[PPT] - Multiuser Positioning Ronald Raulefs, Siwei Zhang, Wei Wang DLR: PowerPoint Presentation

SLIDE 1

Multiuser Positioning Ronald Raulefs, Siwei Zhang, Wei Wang DLR: www.dlr.de/kn WHERE2 Project: www.ict-where2.eu

www.DLR.de • Chart 1 > Multiuser Positioning > Ronald Raulefs • 4th COST Workshop > 09.10.2013

SLIDE 2

Motivation: Indoor Navigation

www.DLR.de • Chart 2 > Multiuser Positioning > Ronald Raulefs • 4th COST Workshop > 09.10.2013

Base station Base station

SLIDE 3

Motivation: Indoor Navigation

www.DLR.de • Chart 3 > Multiuser Positioning > Ronald Raulefs • 4th COST Workshop > 09.10.2013

Base station Base station

SLIDE 4

Motivation: Indoor Navigation

www.DLR.de • Chart 4 > Multiuser Positioning > Ronald Raulefs • 4th COST Workshop > 09.10.2013

Base station Base station

SLIDE 5

Motivation: Indoor Navigation

www.DLR.de • Chart 5

Base station Base station GNSS may be unavailable or insufficient, but devices are seldom alone Questions:

How to improve positioning?
How to range simultaneously?

SLIDE 6

Outline

Motivation
Cooperative positioning and related work
Multicarrier Ranging and Positioning Cramer-Rao Lower Bound
Multiuser positioning concept
Measurement evaluation
Conclusions & Outlook

www.DLR.de • Chart 6

SLIDE 7

Position Error in m 10 8 6 4 2 29m 22m

Core building wall with 10dB penetration loss Office wall with 3dB penetration loss

BS

1

BS

2

BS

3

Cooperative Positioning: Deep Indoors

www.DLR.de • Chart 7

Ranging variance: 1m

SLIDE 8

Related Work

www.DLR.de • Chart 8

Communication Standards – Device-2-Device:

WiFi-direct (Android [today])
LTE-direct (R12 discussion)
Coop. Pos. with link

selection in WSN (Wymeersch et al. [2012]) Multicarrier Ranging (Luise et al. [2009])

Resource Allocation in
Coop. Pos.
Multiuser D2D Ranging

SLIDE 9

Ranging Measuring the Wireless Channel

www.DLR.de • Chart 9

Received Signal Strength (RSS)

– Corrupted by the propagation effects

Time of Arrival (TOA)

– Requires synchronization between transmitter and receiver

0dBm
-80dBm
ta=3.4µs

SLIDE 10

Cramér-Rao Lower Bound (CRLB) for Ranging and Positioning with Multicarrier Signals

www.DLR.de • Chart 10

SLIDE 11

System Model

Distance between cooperating node and node of interest 1:

𝑒=||𝒚↓ 𝒚↓1 −𝒚| 𝒚||=√⁠(𝑦↓1 −𝑦)↑2 +(𝑧↓1 −𝑧)↑2

OFDM Signal:

s(t)=1/√⁠𝑂 ∑𝑜=− ¡𝑂/2 ↑𝑂/2 ▒𝑇↓𝑜 𝑓↑𝑘2𝜌nf↓𝑡𝑑 𝑢

Received signal under line-of-sight conditions (single slope path

loss model):

r(𝑢)=𝑏↓0 (𝑒/𝑒↓0 )↑−𝛿/2 𝑡(𝑢−𝑒/𝑑↓0 )+𝑨(𝑢)

www.DLR.de • Chart 11

𝑇↓𝑜 is the n’th signal sample of the transmitted multicarrier signal
‑ ¡𝑜 ¡ subcarrier index
‑ ¡𝑔↓𝑇𝐷 ¡subcarrier spacing
𝑂 number of subcarriers
𝛿 attenuation / path loss exponent (𝛿 ¡ = 2 for free space)
𝑏↓0 path loss at the reference distance 𝑒↓0
z Gaussian thermal noise
𝑑↓0 speed of light

