Smart mm-Wave Beam Steering Algorithm for Fast Link Re-Establishment under Node Mobility in 60 GHz Indoor WLANs Avishek Patra, Ljiljana Simić and Petri Mähönen Institute for Networked Systems, RWTH Aachen University, Aachen, Germany
Feb 08, 2017
Smart mm-Wave Beam Steering Algorithm for Fast Link Re-Establishment under Node Mobility in
60 GHz Indoor WLANsAvishek Patra, Ljiljana Simić and Petri Mähönen
Institute for Networked Systems, RWTH Aachen University, Aachen, Germany
OUTLINE
INTRODUCTION TO MM-WAVE NETWORKS
BEAM STEERING PROBLEM
MOTIVATION FOR SMARTER SOLUTION
60 GHz CONNECTIVITY PRE-STUDY
SMART MM-WAVE BEAM STEERING ALGORITHM
SIMULATION SCENARIOS
RESULTS
CONCLUSIONS & FUTURE WORKS
INTRODUCTION TO MM-WAVE NETWORKS
• large unlicensed spectrum in mm-wave bands, e.g. 60 GHz
• exploiting mm-wave bands for multi-Gbps wireless connectivity in WLAN, e.g. IEEE 802.11 ad
FILLER
• Challenges:
• high signal attenuation inherent at mm-wave frequencies!
FILLER
• Solution:
• highly directional beamforming antennas
⇒ increase transmission range 1
• But in mm-Wave WLAN…
1) directional link formation = only when Tx and Rx antenna sectors are both steered in correct directions
2) sector misalignments OR signal interruption ⇒ link breakage
3) node mobility ⇒ more link breakagesFILLER
FILLER
⇒ link establishment and maintenance much more challenging using directional antennas cf. vs. traditional omnidirectional antennas
2
CONTD.INTRODUCTION TO MM-WAVE NETWORKS
• low latency, fast beam steering to re-establish links essential for seamless connectivity and maintaining QoS
FILLER
• State of the Art:
Simple exhaustive sequential scanning of Tx and Rx antenna sectors, e.g. in IEEE 802.11 ad
FILLER
• Our work:
Smart beam steering algorithm with reduced Tx-Rx sector pair search space
3
CONTD.INTRODUCTION TO MM-WAVE NETWORKS
if RSS > Threshold for given AP-UE sector pair:
⇒ link established
⇒ “feasible sector pairs”
e.g. S and S
4
...
user equipment (UE)
S4
S3S2S1
S I
S i ...
S1
S JS j
S3 S2S4
...
i
j
APΘ = 360°I
UEΘ = 360°J
APΘ
UEΘ...
Pair = {S , S }3 4
access point (AP)
AP, f UE, fFeasible sector
BEAM STEERING PROBLEM
3 4
• links may be LOS or NLOS
FILLER
depends on the material properties of the surrounding indoor environment
FILLER
• a given AP-UE pair may have multiple feasible sector pairs
BEAM STEERING PROBLEM
5
8
UE1
UE3
UE2
5
6
4
AP
LOS = if no blockage and close enough
NLOS = reflected signals, penetrations
CONTD.
⇒
BEAM STEERING PROBLEM
• (re-) establishing link = searching until a feasible sector pair is found
FILLER
• link (re-) establishment latency ∝ # sector pairs searched before formation of link
FILLER
• Existing proposals: Exhaustive sequential scanning (# sector pairs searched = total # sector pairs), e.g. IEEE 802.11 ad
• Finds optimal feasible sector pair
• But… high latency (∝ total # sector pairs)
6
CONTD.
MOTIVATION FOR SMARTER SOLUTION
• link re-establishment latency increases with increase of antenna directionality
• especially under node mobility conditions FILLER
⇒ highly detrimental to QoSFILLER
Our aim: maintaining seamless connectivity and QoS in 60 GHz WLAN requires frequent faster beam re-steering methodsFILLER
Our work: develop faster beam steering algorithm by smartly restricting feasible sector pair search space
7
SMART MM-WAVE BEAM STEERING ALGORITHM
• Algorithm idea…
“Smart beam steering algorithm for link re-establishment that searches for a new feasible sector pair over a reduced search space in the vicinity of the previously known valid sector orientation (previous feasible sector pair).”
FILLER
⇒ use of historical information, i.e. previous feasible sector pair
FILLER
⇒ reduced search in vicinity of previous feasible sector pairFILLER
8
SMART MM-WAVE BEAM STEERING ALGORITHM
• idea based on a look at the 60 GHz connectivity… 9
(a) Exhaustive sequential scan (b) Our WorkPrevious feasible
sector pairs
• study AP-UE link formation in indoor scenarios
• determining all feasible sector pairs between AP and UEFILLER
o indoor layouts with realistic material properties (for 60 GHz), area of 10 x 10 m2
o AP – centrally located, UE – different locations at every 1 m through indoor layouts
o ray-tracing signal propagation simulation using WinProp
o simulations done for every AP-UE sector pair and every UE location in the indoor layouts
10
60 GHz CONNECTIVITY PRE-STUDY
FILLER
60 GHz CONNECTIVITY PRE-STUDY CONTD.
Indoor layouts (1) free space, (2) home,
(3) office, and (4) conference hall
Transmission power 0 dBm
Antenna gain (AP + UE) 25 dBi
Receiver sensitivity threshold – 78 dBm
AP Beamwidth Case 1: 30⁰ ; Case 2: 10⁰
UE Beamwidth Case 1: 30⁰ ; Case 2: 90⁰
11
60 GHz CONNECTIVITY PRE-STUDY
-20-30
-40
-50
-60
-70-80
Received Power[dBm]
12
CONTD.
