Tracking Prasun Dewan Department of Computer Science University of North Carolina [email protected].

59
Tracking Prasun Dewan Department of Computer Science University of North Carolina [email protected]

Transcript of Tracking Prasun Dewan Department of Computer Science University of North Carolina [email protected].

Page 1: Tracking Prasun Dewan Department of Computer Science University of North Carolina dewan@unc.edu.

Tracking

Prasun Dewan

Department of Computer Science University of North Carolina

[email protected]

Page 2: Tracking Prasun Dewan Department of Computer Science University of North Carolina dewan@unc.edu.

2

Triangulation

Need to solve for x, y, z Assume orientation not an issue

Need distance to three points with known coordinates Can solve for x, y, z

Page 3: Tracking Prasun Dewan Department of Computer Science University of North Carolina dewan@unc.edu.

3

Issues

What are the three known points? How to determine distances? Expense Privacy

Page 4: Tracking Prasun Dewan Department of Computer Science University of North Carolina dewan@unc.edu.

4

GPS

Satellites are known points Their current location known 24 hrs in advance

upto accuracy of a few meters Used for tuning?

They also broadcast their position Measure time takes for signal to each receiver

Signal frequency 1575.42 MHz and 1227.6 MHz Code division multiple access to eliminate

interference Time of flight of signal gives distance

Page 5: Tracking Prasun Dewan Department of Computer Science University of North Carolina dewan@unc.edu.

5

Clock Synchronization

Clocks of satellites synchronized Clock of receiver not synchronized Offset same for all satellites One more variable Need four satellites

Page 6: Tracking Prasun Dewan Department of Computer Science University of North Carolina dewan@unc.edu.

6

Excerpt from Hopper’s Slides

Start of excerpt

Page 7: Tracking Prasun Dewan Department of Computer Science University of North Carolina dewan@unc.edu.

7

Sentient ComputingUbiquitous Computing vision

Computing devices everywhere

Access to applications anywhere

Whatever is on hand is available

Sentient Computing visionUbiquitous Computing made context-aware

Physical context used for automatic control

Sensors and space are part of computing systems

Page 8: Tracking Prasun Dewan Department of Computer Science University of North Carolina dewan@unc.edu.

8

Programming With Space

The componentsNotions and representations of

physical spaceData and computational modelsSensor information

User interface the real world

Page 9: Tracking Prasun Dewan Department of Computer Science University of North Carolina dewan@unc.edu.

9

Components for Programming With Space

Devices

Platforms

Sensors

Networks

+Architecture

Conduits

Page 10: Tracking Prasun Dewan Department of Computer Science University of North Carolina dewan@unc.edu.

10

Components for Programming With Space

Devices

Platforms

Sensors

+Architecture

Page 11: Tracking Prasun Dewan Department of Computer Science University of North Carolina dewan@unc.edu.

11

Sensors: Location Information

Containment GSM, UMTS, broadband radio Active badge

Proximity Bluetooth, IrDA PICOnet

Co-ordinate GPS Active bat

Page 12: Tracking Prasun Dewan Department of Computer Science University of North Carolina dewan@unc.edu.

12

Sensors: Location Information

Containment GSM, UMTS, Broadband Radio Active Badge

Proximity Bluetooth, IrDA PICOnet

Co-ordinate GPS Active bat

Page 13: Tracking Prasun Dewan Department of Computer Science University of North Carolina dewan@unc.edu.

13

Containment: Active Badge

Infra-Red Network 10 meter range

diffuse

room-scale location

Page 14: Tracking Prasun Dewan Department of Computer Science University of North Carolina dewan@unc.edu.

14

Sensors: Location Information

Containment GSM, UMTS, Broadband Radio Active Badge

Proximity Bluetooth, IrDA PICOnet

Co-ordinate GPS Active Bat

Page 15: Tracking Prasun Dewan Department of Computer Science University of North Carolina dewan@unc.edu.

15

Sensors: Location Information

Containment GSM, UMTS, broadband radio Active badge

Proximity Bluetooth, IrDA PICOnet

Co-ordinate GPS Active bat

Page 16: Tracking Prasun Dewan Department of Computer Science University of North Carolina dewan@unc.edu.

16

Ultrasonic Location System

Mobile transmitter (Bat)

Fixed receiversCeiling

Active BatsUltrasonic transponder

Measure pulse time-of-flight

Radio synchronised

Page 17: Tracking Prasun Dewan Department of Computer Science University of North Carolina dewan@unc.edu.

