Secondary Spectrum Usage for Mobile Devices

Better Mobility Support
for White Space Devices
Aug 19, 2015
Amjad Yousef Majid
4178890
Supervisor: dr. P. Pawełczak
1

What is White Space Spectrum?
2
US TV Available Spectrum (Example)
● WS spectrum is an
unused spectrum

WS Spectrum to the Rescue
Cisco forecasts 24.3 exabytes per month of mobile data traffic by 2019
http://goo.gl/zHqYg1
3

Obtaining WS Spectrum Information
● Spectrum Sensing
○ Lack of guarantees
○ Technical challenges
● WS Databases
○ Knows exactly about the incumbent services and
their protection requirements
4
Device, WSDB communication

Problem Statement (1)
How to provide WS spectrum to Mobile devices
in an efficient way compared to the current
WSDB query technique?
5

The Research Outline
6
WSDB
Comparative Study Optimization Process
Mobile access
optimization
Information
optimization
1
23

Experimental Setup
● WSDBs profiling (response time, message size)
○ Ubuntu 14.04, CURL+Matlab, Eduroam, MacBook Air (stationary)
● WSDB on smartphone (energy consumption, delay)
○ NEAT, Samsung Galaxy S3, KPN
● > 6k lines of code
● >1M queries
7

WSDBs Used in Experiments
8
● Seven White Space Databases (WSDB)
● Distributed in three countries
● WSDBs use three message formatting
● Two WSDBs support Multi-Location query

Results: Current Query Technique 1
9
● Variant and fix sizes
● Large differences
● SBO very slow
● GGL most efficient
● NOM most consistent
with 0.1 sec variance

Optimizing Mobile Devices
Access to WSDBs
10
2

Multi-Location WSDB Query
11
Multi-Location Query: querying multiple locations in
the same (JSON) request
{location} → [{location1},{location2},...]
● 100m vs 300km
Google: “This isn’t a supported mode of operation,
though it’s nice to find out that it does appear to work.”

Results: ML Query Response Time
12
● Jump in delay between
1 and 2 locations
● After 2 locations delay
increases marginally
● ML Query increases
message size, but
compression is possible

Results: Time-Energy Models
e(t) = a t + b
a = 0.94
b = -0.58
t(n) = a n + b
a = 0.087
b = 3
13
e(n) = c n + d
c = 0.082
b = 2.55

WSDB ML Query scenarios
● AML: Area ML query
○ Specify area for which you will move (PAWS)
● OML: “Oracle” ML query
○ Know exactly where you plan to move
● ODML: On-demand ML query
○ Query as you go (status quo) 14

ML Query Algorithm: Nuna
15
● Idea:
○ Converting Query time to Query distance
D = S T
○ Find the size of the ML Query
E = (A N - D) / D [e.g. A: 100m (US)]
○ Find direction to distribute locations
Longitude/Latitude changing factors
latChFac = a / (a + o)
A N
D (AN - D)
o
a
Movement
lat
long

16
Implementation
● First stage: Current location query
● Second stage: Specify direction of movement,
first multi-location query
● Next stage: Check direction, multi-location query
or back to first stage
Algorithm: ~400 lines of code; the App ~1500 (Java)
ML Query Algorithm: Nuna

Nuna: Example Run
17
Setup:
● Outside of a car
● Path: 12 km
● Speed: < 80 km/h
We drove over 300 km to test and improve Nuna

Results: Nuna Testing
Querying
Querying+GPS
18
● ODML reduces energy
consumption to half
● Number of queried locations
increases quadratically with
distance

Optimizing WSDB Information
19
By providing local readings
3

Portable Spectrum Sensing Platform
20
OTG Switch
RTL-SDR
Custom app
of 1100
lines of code
rtl power

Problem Statement (2)
What is the energy cost of sensing the
spectrum using a RTL-SDR dongle connected
to a smartphone and sending the data to a
WSDB?
21

Energy consumption of PoSSP
22
● rtl power, FFT size, BW
● PoSSP energy consumption is
independent when the FFT size is
below 1MHz
● PoSSP consumes energy the
most when the FFT is 1MHz
● Above 50 hops the algorithm
stops sensing and returns -Inf

● Two papers are to be submitted for possible publication
● We are almost done with final version of describing Nuna to be patented
● Follow up research tasks are suggested from two WSDBs operators
○ a
○ b
24

Results: Nuna on Various Paths
26
CP: Circular path, 3km
LP: Long-lines path, 3km
Five iterations around
each path
CP LP

Results: Google WSDB Delay
AK: Alaska
US: Mainland US
rpc: Remote Procedure Call
28
● Smaller database,
faster response

Response: Location Dependency
29

Nominet Conclusion
In conclusion, although the regulatory framework currently
allows transmissions in mobility, practical considerations
suggest that more technical work needs to be done to
support it.
● Batch query will take time.
● When a device change locations it needs to query
again.
Mobility Transmission in TVWS.
31

GGL query message
curl -XPOST https://www.googleapis.com/rpc -H "Content-Type: application/json" --data '{
"jsonrpc": "2.0",
"method": "spectrum.paws.getSpectrum",
"apiVersion": "v1explorer",
"params": {
"type": "AVAIL_SPECTRUM_REQ",
"version": "1.0",
"deviceDesc": { "serialNumber": "your_serial_number", "fccId": "TEST", "fccTvbdDeviceType": "MODE_1" },
"location": { "point": { "center": {"latitude": 42.0986, "longitude": -75.9183} } },
"antenna": { "height": 30.0, "heightType": "AGL" },
"owner": { "owner": { } },
"capabilities": { "frequencyRanges": [{ "startHz": 800000000, "stopHz": 850000000 }, { "startHz": 900000000,
"stopHz": 950000000 }] },
"key": "your_API_key"
}, "id": "any_string" }
32

Future work
● Finding the most common frequency band
within Area ML query
● Spectrum management for mobile SU’s
through local sensing
● Reducing the dependency of Nuna on GPS
33

Problem Statement
● WSDBs implemented using IETF PAWS
● Implementation of IETF PAWS has a poor
support for mobility
34

Secondary Spectrum Usage for Mobile Devices

More Related Content

What's hot (20)

Similar to Secondary Spectrum Usage for Mobile Devices (8)

Recently uploaded (20)

Secondary Spectrum Usage for Mobile Devices