[3] C.Mensing, A. Dammann, โPositioning with OFDM based communications systems and GNSS in critical scenariosโ, 2008
[4] E.Staudinger, โA generic OFDM based TDoA positioning testbed with interference mitigation for subsample delay estimationโ, 2011
[5] P.Thevenon, D.Serant, O.Julien, C.Macabiau, โPositioning using mobile TV based on DVB-SH standardโ, 2014
[6] D.Serant, P.Thevenon, O.Julien, C.Macabiau, โDevelopment and validation of an OFDM/DVB-T sensor for positioningโ, 2010
[7] S. Jaeckel, L. Raschkowski, โQuaDRiGa: A 3-D Multicell Channel Model with Time Evolution for Enabling Virtual Field Trials," 2014.
[8] 3GPP, "TS 38.211 Physical channels and modulation".
[9] J-J Van de Beek, M. sandell, P.O. Bรถrjesson, โML estimation of Time and Frequency Offset in OFDM Systemsโ, 1997
[10] D. Serant, D. Kubrak, M. Monnerat, G. Artaud et L. Ries, ยซField test performance assessment of GNSS/INS ultra-tight coupling scheme targeted to mass-market applications,ยป chez 2012 6th ESA Workshop on Satellite Navigation Technologies (Navitec 2012)
[11] E. Kaplan and C. Hegarty, Understanding GPS/GNSS: Principles and Applications, 2017.
[12] M. Ait Ighil, J. Lemorton and al, "SCHUN - A hybrid and mobil satellite channel simulator enhanced for multipath modelling applied to satellite navigation systems," 7th European Conference on Antennas and Propagation (EuCAP), 2013.
[13] A. Tobie, A. Garcia-Pena, P. Thevenon and al, "Processed 5G Signals Mathematical Models for Positioning considering a Non-Constant Propagation Channel.," in VTC 2019-Fall2019 IEEE 90th Vehicular Technology Conference, Honolulu, Sep 2019.
[14] A. Tobie, A. Garcia-Pena, P. Thevenon and al, "Mathematical Models and Statistics of Processed 5G Signals for Ranging Based Positioning for a Realistic Propagation Channel," Navigation Solution, submitted in September 2020
80
The French Civil Aviation University
La rรฉfรฉrence aรฉronautique
Ecole Nationale de lโAviation Civile
www.enac.fr
Rรฉf: Version: Date:
81/21
References [15] ITU ยซ Setting the scene for 5G: opportunities and challenges ยป - 2018
81