17

DSP Ceiling Array

25,000 MIPS to cover AT&T Laboratories Cambridge!

Page 18: Tracking Prasun Dewan Department of Computer Science University of North Carolina dewan@unc.edu.

18

Components for Programming With Space

Devices

Platforms

Sensors

Networks

+Architecture

Conduits

Page 19: Tracking Prasun Dewan Department of Computer Science University of North Carolina dewan@unc.edu.

19

Telephone318

Computer“Pumpkin”

Computer“Papaya” Person

“Mike”

Person“Pete”

Representing the Real World

Model real world as collection of objects

Computer“Plantain”

Person“Andy”

Follow-mePhonebook

MobileDesktop

Telephone241

Telephone217

CTIswitch

Resourcemonitor

Keyboardmonitor

Locationservice

Applications

Software objects

Sensors

Objects maintain state using sensor data Applications query relevant sets of objects

Page 20: Tracking Prasun Dewan Department of Computer Science University of North Carolina dewan@unc.edu.

20

Data Model Visualisation

Page 21: Tracking Prasun Dewan Department of Computer Science University of North Carolina dewan@unc.edu.

21

Spatial Monitoring

Vague spatial facts formalised as geometric containment and overlapping relationships between spaces

X

M

‘X is holding the microphone M’‘X can be seen by

camera B but not by camera A’

A

B

X

Page 22: Tracking Prasun Dewan Department of Computer Science University of North Carolina dewan@unc.edu.

22

Spatial Indexing

Generates all positive/negative overlapping or containment events

thro

ug

hp

ut

(‘00

0 u

pd

ates

s-1)

1

3

2

4

population (‘000)10 20 30

non-overlapping spaces

overlapping spaces

Page 23: Tracking Prasun Dewan Department of Computer Science University of North Carolina dewan@unc.edu.

23

Putting It All TogetherMove user’s desktop to screen in front of them

Visible

A

Visib

le

B

Visible

C

Callbacks

Registration+ve Containment (Andy)-ve Overlapping (Andy)

-ve Overlapping(Andy,”Visible B”)

CLEAR DESKTOP FROM B

-ve Overlapping(Andy,”Visible A”)

CLEAR DESKTOP FROM A

+ve Containment(Andy,”Visible B”)

MACHINE B: NOT IN USEMOVE DESKTOP TO B

+ve Containment(Andy,”Visible C”)MACHINE C: IN USE NO ACTION

Page 24: Tracking Prasun Dewan Department of Computer Science University of North Carolina dewan@unc.edu.

24

Example Applications

Corporate memoryRecord me / what’s around meAnnotate multimedia stream

Camera field-of-view

Flat display

Compositedisplay

“Plonk-and-play” systemsSpatial configuration determines

logical configuration

No need to know device IDs

Automatic personalisation

Page 25: Tracking Prasun Dewan Department of Computer Science University of North Carolina dewan@unc.edu.

25

Sentient Computing: New User Interfaces

Non-user interfaces!

Objects and people are cursors in the real-world of icons

Aural and visual feedback

Page 26: Tracking Prasun Dewan Department of Computer Science University of North Carolina dewan@unc.edu.

Nissanka B. Priyantha Anit Chakraborty

Hari Balakrishnan

MIT Lab for Computer Science

http://nms.lcs.mit.edu/

The Cricket Location-Support System

Page 27: Tracking Prasun Dewan Department of Computer Science University of North Carolina dewan@unc.edu.

27

Motivation

Emergence of pervasive computing environments

Context-aware applications Location-dependent behavior

User and service mobility Navigation via active maps Resource discovery

Cricket provides applications information about geographic spaces they are in

Page 28: Tracking Prasun Dewan Department of Computer Science University of North Carolina dewan@unc.edu.

28

Design Goals

Preserve user privacy Operate inside buildings Recognize spaces, not just physical

position Good boundary detection is important

Easy to administer and deploy Decentralized architecture and control

Low cost and power consumption

Page 29: Tracking Prasun Dewan Department of Computer Science University of North Carolina dewan@unc.edu.

29

Traditional Approach

Controller/Location database

Base stations

ID = u

Transceivers

• Centralized architecture• User-privacy issues• High deployment cost

ID = u ? ID = u ? ID = u ?

ID = u ?

Page 30: Tracking Prasun Dewan Department of Computer Science University of North Carolina dewan@unc.edu.

30

Cricket Architecture

Beacon

Listener

SpaceA

SpaceB

SpaceC

I am atC

• Decentralized, no tracking, low cost• Think of it as an “inverted BAT”!

