8/10/2019 110124 Location

1/29

Mobile Location

Technologies

Jeff HightowerIntel Labs

8/10/2019 110124 Location

2/29

A smartphone without location

sensing is like a laptop without WiFi.

8/10/2019 110124 Location

3/29

Why?

8/10/2019 110124 Location

4/29

Maps and Way Finding

Where am I?

Map View, Nearby stuff

How do I get to X?

Directions & Navigation

What is around here?

Nearest Starbucks

Local search

8/10/2019 110124 Location

5/29

Social-Mobile Services

Who is around here?

Where do my friends go?

What is a good exercise route?

How far did I walk today?

8/10/2019 110124 Location

6/29

IT Management & Asset Tracking

Inventory Tracking

Finding lost and stolen devices

Virus breakout tracking Location as tool in computer virus epidemiology

Controlling wireless network access

e.g. Access denied beyond 20m of building

Monitoring device usage

Measure mobility, usage by device class

8/10/2019 110124 Location

7/29

8/10/2019 110124 Location

8/29

Location Sensing Technologies

Technology Application Domain

Satellite Positioning Systems Outdoor navigation

Manual User-entry Location-based web services

Cell-tower Triangulation Web and Fee-based locationtransactions

802.11 Fingerprinting Process management

e.g. HospitalsBeacon-based Location Indoor and fast TTFF mobile

computing

8/10/2019 110124 Location

9/29

Long Range Navigation (LORAN)

Coverage Outdoors, high seas

Accuracy 200-400 meters

Infrastructure cost High

Per-client cost LowPrivacy High

Application Domain Aircraft & Vessels

Simplified LORAN TD lines

8/10/2019 110124 Location

10/29

Global Navigation Satellite Systems (GNSS)

(e.g. GPS, GLONASS, Galileo, Compass)

Coverage Outdoors (line of sight)

Accuracy 10m

Infrastructure cost High

Per-client cost MediumPrivacy High

Application Domain Outdoor navigation

8/10/2019 110124 Location

11/29

GPS Variants

WAAS (LAAS) Improve accuracy to 3 meters (LAAS to 10cm)

Assisted GPS (A-GPS) Uses data network, faster lock times, comparable coverage

Relaxed GPS Loosen the GPS algorithm requirements improve coverage

at the cost of some accuracy

Can work indoors, but with >50m error

Soft-GPS GPS antenna + A/D + CPU

Slight improvement in coverage, time to lock

8/10/2019 110124 Location

12/29

Manual Entry

Coverage Populated Areas

Accuracy 10m-50km

Infrastructure cost Low

Per-client cost LowPrivacy High

Application Domain Location-based web services

8/10/2019 110124 Location

13/29

Cell-Tower Triangulation

Coverage Populated Areas

Accuracy 50-150m

Infrastructure cost Low

Per-client cost MediumPrivacy Low

Application Domain Web services and Fee-basedlocation transactions

8/10/2019 110124 Location

14/29

8/10/2019 110124 Location

15/29

Radio Beacon Location

Coverage Populated Areas + Outdoors

Accuracy 5-150M

Infrastructure cost Low

Per-client cost LowPrivacy Low-High

Application Domain Mobile computing, fast TTFF

8/10/2019 110124 Location

16/29

8/10/2019 110124 Location

17/29

Signal Strength is a Mediocre Indicator of Distance

802.11 signal

strength by

distance

GSM signal

strength by

distance

0%

10%

20%

30%

40%

50%

60%

70%

80%

90%

100%

0 50 100 150 200 250 300 350 400 450 500

Meters from GSM Cell Tower

SignalStrength

Line shows median signal strength

Bars denote 25th to 75th percentile values

-100

-95

-90

-85

-80

-75

0 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 160 170 180

Meters from 802.11 Access Point

SignalStrength(dB)

Line shows median signal strength

Bars denote 25th to 75th percentile values

8/10/2019 110124 Location

18/29

GSM Response Rate By Distance

0%

10%

20%

30%

40%

50%

60%

50 100 150 200 250 300 350 400 450 500

Distance from Cell Tower (meters)

ChanceofAssociatingwithC

ell

802.11 Response Rate By Distance

0%

5%

10%

15%

20%

25%

30%

35%

40%

45%

5 15 25 35 45 55 65 75 85 95 105

115

125

135

145

155

165

175

Distance from 802.11 Access Point (meters)

ChanceofReceivingBeaconFrame

Response Rate Another Indicator of Distance

802.11

response rate

by distance

GSM response

rate by

distance

Response rate = 1 - loss rate of beacon frames

8/10/2019 110124 Location

19/29

Self-Mapping Radio Beacons

Known beacon Inferred beacon

- Grows beacon database using everyday radio traces

- Accuracy and coverage improve over time

8/10/2019 110124 Location

20/29

One Users Experience with Self-Mapping

Self-mapping with sporadic GPS for one volunteer

Day

Accuracy(m)

Cove

rage

0%

20%40%

60%

80%

100%

0 5 10 15 20 25 30

Coverage of self-mapping

0

20

40

60

80

100

0 5 10 15 20 25 30

Accuracy of self-mapping

8/10/2019 110124 Location

21/29

Location in Todays Smartphones

A-GPS

Accurate navigation

and tracking

Cell-ID Lookup

Highly availablefallback

Beacon

LocationFast TTFF

8/10/2019 110124 Location

22/29

Hot Topics in Mobile Location

Computer vision and Indoor Location

Mobile Augmented Reality

Discovering the Places people go Mobile [Push] Advertising

8/10/2019 110124 Location

23/29

23

Computer Vision Location

Accuracy of 30cm and

10o, 80% of the time

4 fps with GPU

Google starting major3D indoor mapping

effort, startup out of

Cambridge

Bundle

adjustment

algorithm

Camera image sequenceMetrically accurate 3D map

3D mapping

3D localization

http://../Documents/work/past/cse590b%20lecture%20Jan%202011/mapping.wmvhttp://../Documents/work/past/cse590b%20lecture%20Jan%202011/mapping.wmv

8/10/2019 110124 Location

24/29

Use your mobile phones camera to

estimate location/orientation and find web

content about where you are right now!

Mobile Augmented Reality

8/10/2019 110124 Location

25/29

25

Knowing types and sequences of

places we go is valuable

Predict likely destinations

Build personal quick-lists

Develop behavior models and

detect changes

Discovering the Places People Go

Manual Check-ins to places is the

commercial state-of-the-art approach

Research underway to augment

check-ins with automatic methods for

place detection and recognition

8/10/2019 110124 Location

26/29

26

Place Learning Two Approaches

Geometry-based Fingerprint-based

Input

Location coordinates

(e.g., GPS, WiFi/Cell tower

triangulation)

Radio environment

(e.g., currently visible cell towers,

WiFi access points)

ProsTightly coupled with the

geographical location of the place

Does not depend on the

underlying positioning systems

accuracy (especially indoors)

Cons

Depends on the underlying

positioning systems accuracy andavailability

Radio environment may changeover time (affecting recognition,

not necessarily detection)

8/10/2019 110124 Location

27/29

27

Sample Trace of WiFi APs encountered

8/10/2019 110124 Location

28/29

28

Sample Trace of WiFi APs encountered

8/10/2019 110124 Location

29/29

29

Results from a 4 week place learning experiment

Fingerprint-based techniques outperform geographic techniques due to the

challenge of accurately clustering coordinates

Many indoor places merged

as a single visit