1. IEEE WCNC 2014, Istanbul, Turkey
Effective Transceiver Selection for Mobile Multi-Directional Free-Space-Optical Modules
Abdullah Sevincer and Murat Yuksel
Department of Computer Science and Engineering
University of Nevada, Reno
Project Website
IEEE WCNC 2014, Istanbul, Turkey

2. Collaborators & Sponsors
Faculty
Mona Hella Rensselaer Polytechnic Institute
Nezih Pala Florida International University
Students
Mahmudur Khan (Ph.D.), UNR
Prabath Palathingal (M.S.), UNR
Alumnus
Abdullah Sevincer (Ph.D.), Intel
Behrooz Nakhkoob (Ph.D.), RPI
Mehmet Bilgi (Ph.D.), Microsoft
Michelle Ramirez (B.S.), US Air Force
Acknowledgments
This work was supported by the U.S. National Science Foundation under awards and and DARPA under contract W31P4Q-08-C-0080

3. Motivation
Free-space-optical (FSO) communication has the potential to serve as a complementary technology to RF for the future wireless networking.
Multi-element spherical modules covered with multiple highly directional FSO transceivers has been shown to work well to handle mobility for FSO communication.
Reducing the modules’ energy consumption becomes a crucial issue for the FSO modules with many transceivers.
Need to find an efficient algorithm to reduce power consumption while satisfying throughput performance.

4. Wireless Capacity – NOW!
Scary trends in mobile wireless demand
2+ times increase per year since 2007.
“18-fold by 2016!” Cisco, February 2012.
“More than 80% is landing on WiFi”, excluding iPhone!
Opportunistic networking is well accepted by the users!

5. Free-Space-Optical (FSO): open spectrum
2.4GHz, 5.8GHz, 60GHz, > 300 GHz
FSO usage:
point-to-point links
interconnects
indoor infrared communications
DoD use of FSO:
Satellite communications
AirForce RIKA; DARPA THOR, ORCL, ORCA: air-to-ground, air-to-air, air-to-satellite
802.11a/g,
802.16e,
Cellular (2G/3G)

6. Optical Wireless: Commodity components
Digital Data
ON-OFF Keyed Light Pulses
Transmitter
(Laser/VCSEL/LED)
Receiver
(Photo Diode/ Transistor)
IrDAs…
Lasers…
LEDs…
VCSELs…
Many FSO components are very low cost and available for mass production.

7. Optical Wireless: Why?
More Secure: Highly directional => low probability of interception
Small size and weight: Dense packaging is possible
Very low cost and reliable components, e.g. HBLEDs
35-65 cents a piece, and $2-$5 per transceiver package + upto 10 years lifetime
Very low power consumption (100 microwatts for Mbps!)
Even lower power for 1-10 Mbps
4-5 orders of magnitude improvement over RF
Huge spatial reuse => multiple parallel channels for huge bandwidth increases

8. FSO Issues/Disadvantages
Limited range (no waveguide, unlike fiber optics)
Need line-of-sight (LOS)
Any obstruction or poor weather (fog, heavy rain/snow) can increase BER in a bursty manner
Bigger issue: Need tight LOS alignment:
LOS alignment must be maintained with mobility or sway!
Effects of relative distance and mobility
Received power
Spatial profile:
~ Gaussian drop off
Can we leverage FSO’s benefits while solving the issues?

9. FSO Modules: Spherical Designs
How to handle mobility under LOS alignment requirement?
Mobile FSO =
Directionality + Angular Diversity + Electronic Steering
Multi-transceiver spherical FSO designs.
Need a distributed protocol for this! A B
Bidirectional LOS
ACM/Springer WINET 2009

10. FSO Modules: Spherical Designs
Need to autonomously manage (electronically steer) multiple data transmissions to leverage spatial reuse. A B C

11. FSO Modules: Alignment Protocol
LOS Alignment Process:
Step 1: Search Phase
Step 2: Data Transfer Phase
Goal: Provide an FSO link with “seamless” alignment
Steer the data transmission among the transceivers as the nodes move with respect to each other
Need a 3-way handshake among the transceivers to assure a bidirectional alignment
LOS
Network Layer
Link Layer
PHY
Alignment Protocol

12. FSO Modules: Alignment Protocol
Alignment Lists
to steer transmissions with over multiple alignments
updated at every search interval C 1 21
NextHop.
Interface
Local
B.7 19
A.6 11
NextHop.
Interface
Local
B.5 17
A.6 11 A 1 B 1 21
NextHop.
Interface
Local
A.11 11
C.19 7 21
NextHop.
Interface
Local
B.9 14
C.11 6
NextHop.
Interface
Local
B.11 11
C.11 6
NextHop.
Interface
Local
A.14 9
C.17 5

13. Transceiver Selection Problem
Brute force: send search signals at every transceiver
What happens if there are many transceivers?
Too frequent searches (i.e., too many active transceivers)
Energy consumption
Controller scalability
Too infrequent searches
May not detect the FSO communication opportunities
Brute force alignment protocol cannot scale to 100s of transceivers..