-20-30
-40
-50
-60
-70-80
Received Power[dBm]
-20-30
-40
-50
-60
-70-80
Received Power[dBm]
-20-30
-40
-50
-60
-70-80
Received Power[dBm]
1. free space
3. office
2. home
4. conference hall
0 2 4 6 8 100
1
2
3
4
5
6
7
8
9
10
0 2 4 6 8 100
1
2
3
4
5
6
7
8
9
10
60 GHz CONNECTIVITY PRE-STUDY CONTD.
13
Case 1: AP – 30⁰; UE – 30⁰ Case 2: AP – 10⁰; UE – 90⁰
• RSS at UE > Receiver sensitivity threshold ⇒ link established• Complete set of feasible sector pairs for home layout:
* 1 hop = 1 m
• studying beam steering requirements (from previous to new feasible sector pair) for 1–3 hop* UE movements
FILLER0 5 10 15 200
0.2
0.4
0.6
0.8
1
CDF
homeAP+UE
1 – Hop 2 – Hop 3 – Hop 1 – Hop 2 – Hop 3 – Hop
(θ 30 ;θ 30 )AP UE
(θ 10 ;θ 90 )AP UE (θ 10 ;θ 90 )AP UE
(θ 10 ;θ 90 )AP UE (θ 30 ;θ 30 )AP UE
(θ 30 ;θ 30 )AP UE
homeAP+UE98%-ile case
60 GHz CONNECTIVITY PRE-STUDY CONTD.
14
• avg. total (AP+UE) beam steering requirement (1-hop movements) for 98% cases ≈ 6
FILLER
• ‘insight’ for smart beam steering algorithm based on reduced search around previous feasible sector pair
Average beam steering requirement
CDF
SMART MM-WAVE BEAM STEERING ALGORITHM
• Algorithm details…• start new feasible sector pair search for around previous pair
using a reduced search width parameter
• reduced search width parameter = combined search width for AP and UE
• individual search width for AP & UE 1/∝ UE & AP beamwidth respectively
• reduced search sector pairs arranged ∋ sector pairs requiring least movement searched first ⇒ ensures coordination
• if feasible sector pair not found within reduced space, retort to exhaustive sequential scan for unchecked sector pairs
15
CONTD.
SMART MM-WAVE BEAM STEERING ALGORITHM
16
CONTD.
= 6 for this
work
Initialize reduced search width parameter
Compute AP & UE search widths
Obtain & sort reduced search sector pairs
Select first reduced search sector pair
For selected sector pair,
RSS > threshold? Select next
reduced search
sector pair
All reduced search sector pairs
checked?
NO
NO
Obtain unchecked sector pairs(all sector pairs – reduced search sector pairs)
Select first unchecked sector pair
For selected sector pair,
RSS > threshold? Select next unchecked
sector pair
YES
All unchecked sector pairs
checked?NO
NO
No link for given AP & UE locations
YES
New feasible sector pair found
YES YES
SIMULATION SCENARIOS
• different indoor layouts – home, office, and conference hall
• mobility simulation through ‘walks’ (AP static, UE moved by 1-hop at a time)
• straight walks and random walks
• for random walks, orientation-unaware UEs and orientation-aware UEs• orientation-unaware UEs – previous feasible sector pair info.
corrupted at turnings
• orientation-unaware UEs – no ambiguity about previous feasible sector pair
17
0 1 2 3 4 5 6 7 8 9 100
1
2
3
4
5
6
7
8
9
10
0 1 2 3 4 5 6 7 8 9 100
1
2
3
4
5
6
7
8
9
10
SIMULATION SCENARIOS
18
0 1 2 3 4 5 6 7 8 9 100
1
2
3
4
5
6
7
8
9
10homewalks
officewalks
conference hall walks
CONTD.
RESULTS
• Performance metrics:
1. Search space and latency reduction – comparing # sector pairs searched vs. total # sector pairs
2. Link optimality – comparing RSS for optimal feasible sector pair and selected feasible sector pair
FILLER
• Results:
A. Straight walk in home layout
B. Random walk in home layout
C. Overall (all straight and random walks in all layouts)
19
A. Straight walk
Home layout:
0 1 2 3 4 5 6 7 8 9 100
1
2
3
4
5
6
7
8
9
10
Case 1: AP – 30⁰ UE – 30⁰
RESULTS
0 1 2 3 4 5 6 7 8 9 100
1
2
3
4
5
6
7
8
9
10
CONTD.
