The millimeter-wave (mm-wave) bands are currently being explored for multi-Gbps wireless local area networks (WLANs). Directional antennas are required to overcome the high attenuation inherent at the mm-wave frequencies. However, directionality makes link maintenance and establishment tasks complex, especially under node mobility, as slight misalignment of antenna beams between nodes leads to link disruption. Consequently, low latency beamsteering algorithms are needed for fast link re-establishment to support seamless data provisioning. Solutions based on exhaustive sequential scanning induce high latency, thereby disrupting communication. On the other hand, existing low latency proposals typically consider only static links, depend on additional hardware, or require a priori information about the network environment. In this paper, we propose a generic, fast mm-wave beamsteering algorithm that utilizes the previous valid link information to initiate the feasible antenna sector pair search and adaptively increases the sector search space around it to re-establish a link. Additionally, we experimentally evaluate the performance of our algorithm through measurements conducted in a real indoor environment using 60 GHz packet radio transceivers. The results show that, compared to exhaustive sequential scanning, our algorithm reduces the required sector search space, and thereby the link re-establishment latency, by 89% on average compared to exhaustive sequential scanning.
Experimental Evaluation of a Novel Fast Beamsteering Algorithm for Link Re-Establishment in mm-Wave Indoor WLANs
1. Experimental Evaluation of a
Novel Fast Beamsteering Algorithm for
Link Re-Establishment in mm-Wave Indoor WLANs
Avishek Patra, Ljiljana Simić and Marina Petrova
Institute for Networked Systems, RWTH Aachen University, Aachen, Germany
2. OUTLINE
INTRODUCTION TO MM-WAVE NETWORKS
BEAMSTEERING PROBLEM
FAST BEAMSTEERING ALGORITHM
EXPERIMENTAL EVALUATION METHODOLOGY
PERFORMANCE RESULTS & ANALYSIS
CONCLUSIONS & FUTURE WORKS
3. INTRODUCTION TO MM-WAVE NETWORKS
• mm-wave bands for multi-Gbps connectivity
FILLER
• Challenge: high attenuation!
FILLER
• Solution: directional antennas ⇒increase range
FILLER
• Further challenges:
1) directional link formation ⇒ only if TX & RX antennas
steered to feasible directions
2) link disruptions due to antenna misalignments, link
interruption, mobility, . . .
1
4. FILLER
• State of the Art:
exhaustive sequential scanning of Tx & Rx antenna sectors
⇒ e.g. exhaustive scanning-like algorithm in IEEE 802.11 ad
⇒ high latency, QoS degradation
FILLER
• Motivation:
low latency, fast beamsteering to re-establish links
essential for seamless connectivity and maintaining QoS
FILLER
2
CONTD.INTRODUCTION TO MM-WAVE NETWORKS
5. FILLER
• Solutions in literature:
• dependent on:
• focus only on static networks
FILLER
• Untackled problem:
• UE mobility ⇒ induces frequent link disruptions
FILLER
3
CONTD.INTRODUCTION TO MM-WAVE NETWORKS
additional hardware,
secondary link knowledge,
environmental information . . .
6. • Our work:
FILLER
• we propose a generic, fast beamsteering algorithm that:
doesn’t depend on extra hardware or information,
addresses UE mobility-induced disruption,
uses available last valid link information only!
FILLER
• performance evaluation in real indoor environment
using 60 GHz packet radio transceiver
4
CONTD.INTRODUCTION TO MM-WAVE NETWORKS
7. user
equipment
(UE)
S4
S3
S2S1
S I
S i
...
S4
S j
S2 S3
S1
...
i
j
UEΘ = 360°
J
Θ
...
AP
Pair 1 = {S , S }
3 1
AP, f UE, f
Feasible sector
Pair 2 = {S , S }
4 2
AP, f UE, f
Feasible sector
SJ
...
ΘUE
AP
AP
AP AP
AP
AP
UE
UE
UE
UE
UE
UE
access
point (AP)
user
equipment
(UE)
• if RSS > RX Sensitivity
Threshold for AP-UE sector:
FILLER
⇒ link established
FILLER
⇒ feasible sector pair
FILLER
e.g.
