The 5G Localisation Waveform

Ronald Raulefs, Armin Dammann, Thomas Jost, Michael Walter, Siwei ZhangGerman Aerospace Center (DLR)

ETSI Workshop on Future Radio Technologies

27-28 January 2016

Our Goal: Ubiquitous Navigation

Where GPS fails, we likely have dense mobile radio networks!

> The 5G Localisation Waveform > Ronald Raulefs, Armin Dammann, Thomas Jost, Michael Walter, Siwei Zhang > 27-28 January 2016DLR.de • Chart 2

Today‘s Mobile Radio Positioning

Propagation Delay Based Positioning

BS2MT3

BS3

BS1

MT2

MT1

„Time Difference of Arrival“ (TDOA) Requires 3 base stations

Propagation Model (BS MT)

Path Loss

Shadow Fading

> The 5G Localisation Waveform > Ronald Raulefs, Armin Dammann, Thomas Jost, Michael Walter, Siwei Zhang > 27-28 January 2016DLR.de • Chart 3

What 5G Can Offer for Positioning• Higher FrequenciesConnectivity range decreases But if connected, it‘s LoS

• High Signal BandwidthsHigher ranging accuracy Less vulnerable against multipath propagation

• Device-to-Device (D2D) Communications Additional ranging measurements

• Much more 5G devices than todayHigh number of devices in com rangeNumber of range measurements grows quadratically

# BSMT links: NMT NBS# D2D links (max): NMT (NMT-1)

to solve for

3 NMT unknowns (xm, ym, Tm)

Cramer Rao Lower Bound for ranging

equivalent signal bandwidth

> The 5G Localisation Waveform > Ronald Raulefs, Armin Dammann, Thomas Jost, Michael Walter, Siwei Zhang > 27-28 January 2016DLR.de • Chart 4

Cooperative Positioning

BS2MT3

BS3

BS1

MT2

MT1

Centralized, Network Centric• MTs transmit to a central

positioning unit. • Central unit calculates position

solutions for the MTs jointly.

Decentralized, MT Driven• Each MT calculates its own

position based on its ownobservations.

• MTs share their estimates(position and time base offset)

Works even if there are less than 3 base stations hearable!

> The 5G Localisation Waveform > Ronald Raulefs, Armin Dammann, Thomas Jost, Michael Walter, Siwei Zhang > 27-28 January 2016DLR.de • Chart 5

Performance EvaluationNon Coop. Positioning Accuracy

• Random distribution of a MT in triangular area between 3 BSs

If the channel is in NLoS:

• Best case (solid lines): Consider as LoS (no NLoS bias)

• Worst case (dashed lines): Not connected

There is a large outage!

dBS

outage probability: 82%

> The 5G Localisation Waveform > Ronald Raulefs, Armin Dammann, Thomas Jost, Michael Walter, Siwei Zhang > 27-28 January 2016DLR.de • Chart 6

worst case

best case

Performance ResultsCoop. Positioning Accuracy, Outage Probability

approx. 3000 footballers per km²

> The 5G Localisation Waveform > Ronald Raulefs, Armin Dammann, Thomas Jost, Michael Walter, Siwei Zhang > 27-28 January 2016DLR.de • Chart 7

1000 m²

1000 MTs per km² means 1 MT per 1000 m²

worst case

best case

5G Localization WaveformsRanging Error Evaluation

> The 5G Localisation Waveform > Ronald Raulefs, Armin Dammann, Thomas Jost, Michael Walter, Siwei Zhang > 27-28 January 2016DLR.de • Chart 8

Cramer-Rao Lower Bound (CRLB)

with squared equivalent bandwidth

Is tight for high SNR and does not account forthe „threshold effect“. Therefore we use the

Ziv-Zakai Lower Bound (ZZLB)

with the localization signal’s autocorrelation function φ(τ) and the Gaussian Q-function

Example:

Synchronization with LTE Primary Synchronization Sequences (PSS),

CRLB is tight for high SNRs Simulation results show threshold effect

5G Localization WaveformsTriangular Waveform

> The 5G Localisation Waveform > Ronald Raulefs, Armin Dammann, Thomas Jost, Michael Walter, Siwei Zhang > 27-28 January 2016DLR.de • Chart 9

Power Spectrum Density

Autocorrelation

• ZZLB accounts for “threshold effect” ZZLB and CRLB diverge

• α = 1 provides the optimum

Ranging Error Performance

5G Localization WaveformsDircac-Rectangular Waveform

> The 5G Localisation Waveform > Ronald Raulefs, Armin Dammann, Thomas Jost, Michael Walter, Siwei Zhang > 27-28 January 2016DLR.de • Chart 10

Power Spectrum Density Ranging Error Performance

Autocorrelation

• Provides maximum squared equivalent bandwidth β2 = B2/4 for γ = 1.

• Optimum γ for different SNR ranges

5G Localization WaveformsDolph-Chebyshev Waveform

> The 5G Localisation Waveform > Ronald Raulefs, Armin Dammann, Thomas Jost, Michael Walter, Siwei Zhang > 27-28 January 2016DLR.de • Chart 11

Power Spectrum Density Ranging Error Performance

Autocorrelation

• Parameter a controls ACF sidelobeattenuation

• Optimum a for different SNR ranges

5G envisages properties which are beneficial for cooperative positioning• D2D, M2M communication is the big step for positioning• Our example: 1000…1500 Devices per km² provide sub-meter accuracy

5G localisation waveform design for optimal ranging/positioning performance• Power spectrum density form determines ranging performance• Higher (equivalent) signal bandwidth better performance at high SNRs• Take care about threshold effect choose appropriate sidelobe suppression

5G Positioning?

Yes, 5G can!With promising positioning performance achieved by

optimized localisation waveforms & cooperative positioning

Conclusion

> The 5G Localisation Waveform > Ronald Raulefs, Armin Dammann, Thomas Jost, Michael Walter, Siwei Zhang > 27-28 January 2016DLR.de • Chart 12