SLIDE 12

Joint CRLB TOA and Received Signal Strength

CRLB(TOA) = 𝜏↑2 /𝑔↓𝑇𝐷↑2 /2 ¡𝑔↓𝑑↑2 𝑒↑2 ∑𝑜=−𝑂/2 ↑𝑂/2 ▒𝑜↑2 |𝑇(𝑜)|↑2
CRLB(RSS) = 𝜏↑2 /𝑑↓0↑2 /8 ¡𝜌↑2 𝑔↓𝑑↑2 𝑒↑4 ∑𝑜=−𝑂/2 ↑𝑂/2 ▒|𝑇(𝑜)|↑2
CRLB(RSS,TOA) = 𝜏↑2 /𝑑↓0↑2 /8 ¡𝜌↑2 𝑔↓𝑑↑2 𝑒↑4 ∑𝑜=−𝑂/2 ↑𝑂/2 ▒|𝑇(𝑜)|

↑2 + ¡𝑔↓𝑇𝐷↑2 /2 ¡𝑔↓𝑑↑2 𝑒↑2 ∑𝑜=−𝑂/2 ↑𝑂/2 ▒𝑜↑2 |𝑇(𝑜)|↑2

‑ 𝑔↓𝐷 ¡: Carrier frequency used to include free space path loss

www.DLR.de • Chart 12

Subcarrier index
Signal power
Bandwidth (𝑔↓𝑇𝐷 (𝑂+1))
𝑔↓𝐷 ¡: Carrier frequency
Signal power

SLIDE 13

Ranging Bounds for TOA and RSS – Path Loss Integrated

www.DLR.de • Chart 13

𝐾↓𝑈𝑃𝐵 =2𝑏↑2 /𝜏↑2 (𝑒/𝑒↓0 )↑−𝛿 4 ¡𝜌↑2 𝑔↓𝑇𝐷↑2 /𝑑↑2 ∑𝑜=−{𝑂−1/2 }↑{𝑂−1/2 }▒𝑜↑ 𝐾↓𝑆𝑇𝑇 =2𝑏↑2 /𝜏↑2 (𝑒/𝑒↓0 )↑−𝛿 𝛿↑2 /4𝑒↑2 ∑𝑜=−{𝑂−1/2 }↑{𝑂−1/2 }▒𝑇[𝑜]↑2 ¡ ¡ 𝐾↓𝐾𝑝𝑗𝑜𝑢 =𝐾↓𝑈𝑃𝐵 +𝐾↓𝑆𝑇𝑇 CRLB↓Joint = ¡1/𝐾↓𝐾𝑝𝑗𝑜𝑢 Fisher Information Matrix Cramer-Rao Lower Bound

SLIDE 14

Bandwidth affects Ranging CRLB

Different bandwidths:

More bandwidth

reduces CRLB

Parameters:
𝑄↓0 =−30 ¡𝑒𝐶𝑛/

𝑡𝑣𝑐𝑑𝑏𝑠𝑠𝑗𝑓𝑠

𝑔↓𝑇𝐷 =15 ¡𝑙𝐼𝑨
AWGN channel

www.DLR.de • Chart 14

SLIDE 15

Different Zones of the CRLB

www.DLR.de • Chart 15

Shorter distances favor RSS estimator à cooperative links! Triangular zone: Jointly RSS plus TOA – here for 20MHz

SLIDE 16

CRLB Positioning Bound

www.DLR.de • Chart 16

DOI10.1002/ett.2572 - Raulefs, Zhang, Mensing: “Bound-based spectrum allocation for cooperative positioning”

Mobility and limited ranging increases uncertainty
Simultaneous ranging possible?

Important aspects:

Ranging performance
Geometric constellation

SLIDE 17

LTE Frame Structure

Todays multicarrier based cellular

communication systems, such as LTE, use for synchronization and positioning:

Primary and Secondary synchronization

signal à Chunk of signals (low bandwidth)

Positioning Reference Signals (PRS)

are scattered over the available bandwidth à Grid of reference signals (scale with communication load)

Can we apply such a scheme to range

simultaneously multiple users?

What is the impact on the CRLB bound in case
nly limited subcarriers are used?

www.DLR.de • Chart 17

1MHz Primary synchronization signal Secondary synchronization signal Cell specific reference signal Positioning reference signal Data Control signals 5 ms LTE frame structure (10 subframes)

SLIDE 18

Multiuser Ranging - Sharing: Spectrum Allocation

1. Chunks

Subcarrier(f) Full band Nodes (spatial)

www.DLR.de • Chart 18

Chunk: Like Primary/ Secondary synchronization Signals in LTE

SLIDE 19

Multiuser Ranging - Sharing: Spectrum Allocation

2. Grid

Subcarrier(f) Nodes (spatial)

www.DLR.de • Chart 19

Grid: Like Positioning Reference Signals (PRS) in LTE Full band

SLIDE 20

RSS CRLB with Block Fading Channel

www.DLR.de • Chart 20

Block Fading Channel
Grid and Chunk scheme perform

similar

Full band uses more signal power
Channel is known!
1540 subcarriers (grid: every 77th
r chunk: 77 subcarriers)