20
Received power using optimal sector pair, RSS opt
Received power using selected sector pair, RSS SBS
Minimum received power threshold,
Selected sector pair search space size,|P|SBS
Complete sector pair search space size, |M| ( |M|= |P| )EX
RESULTS
21
CONTD.
0 1 2 3 4 5 6 7 8 9 100
1
2
3
4
5
6
7
8
9
10
0 1 2 3 4 5 6 7 8 9 100
1
2
3
4
5
6
7
8
9
10
A. Straight walk
Home layout:
Case 2: AP – 10⁰ UE – 90⁰
Received power using optimal sector pair, RSS opt
Received power using selected sector pair, RSS SBS
Minimum received power threshold,
Selected sector pair search space size,|P|SBS
Complete sector pair search space size, |M| ( |M|= |P| )EX
RESULTS
A. Straight walk in home layout:
22
Avg. search reduction* Avg. RSS difference **Case 1 90% ~ 0.03 dBCase 2 75% ~ 0.00 dB
* Total # of sector pairs = 144 (both cases)** RSS (optimal sector pair) – RSS (selected sector pair)
CONTD.
RESULTS
23
CONTD.
0 1 2 3 4 5 6 7 8 9 100
1
2
3
4
5
6
7
8
9
10
0 1 2 3 4 5 6 7 8 9 100
1
2
3
4
5
6
7
8
9
10
B. Random walk
Home layout:
Selected sector pair search space size,|P| , for direction– unaware UE
SBS
Received power using optimal sector pair, RSS opt
Received power using selected sector pair, RSS , for direction–unaware UE
SBS
Received power using selected sector pair, RSS , for direction–aware UE
SBS
Minimum received power threshold,
Selected sector pair search space size,|P| , for direction–aware UE
SBS
Complete sector pair search space size, |M| ( |M|= |P| )EX
Case 1: AP – 30⁰ UE – 30⁰
RESULTS
24
CONTD.
0 1 2 3 4 5 6 7 8 9 100
1
2
3
4
5
6
7
8
9
10
0 1 2 3 4 5 6 7 8 9 100
1
2
3
4
5
6
7
8
9
10
Selected sector pair search space size,|P| , for direction– unaware UE
SBS
Received power using optimal sector pair, RSS opt
Received power using selected sector pair, RSS , for direction–unaware UE
SBS
Received power using selected sector pair, RSS , for direction–aware UE
SBS
Minimum received power threshold,
Selected sector pair search space size,|P| , for direction–aware UE
SBS
Complete sector pair search space size, |M| ( |M|= |P| )EX
B. Random walk
Home layout:
Case 2: AP – 10⁰ UE – 90⁰
RESULTS
B. Random walk in home layout:
max. RSS diff. between orientation unaware and aware UE = 0.17 dB
25
Avg. search reduction Avg. RSS differenceOrientation unaware UEs Case 1 75% ~ 1.44 dB Case 2 83% ~ 0.00 dBOrientation aware UEs Case 1 86% ~ 1.44 dB Case 2 86% ~ 0.00 dB
CONTD.
C. Overall result (search space and latency reduction)
1 2 3 4 5 60
30
60
90
120
150
Conference Hall
Home Office Home Office Conference Hall
θ 30AP θ 30UE θ 10AP θ 90UE
Walk AWalk BWalk CWalk DOverall Average |P|Average |P| θ 30AP θ 30UE ( ; )Average |P| θ 10AP θ 90UE ( ; )|P
|
SBS
SBS
|P|EX
SBS
RESULTS
26
CONTD.
Avg. reduction = 86% (7-fold)
Avg. reduction (Case 1) = 89% (7-fold)
Avg. reduction (Case 2) = 83% (7-fold)
Worst case reduction = 66% (3-fold)
C. Overall result (link optimality)
RESULTS
27
CONTD.
1 2 3 4 5 60
0.03
0.06
0.09
0.12
0.15
0.18
Conference Hall
Home Office Home Office Conference Hall
θ 30AP θ 30UE θ 10AP θ 90UE
Walk AWalk BWalk CWalk DOverall Average (RSS - RSS )
(RSS
- RSS
)[d
B]
opt SBS
opt
SBS
0.03
0.06
1.20
1.50
1.80 Avg. RSS diff. = 0.02 dB
Avg. RSS diff. (Case 1) = 0.02 dB
Avg. RSS diff. (Case 2) = 0.02 dB
Worst case RSS diff. = 1.44 dB
CONCLUSIONS & FUTURE WORKS
• low latency, fast beam steering algorithm that smartly reduces feasible sector pair search space
• search limited based on (static) reduced search width parameter
• 7-fold (avg.) / 3-fold (worst case) reduction in search space and link re-establishment latency
• re-established links nearly optimal (avg. RSS diff. < 0.03 dB)
• incorporation of adaptive reduced search width parameter
• performance in outdoor scenarios28