FILLER
• multiple feasible sector pairs
for a AP-UE location
5
BEAM STEERING PROBLEM
LOS Link NLOS Link
3 1 4 2{S ,S } & {S ,S }AP UE AP UE
8. BEAM STEERING PROBLEM
• re-establishing link ⇒ search until feasible sector pair found
FILLER
FILLER
FILLER
• for exhaustive sequential scanning
⇒ # sector pairs searched = total # sector pairs
• finds optimal sector pair . . . but high latency
• latency increases with antenna directionality increase
6
CONTD.
re-establishment
latency ∝ # of sector pairs
searched
9. FAST BEAMSTEERING ALGORITHM
• Algorithm idea: “…initiate search in vicinity of the previously
valid (i.e., before disruption) sector pair…”
FILLER
→ start search from the previous feasible sector pair
→ increase search space till new feasible sector pair found
→ AP-UE coordination? Before start, sort all sector pairs
such that sector pairs nearest to previous feasible sector
pair checked first
→ stop as soon as a new feasible sector pair found
FILLER
7
[1]
[1] – Patra et al. “Smart mm-Wave Beam Steering Algorithm for Fast Link Re-
Establishment under Node Mobility in 60 GHz Indoor WLANs”
10. 8
CONTD.
access point
(AP)
user equipment
(UE)
p1
p7p6
p5
p3
p4
p2
p8
q5
q4
q2
q7 q6
q8
q1
q3
UE mobility
causes link
disruption
p1
p8
p7p6
p5
p3
q5
q4
q2
q7 q6
q8
p4
p2
q1
q3
p1
p8
p7p6
p5
p3
q5
q4
q2
q7 q6
q8
p4
p2
q1
q3
M = [ {p0, q0},1 1 {p0, q0},2 1 {p0, q0},8 1
{p0, q0},1 2 {p0, q0},1 8 {p0, q0},3 1
{p0, q0},7 1 {p0, q0},2 2 {p0, q0},8 2
{p0, q0},2 8 {p0, q0},8 8 {p0, q0},1 3
{p0, q0}, . . .1 7
S
{p0, q0} ]5 5
Sorted
sector
pairs
Searched
sector pairs
New
feasible
sector pairs
Previous feasible AP sector
Previous feasible UE sector
Searched AP sector
Searched UE sector
Feasible AP-UE links
New feasible AP sector
New feasible UE sector
FAST BEAMSTEERING ALGORITHM
AP AP AP
UE
UE UE
Established link Link disrupted Link re-established→→
11. EXPERIMENTAL EVALUATION METHODOLOGY
9
experimental evaluation in real indoor environment
using 60 GHz packet radio transceivers
FILLER
• UE mobility along ‘walks’
• link disruption along walks ⇒ algorithm triggered
• re-establish link using (i) previous feasible sector pair
knowledge and (ii) derived feasible sector pair information
FILLER
RSS measured for various UE locations
⇒ AP-UE feasible sector pairs determined
12. Measurement
points
60 GHz AP60 GHz UE
EXPERIMENTAL EVALUATION METHODOLOGY
10
CONTD.
Indoor environment:
AP
T U VC
EFGH
I
J
K L M N O P
Q
R
SD
A
B
Concrete
Plasterboard
Brick
Glass
Wood
Metalized Glass
indoor environment with UE locations
(A – V) for RSS measurement
15. EXPERIMENTAL EVALUATION METHODOLOGY
AP
C
DEFGH
I
J
OK
R
Q
P
S
T
NL
13
CONTD.
Performance evaluation:
FILLER
• UE moves along walks ⇒ link disruption
• link re-establishment using:
1. previous feasible sector pair
2. feasible sector pairs (from RSS
measurement)
• algorithm evaluated in non-real time
⇒ real-time evaluation highly time
consuming ⇒ mechanical beamsteering
AP
A
B
C
DEFGH
I
J
ONMLK
R
Q
P
S
VUT
AP
A
B
C
ONM
R
Q
P
S
VUT
Walk I
Walk II
Walk III
16. PERFORMANCE RESULTS & ANALYSIS
• Performance metrics:
1. #. of sector pairs searched (|P|)
2. search space reduction* (in %)
3. RSS difference*
4. data rate
14
* Comparing our algorithm with exhaustive sequential scanning
Search space and
latency reduction
Re-established link
quality
17. 0 2 4 6 8 10 12 14 16 18 20 22
-90
-80
-70
-60
-50
-40
-30
UE Positions
RSS[dBm]
0 2 4 6 8 10 12 14 16 18 20 22
0
24
48
72
96
120
144
|P|
PERFORMANCE RESULTS & ANALYSIS
Walk I
15
Receiver sensitivity threshold
|P|
RSS
|P|
RSS
exhaustive sequential
scanning
fast beamsteering
algorithm
AP
A
B
C
DEFGH
I
J
ONMLK
R
Q
P
S
VUT
Avg. # sector pairs
searched = 15.6
Avg. search space
reduction = 84%
Avg. RSS
difference = drop
by 5.8 dB
CONTD.