20 40 60 80 100 10

4

10

3

10

2

10

1

10 10

1

Distance[m] CRLB[m] RSS CRLB Full RSS CRLB Grid 1 RSS CRLB Grid 2 RSS CRLB Grid 3 RSS CRLB Chunk 1 RSS CRLB Chunk 2 RSS CRLB Chunk 3

SLIDE 21

Using distance dependent path loss model to determine distance

𝑄(𝑒)=𝑏↓0 ¡(𝑒↓0 /𝑒 )↑−𝛿

If link budget (TX power, antenna gains, path loss, etc.) is known

à distance estimation possible Indoors:

LoS condition between devices is reasonable
Short ranges à RSS-CRLB outperforms TOA-CRLB

𝑄↓0 :

Averaged received power at distance 𝑒↓0

𝑒: ¡

Distance in m

𝛿=1.13

Environmental path-loss exponent (determined by measurement campaign)

Received Signal Strength Ranging

www.DLR.de • Chart 21

SLIDE 22

Received Signal Strength Ranging: Channel Snapshot

www.DLR.de • Chart 22

Full-band: 120MHz bandwidth Distance estimation using the chunk scheme:

high variance for medium

distances

lower variance for short

distances Distance estimation for grid and full band overlap à Coherence bandwidth ~20MHz à Grid = Sampled channel fits path loss model

SLIDE 23

5 10 15 20 0.5 1 1.5 2 2.5 Selected Chunk or Grid Scheme RMSE[m]

Received Signal Strength Ranging with Real Data

Path loss for single slope model: P(𝑒)=𝑏↓0 (𝑒↓0 /𝑒 )↑−𝛿 , ¡𝛿=1.13 120MHz channel bandwidth (1 track) Chunk: Mismatch compared to the derived path loss model Grid and full band approach fits fairly well (same performance as full band)

www.DLR.de • Chart 23

SLIDE 24

Conclusions

Cooperative positioning improves performance significantly (esp.

indoors)

Received signal strength (RSS) is a simple non-synchronous scheme

ranging scheme

Multiuser positioning concept:
Frequency division multiple access
Benefits for grid vs. chunk scheme to represent the broadband path

loss model

www.DLR.de • Chart 24

SLIDE 25

www.DLR.de • Chart 25 www.DLR.de • Chart 25 > Multiuser Positioning > Ronald Raulefs • 4th COST Workshop > 09.10.2013

Multiuser Positioning Ronald Raulefs, Siwei Zhang, Wei Wang DLR: www.dlr.de/kn WHERE2 Project: www.ict-where2.eu

Motivation: Indoor Navigation

Base station Base station

Motivation: Indoor Navigation

Base station Base station

Motivation: Indoor Navigation

Base station Base station

Motivation: Indoor Navigation

Base station Base station GNSS may be unavailable or insufficient, but devices are seldom alone Questions:

Outline

Position Error in m 10 8 6 4 2 29m 22m

Core building wall with 10dB penetration loss Office wall with 3dB penetration loss

BS

BS

BS

Cooperative Positioning: Deep Indoors

Ranging variance: 1m

Related Work

Communication Standards – Device-2-Device:

selection in WSN (Wymeersch et al. [2012]) Multicarrier Ranging (Luise et al. [2009])

Ranging Measuring the Wireless Channel

– Corrupted by the propagation effects

– Requires synchronization between transmitter and receiver

Cramér-Rao Lower Bound (CRLB) for Ranging and Positioning with Multicarrier Signals

System Model

𝑒=||​𝒚↓ 𝒚↓1 −𝒚| 𝒚||=√⁠​(​𝑦↓1 −𝑦)↑2 +​(​𝑧↓1 −𝑧)↑2

s(t)=​1/√⁠𝑂 ∑𝑜=− ¡​𝑂/2 ↑​𝑂/2 ▒​𝑇↓𝑜 ​𝑓↑𝑘2𝜌​nf↓𝑡𝑑 𝑢

loss model):

r(𝑢)=​𝑏↓0 ​(​𝑒/​𝑒↓0 )↑−​𝛿/2 𝑡(𝑢−​𝑒/​𝑑↓0 )+𝑨(𝑢)

