Sensor Assisted Wireless Communication

1
Sensor Assisted Wireless Communication

• Naveen Santhapuri, Justin Manweiler, Souvik Sen,
• Xuan Bao, Romit Roy Choudhury
• Srihari Nelakuditi

2
Context

• 4.2 billion mobile phones, 50 million iPhones,
• 1 million iPads in 28 days, Androids, Slates, etc
  
• Projection 39x increase in mobile traffic by 2015

3
Different from Laptops

• These devices are always-on, and
• always-with their human owners

4
Wireless
Wired
Mobile Wireless
Wireless
5
Mobile Wireless brings Challenges

• Humans move through various environments
• Devices subject to diverse communication contexts

Office
Home
6
Mobile Wireless brings Challenges

• Humans move through various environments
• Devices subject to diverse communication contexts

Disconnected
3G/EDGE
4G/WiFi
WiFi/Bluetooth
WiFi/3G/4G
Office
Home
High Mobility
Stationary
Low Mobility
Stationary
7
Great Expectations

• Users expect devices to adapt to the context

Disconnected
3G/EDGE
4G/WiFi
WiFi/Bluetooth
WiFi/3G/4G
Office
Home
High Mobility
Stationary
Low Mobility
Stationary
8
Great Expectations

• Users expect devices to adapt to the context

Example1 The phone should turn itself off in
the subway, turn back on at stations or at
destination.
Disconnected
3G/EDGE
4G/WiFi
WiFi/Bluetooth
WiFi/3G/4G
Office
Home
High Mobility
Stationary
Low Mobility
Stationary
9
Great Expectations

• Users expect devices to adapt to the context

Example1 The phone will turn itself off in the
subway, turn back on at stations or at
destination.
Example2 The phone should discern the RF
environment, and jump to the optimal frequency
channel
Disconnected
3G/EDGE
4G/WiFi
WiFi/Bluetooth
WiFi/3G/4G
Office
Home
High Mobility
Stationary
Low Mobility
Stationary
10
In General

• Phones expected to perform
• context-aware communication
• much different from traditional laptop computing

11
Context-Aware Communication

• Innovative research on context-awareness
• Handoffs, adaptive duty cycling, interference
  detection

12
Context-Aware Communication

• Innovative research on context-awareness
• Handoffs, adaptive duty cycling, interference
  detection
• However, most approaches are in-band
• i.e., RF signals used to assess RF context
• In band methods often restrictive
• When will train come to station (for WiFi
  connection)
• Continuous WiFi probing requires high energy
• Difficult to detect primary user in WhiteSpace
  system
• No easy RF signature hard to quickly switch
  channels
• Even difficult to discriminate collision/fading
  in band

13
Our Proposal

• Break away from in-band assessment
• Mobile phones equipped with multiple sensors
• Sensors offer multi-dimensional,
• out of band (OOB) information
• Exploit OOB information to assess context
• Make communication context-aware

14
Examples

• Accelerometer assistance
• Detect user inside subway turn off phone
• Identify nature of movement adapt bitrate
• Detect user driving block a phone call

15
Examples

• Accelerometer assistance
• Detect user inside subway turn off phone
• Identify nature of movement adapt bitrate
• Detect user driving block a phone call
• Acoustic assistance
• Microwave oven hums nearby switch WiFi
  channel
• Hear ambulance siren escape from WhiteSpace
  freq.

16
Examples

• Accelerometer assistance
• Detect user inside subway turn off phone
• Identify nature of movement adapt bitrate
• Detect user driving block a phone call
• Acoustic assistance
• Microwave oven hums nearby switch WiFi
  channel
• Hear ambulance siren escape from WhiteSpace
  freq.
• Multi-dimensional assistance
• Sense which users will leave WiFi hotspot sooner
  priotitize WiFi traffic to save 3G

17
Observe that

• Sensor assisted apps
• Already in use
• E.g., Display off when talking
• on phone (proximity sensor)
• E.g., Ambience-aware ringtones

17
18
Observe that

• Sensor assisted apps
• Already in use
• E.g., Display off when talking
• on phone (proximity sensor)
• E.g., Ambience-aware ringtones
• Sensor-assisted communications
• Relatively unexplored