18. 0 2 4 6 8 10 12 14 16 18 20 22
-90
-80
-70
-60
-50
-40
-30
UE Positions
RSS[dBm]
0 2 4 6 8 10 12 14 16 18 20 22
0
24
48
72
96
120
144
|P|
AP
A
B
C
ONM
R
Q
P
S
VUT
PERFORMANCE RESULTS & ANALYSIS
Walk II
16
CONTD.
Avg. # sector pairs
searched = 23.3
Avg. search space
reduction = 84%
Avg. RSS
difference = drop
by 6.3 dB
Receiver sensitivity threshold
|P|
RSS
|P|
RSS
exhaustive sequential
scanning
fast beamsteering
algorithm
19. 0 2 4 6 8 10 12 14 16 18 20 22
-90
-80
-70
-60
-50
-40
-30
UE Positions
RSS[dBm]
0 2 4 6 8 10 12 14 16 18 20 22
0
24
48
72
96
120
144
|P|
AP
C
DEFGH
I
J
OK
R
Q
P
S
T
NL
PERFORMANCE RESULTS & ANALYSIS
Walk III
17
CONTD.
Avg. # sector pairs
searched = 3.9
Avg. search space
reduction = 97%
Avg. RSS
difference = drop
by 9.3 dB
Receiver sensitivity threshold
|P|
RSS
|P|
RSS
exhaustive sequential
scanning
fast beamsteering
algorithm
20. PERFORMANCE RESULTS & ANALYSIS
• Overall results:
FILLER
FILLER
• link re-establishment latency highly reduced
• re-established link not always optimal ⇒ data rate reduces
• tradeoff between re-establishment latency and link quality!
18
CONTD.
# sector pairs
searched
search space
reduction
RSS
difference
Data rate*
Average 14 89% 07 dB (less) –
Worst case 24 83% 25 dB (less) 2.1 Gbps
* Considering IEEE 802.11 ad OFDM PHY
21. CONCLUSIONS & FUTURE WORKS
FILLER
• generic, low latency beamsteering algorithm; tackles UE
mobility issue
• experimental evaluation in indoor using 60 GHz packet radio
transceivers
• 89% (avg.) reduction in link search & re-establishment latency
• link RSS 7 dB (avg.) less than best case link RSS
⇒ worst case data rate of 2.1 Gbps
FILLER
• presently investigating trade-off between data rate drop and
link re-establishment latency
19
22. QUESTIONS?
For queries, please mail us at:
Avishek Patra avp@inets.rwth-aachen.de
Ljiljana Simić lsi@inets.rwth-aachen.de
Marina Petrova mpe@inets.rwth-aachen.de
20
Editor's Notes
- Basics of mm-wave networks.
- Do not spend much time in this.
Short but strong empahsis on how directional link establishment and beamforming is challenging.
... especially due to possibly easy link disruption chances due to, e.g. Human walking between nodes, slight chasnge of node position, etc.
Present algorithms (like in IEEE 802.11 ad) follows brute-force like mechanisms.
As expected, these mechanisms are simple but tedious and highly time consuming.
High delay in link establishment (after breakage) can degrage QoS, e.g. of video streaming.
We need smart algorithms that can swiftly determine new links and maintain good QoS.
Solutions in literature (for low latency solutions) are often ‘customized‘ and work at the cost of ceratin specfic condition fulfillment.
e.g. 1: Direction of 60 GHz node via signal reception from 2.4 GHz transmitter colocated with the 60 GHz.
e.g. 2: Pre-determination of all 60 GHz links such that when one link fails the nodes ‘know‘ which link to switch to.
No solution for link breakage due to steady UE movement, e.g. A person walking across a room with his smartphone.
We want to mainly tackle mobility induced disruption.
Also, we want a solution that doesn‘t require special a priori information.
Note that we previously had proposed a link re-establishment solution with a focus on node mobility... But the algorithm worked best with a priori information about the surrounding enviornmental layout, i.e. Building structure, etc. (in case of indoor deployment).