14. Transceiver Selection Problem
Recap goals:
Discover opportunities to establish LOS with neighbors
Maintain the existing links
Insights:
Movements of neighbors project to spatially correlated (or close) transceivers
Significant number of transceivers are not relevant

15. Transceiver Selection
Discover: need to explore
Send search from as many transceivers as possible
Maintain: need to exploit
Send search only from the transceivers close to the ones where an alignment already exists
Single Mode: discover and maintain at the same time-scale
Two Mode: discover at larger time-scales
How to keep the network throughput high (FSO links stay on) via minimal search messaging overhead?

16. Simulation Setup ns-2 simulations of TCP traffic
Parameter Name
Default Value
Number of nodes
 49
Number of flows
 49x48
Visibility
 6 km
Number of interfaces
 8, 16, 24
Mobility
 1 m/s
Simulation time
 3000 s
Transmission range & separation of nodes
 30 m
Area
 210 x 210 m
Node radius
 20 cm
Divergence angle
 0.5 radian
Photo detector diameter
 5 cm
LED diameter
 0.5 cm
ns-2 simulations of TCP traffic
Compare throughput for single vs. two mode operations..
Discovery
0.03, 0.15, 0.3, 0.5, 4, and 8 seconds
Maintenance
0.03, 0.15, and 0.3 seconds
Search from the previous and next (+/- 1 )transceivers

17. Order of magnitude fewer searches in the two-mode alignment
Results
Order of magnitude fewer searches in the two-mode alignment

18. Similar throughput with 10 times fewer search messages
Results
Similar throughput with 10 times fewer search messages

19. Similar throughput with 10 times fewer search messages
Results
Similar throughput with 10 times fewer search messages

20. Summary and Future Work
Messaging overhead of LOS alignment protocol can be controlled with simple tricks
10 fold savings are possible by a two-mode design
Immediate future work:
Discovery could be improved: use randomization
More study of the mobility rate
Longer term future work:
A theoretical framework to formulate the tradeoffs
Applicability to directional RF antennas

21. Questions?

22. Seems Inescapable by the Internet
Wireless nodes will soon dominate the Internet.
Currently ~1B nodes, including wireline.
Urgent response to the exploding wireless demand is a necessity.

23. What is RF and FSO
Cellular is Full: wireless airwaves are full”, CNN, Feb 21, 2012.

24. Dense Deployment: No Help Beyond a Point
As we add more RF nodes, per-node throughput diminishes
Dense deployment of many omni-directional antennas increases interference
sqrt(N) as N increases (Gupta, Kumar, 2000)
Can become linear with hierarchical cooperative MIMO imposing constraints on topology and mobility pattern (Ozgur et al., 2006)
None is able to totally eliminate the scaling problem
The RF spectrum is getting saturated.. We need alternative communication spectrum resources for opportunistic usage.

25. Opportunistic Wireless Channel Characteristics
Complements always-on, lower-rate, higher coverage wireless.
Available Capacity
Optical wireless opportunistic channel
RF opportunistic channel
Basic RF channel
Delay-tolerant applications (e.g. , FTP, Video-on-demand)
Basic RF channel (e.g. 2G, 3G, 4G/WiMAX)
Real-time, interactive applications (e.g. chat, VoIP)
Time
This talk will consider a specific PHY technology, free-space-optics (FSO), that could be useful in this context, and its implications on higher-layer protocols.

26. FSO Modules: Alignment Protocol
Send “search” frames periodically
need an “alignment timer”
Receive data frames only after alignment is established
might still get wrong or erroneous frames – leave them to the higher layers
Discard
Discard
Recv(SYN | SYN_ACK |
DATA)
Recv(ACK, j)
Recv(ACK | DATA)
Not Aligned
Sending SYN
Recv(SYN, i)
Sending
SYN_ACK
Target Node = i
Start
Recv(ACK, i)
Alignment Timer
Timeout
Recv(SYN_ACK, i)
Recv(SYN, i)
Sending
ACK
Target Node = i
Aligned
Target Node = i
Process
Data
Recv(DATA, i)
Recv(SYN_ACK |
ACK)
Recv(DATA, j)
Recv(DATA, i)
Recv(SYN | SYN_ACK | ACK)
Recv(DATA, j)
Discard
Discard
State diagram of LOS alignment protocol

27. FSO Prototype: Transceiver
Picture of transceivers.
3-D optical antenna design.

28. FSO Prototype: Mobility ExperimentA B
Node-A
Node-B
~19Kb/s
~20m
frame size 50B
alignment timer 500ms
ACM Mobicom 2010
ACM Mobisys 2010
IEEE ICC 2010

29. Literature
Multiple elements/transceivers in FSO communication in interconnects which communicate over very short distances. [1]
Focus: Mechanical steering or beam steering [2][3]
Optical flow assignment has not been considered to manage multiple different data flows among transceivers during an on going transmission
The main issues:
interference (or cross-talk) between adjacent transceivers due to finite divergence of the light beam
Misalignment due to vibration.
M. Naruse and S. Yamamoto and M. Ishikawa. Real-time active alignment demonstration for free-space optical interconnections. IEEE Photonics Technology Letters, 13:1257–1259, November [1]
Bisaillon and D. F. Brosseau and T. Yamamoto and M. Mony and E. Bernier and D. Goodwill and D. V. Plant and A. G. Kirk. Free-space optical link with spatial redundancy for misalignment tolerance. IEEE Photonics Technology Letters, 14:242–244, February [2]
Canon, [3]