18
19
Sensor Assisted Wireless Communication
19
20
Why Out-of-Band?
Contexts have diverse fingerprints across
multiple sensing dimensions
Sound
Motion
Light
Wireless
Diversity improves context identification (at
least one fingerprint easy to detect)
In-band sensing unable to leverage this diversity
20
21

• Case Study 1
• Microwave Oven Aware Channel Switching

22
Problem

• Microwave ovens operate at 2.4GHz
• Interferes with WiFi receivers
• WiFi transmitters carrier sense and dont
  transmit
• Throughput degrades
• In-band detection difficult
• Microwave interference similar to WiFi

Channel 6
Channel 6
22
23
Acoustic Fingerprint Hum

• Microwave hum is out of band signal
• Detect this acoustic signature
• Switch WiFi to different channel
• When hum stops
• Switch back to original channel

Channel 6
Channel 11
Sound
23
24
Signature Detection
Microwaves distinct acoustic signature in
frequency domain
24
25
Throughput
Throughput comparison across 802.11b/g channels
with and without Microwave
25
26

• Case Study 2
• Activity Aware Call Admission

27
Opportunity

• Phone accelerometer detects user is driving
• Discriminate between driver and passenger

Initiate call
27
28
Opportunity

• Phone accelerometer detects user is driving
• Discriminate between driver and passenger
• Phone blocks call
• Checks if call can be postponed for later
• Can be generalized to other activities

User Driving Continue?
Initiate call
28
29
Accelerometer Signatures
Accelerometer signatures different for driver and
passenger
29
30

• Case Study 3
• Behavior Aware 3G Offloading

31
Problem and Opportunity

• 3G networks overloaded
• Exploit WiFi hotspots to offload 3G load
• Sense user behavior via multiple sensors
• Predict which users likely to exit the hotspot
  soon
• Prioritize WiFi for soon to leave users
• More WiFi traffic less carry-over to 3G

31
32
Dwell Time Prediction

• Phones sense user behavior
• Summarizes sensor readings to AP
• AP runs machine learning algorithm
• Classifies behavior into dwell time buckets
• AP shapes traffic
• Shorter dwell time higher priority

32
33
Studying (60 minutes)
Drive Through (3 minutes)
Grocery Shop (15 minutes)
34
3G Offload
112 MB 3G data saved per hour 2 Behavior Aware AP
1 new 3G user
34
35
Exercise Caution

• Count sensing overheads
• Sensing is not free
• However, sensors may be on cost may amortize
• Out-of-band should provide timely context
• Suitable in our case studies
• Inadequate for some applications
• Treat SAWC as hint rather than solution
• Complementary to in-band sensing

35
36
Summary

• Pervasive communication systems
• Need to be agile to changing contexts
• In band context-awareness may be feasible
• But often expensive, inefficient
• Mobile devices equipped with many sensors
• Together enable a broader view
• We propose to leverage this opportunity via
• Sensor Assisted Wireless Communications (SAWC)

36
37
Out-of-Band in Real Life
Out-of-band information provides useful hints
37
38

• Please stay tuned for more at
• http//synrg.ee.duke.edu
• Thank You

39
Thank You!Questions?
39
40
Continuous in-band context assessment incur
overheads
Todays systems optimize for the common case
Sacrifices performance under atypical contexts
40
41

• In the perspective of
• related work

42
SAWC Classification
Source
Implicit
Explicit
Data
In-band Wireless
Out-of- band
42
43
Context-Awareness

• RF context assessment
• Remains an elusive research problem
• Several approaches use in-band analysis
• i.e., RF signals used to assess RF context
• For example
• Difficult to discriminate between
  collision/fading
• No easy RF signature
• When will train come to station (for WiFi
  connection)
• Continuous RF scanning requires high evergy
• Download more from WiFi before moving out of
  range
• Hard to tell (using RF) how soon user will
  disconnect

44
Mobility Demands Agility

• For example, from home to office
• A user transitions through numerous environments

Office
Home
High Mobility
Stationary
Low Mobility
Stationary
45
Mobility Demands Agility

• For example, from home to office
• A user transitions through numerous environments
• Devices subject to various communication contexts

Disconnected
3G/EDGE
4G/WiFi
WiFi/Bluetooth
WiFi/3G/4G
Office
Home
High Mobility
Stationary
Low Mobility
Stationary