Back to basics! An elaboration on what is a ‘feasible sector pair´ – a sector pair or a pair of TX and RX antenna orientation for which when TX transmits signal through the feasible TX sector and the RX orients along the RX feasible sector, it receives signal from the TX AND the RSS is greater than minimum signal strength detectable by the RX.
One TX-RX pair for their respective positions can have multiple such feasible sector pairs as 60 GHz signal propagates via both LOS and NLOS (mainly reflected) paths. See figure.
Since we need a sector pair that is feasible, we need to ‘check‘ the different sector pair until we find one. Therefore, the delayin link determination is dependent on how many sector pairs we chacked that were NOT feasible till we find a feasible sector pair.
For exhaustive scanning brute force method, we check all sector pairs and don‘t stop till all possible pairs are checked and saving the feasible sector pairs found along the way. Then we chose the feasible sector pair that has highest RSS (optimal sector pair) as the feasible sector pair that re-establishes link.
This method has high latency as it checks through all sector pairs. Higher directionalty means higher no. of sector pairs means higher latency means bad network performance.
Simple concept! When link disrupts, start by checking the sector pair close to the last-valid sector pair.
For example, if a man uis walking with his phone in a room, his direction of walk doesn‘t change drastically and abruptly, but changes more gradually – this ensures that slight orientation changes from previous valid orientation would be enough to re-establish link.
Slowly the search radius is increased, in case the man has made a abrupt direction change.
Of course, this method will not be as effective, say, when the man walks round a corner to the next room. But still, it would be better to smartly check the sector pairs as in this alfgorithm, rather than checking every sector pair.
To minimize re-establishment delay, the algorithm stops as soon as a new feasible sector pair is found AND unlike exhaustive scanning, doesn‘t continue searching till the optimal sector apir is found. Advantage: Reduces latency, Disadvantage: Link not the best one possible.
Figure to explain algorithm.
1. Existing link disrupted by UE movement.
2. Before searching for new link, sort all sector pairs in order such that the nearest sector pairs (w.r.t. the last valid sector pair) are searched first, and then the search distance increases.
3. Search the sector pairs based on sorted set – this ensures UE and AP search is a coordinated manner.
4. Once a new feasible sector pair found, stop search. Search latency proportional to all the searched sector pairs (shaded in green and purple in the ‘Sorted sector pairs‘ set in the figure.
PLEASE READ ATTACHED DOCUMENT BEFORE GOING FURTHER!
Instead of simulation, we do experiment as is provides more accurate performance analysis.
We make the UE node move along walks in indoor rooms (with AP fixed) and as the links are disrupted along the walks, the algorithm determines the new feasible sector pair for the UE at its new position.
To help us evaluate algorithm is a simple and practical manner, we do not actually program the algorithm in the AP and the UE nodes with the algorithm, but actually use the nodes to just measure the UE RSS for 60 GHz transmission from AP at different points along the chosen walks – the reason is explained in the Word document.
Just a brief summary on the process – steps elaborated in further.
Indoor measurement environment. Red points where RSS of UE measured – for every AP and UE sector (i.e. 360° turntable movement for both nodes).
Actual materials of the indoor layout shown in legend.
Technical details of the 60 GHz transceiver. Transceiver chain: GNU Radio <-> USRP <-> SiversIMA.
TX side: GNU Radio generates baseband IQ sample of 500 byte packets -> USRP upconverts baseband signal to IF (1.5 GHz) signal -> SiversIMA upconverts to RF (60 GHz) signal and transmits.
RX side: SiversIMA receives RF (60 GHz) signal and downconverts to IF -> USRP downconverts IF (1.5 GHz) signal to baseband signal -> GNU Radio receives baseband IQ sample of 500 byte packets.
- Feasible sector pairs at all UE points for RSS measurements. Remind audience that this information does not influence the algorithm itself but is used for non-real time processing.
The evaluation methodology in details – as explained in attached Word document.
Notice that the walks chosen are composed of the UE points for which we did our RSS measurements.
- Metrics to verify the effectiveness of our algorithm.
- (Self explainatory).
- (Self explainatory).
- (Self explainatory).
Since the algorithm stops as soon as a new feasible sector pair is found (and doesn‘t bother to look if more feasible sector pairs exist) to reduce latency, the chosen feasible sector pair MAY NOT always be the optimal one – as is always chosen by exhaustive scanning.
So, while we reduce latency, the overall data rate may also drop. But still even in the worst case, the data rate drop is not too bad (remains in Gbps range).