Joint CRLB TOA and Received Signal Strength

↑2 + ¡​𝑔↓𝑇𝐷↑2 /2 ¡​𝑔↓𝑑↑2 ​𝑒↑2 ∑𝑜=−​𝑂/2 ↑​𝑂/2 ▒​𝑜↑2 |𝑇​(𝑜)|↑2

Ranging Bounds for TOA and RSS – Path Loss Integrated

Bandwidth affects Ranging CRLB

Different bandwidths:

reduces CRLB

𝑡𝑣𝑐𝑑𝑏𝑠𝑠𝑗𝑓𝑠

Different Zones of the CRLB

Shorter distances favor RSS estimator à cooperative links! Triangular zone: Jointly RSS plus TOA – here for 20MHz

CRLB Positioning Bound

DOI10.1002/ett.2572 - Raulefs, Zhang, Mensing: “Bound-based spectrum allocation for cooperative positioning”

Important aspects:

LTE Frame Structure

communication systems, such as LTE, use for synchronization and positioning:

signal à Chunk of signals (low bandwidth)

are scattered over the available bandwidth à Grid of reference signals (scale with communication load)

simultaneously multiple users?

1MHz Primary synchronization signal Secondary synchronization signal Cell specific reference signal Positioning reference signal Data Control signals 5 ms LTE frame structure (10 subframes)

Multiuser Ranging - Sharing: Spectrum Allocation

Subcarrier(f) Full band Nodes (spatial)

Chunk: Like Primary/ Secondary synchronization Signals in LTE

Multiuser Ranging - Sharing: Spectrum Allocation

Subcarrier(f) Nodes (spatial)

Grid: Like Positioning Reference Signals (PRS) in LTE Full band

RSS CRLB with Block Fading Channel

similar

𝑄(𝑒)=​𝑏↓0 ¡​(​𝑒↓0 /𝑒 )↑−𝛿

à distance estimation possible Indoors:

​𝑄↓0 :

Averaged received power at distance ​𝑒↓0

𝑒: ¡

Distance in m

𝛿=1.13

Environmental path-loss exponent (determined by measurement campaign)

Received Signal Strength Ranging

Received Signal Strength Ranging: Channel Snapshot

Full-band: 120MHz bandwidth Distance estimation using the chunk scheme:

distances

distances Distance estimation for grid and full band overlap à Coherence bandwidth ~20MHz à Grid = Sampled channel fits path loss model

Received Signal Strength Ranging with Real Data

Path loss for single slope model: P(𝑒)=​𝑏↓0 ​(​𝑒↓0 /𝑒 )↑−𝛿 , ¡𝛿=1.13 120MHz channel bandwidth (1 track) Chunk: Mismatch compared to the derived path loss model Grid and full band approach fits fairly well (same performance as full band)

Conclusions

indoors)

ranging scheme

loss model

𝑒=||𝒚↓ 𝒚↓1 −𝒚| 𝒚||=√⁠(𝑦↓1 −𝑦)↑2 +(𝑧↓1 −𝑧)↑2

s(t)=1/√⁠𝑂 ∑𝑜=− ¡𝑂/2 ↑𝑂/2 ▒𝑇↓𝑜 𝑓↑𝑘2𝜌nf↓𝑡𝑑 𝑢

r(𝑢)=𝑏↓0 (𝑒/𝑒↓0 )↑−𝛿/2 𝑡(𝑢−𝑒/𝑑↓0 )+𝑨(𝑢)

↑2 + ¡𝑔↓𝑇𝐷↑2 /2 ¡𝑔↓𝑑↑2 𝑒↑2 ∑𝑜=−𝑂/2 ↑𝑂/2 ▒𝑜↑2 |𝑇(𝑜)|↑2

𝑄(𝑒)=𝑏↓0 ¡(𝑒↓0 /𝑒 )↑−𝛿

𝑄↓0 :

Averaged received power at distance 𝑒↓0

Path loss for single slope model: P(𝑒)=𝑏↓0 (𝑒↓0 /𝑒 )↑−𝛿 , ¡𝛿=1.13 120MHz channel bandwidth (1 track) Chunk: Mismatch compared to the derived path loss model Grid and full band approach fits fairly well (same performance as full band)