Page 31: Tracking Prasun Dewan Department of Computer Science University of North Carolina dewan@unc.edu.

31

Determining Distance

A beacon transmits an RF and an ultrasonic signal simultaneously RF carries location data, ultrasound is a

narrow pulse Velocity of ultra sound << velocity of RF

RF data(location name)

Beacon

Listener

Ultrasound(pulse)

• The listener measures the time gap between the receipt of RF and ultrasonic signals– A time gap of x ms roughly corresponds to a

distance of x feet from beacon

Page 32: Tracking Prasun Dewan Department of Computer Science University of North Carolina dewan@unc.edu.

32

Uncoordinated Beacons

Multiple beacon transmissions are uncoordinated

Different beacon transmissions can interfere Causing inaccurate distance measurements

at the listener

Beacon A Beacon B

timeRF B RF A US B US A

Incorrect distance

Page 33: Tracking Prasun Dewan Department of Computer Science University of North Carolina dewan@unc.edu.

Handling Spurious Interactions

Combination of three different techniques: Bounding stray signal interference Preventing repeated interactions

via randomization Listener inference algorithms

Page 34: Tracking Prasun Dewan Department of Computer Science University of North Carolina dewan@unc.edu.

34

Bounding Stray Signal Interference

RF range > ultrasonic range Ensures an accompanied RF signal with

ultrasound

tRF A US A

Page 35: Tracking Prasun Dewan Department of Computer Science University of North Carolina dewan@unc.edu.

35

t

S/b

r/v (max)

S - size of space stringb - RF bit rater - ultrasound rangev - velocity of ultrasound

Bounding Stray Signal Interference

(RF transmission time) (Max. RF US separation at the listener)

S r

b v

Page 36: Tracking Prasun Dewan Department of Computer Science University of North Carolina dewan@unc.edu.

36

Bounding Stray Signal Interference

• Envelop ultrasound by RF• Interfering ultrasound causes RF signals to

collide• Listener does a block parity error check

– The reading is discarded

tRF A US A

RF B US B

Page 37: Tracking Prasun Dewan Department of Computer Science University of North Carolina dewan@unc.edu.

37

Preventing Repeated Interactions

Randomize beacon transmissions:

loop:pick r ~ Uniform[T1, T2];delay(r);xmit_beacon(RF,US);

Erroneous estimates do not repeat Optimal choice of T1 and T2 can be calculated

analytically Trade-off between latency and collision

probability

Page 38: Tracking Prasun Dewan Department of Computer Science University of North Carolina dewan@unc.edu.

Inference Algorithms

MinMode Determine mode for each beacon Select the one with the minimum mode

MinMean Calculate the mean distance for each beacon Select the one with the minimum value

Majority (actually, “plurality”) Select the beacon with most number of readings Roughly corresponds to strongest radio signal

Page 39: Tracking Prasun Dewan Department of Computer Science University of North Carolina dewan@unc.edu.

Inference Algorithms

Distance(feet)

Frequency A B

5 10

5

A B

Actual distance (feet) 6 8

Mode (feet) 6 8

Mean (feet) 6.14 6.4

Number of samples 7 10

Page 40: Tracking Prasun Dewan Department of Computer Science University of North Carolina dewan@unc.edu.

40

Closest Beacon May Not Reflect Correct Space

I am atB

Room A Room B

Page 41: Tracking Prasun Dewan Department of Computer Science University of North Carolina dewan@unc.edu.

41

Correct Beacon Positioning

Room A Room B

x x

I am atA

• Position beacons to detect the boundary

• Multiple beacons per space are possible

Page 42: Tracking Prasun Dewan Department of Computer Science University of North Carolina dewan@unc.edu.

42

Implementation Cricket beacon and listener

• LocationManager provides an API to applications

• Integrated with intentional naming system for resource discovery

Page 43: Tracking Prasun Dewan Department of Computer Science University of North Carolina dewan@unc.edu.

43

Implementation Cricket beacon and listener

• LocationManager provides an API to applications

• Integrated with intentional naming system for resource discovery

Micro-controller

RF

US

Micro-controller

RF

USRS232

Page 44: Tracking Prasun Dewan Department of Computer Science University of North Carolina dewan@unc.edu.

44

Static listener performance

Interference

L2

L1

• Immunity to interference– Four beacons within

each others range– Two RF interference

sources

• Boundary detection ability– L1 only two feet

away from boundary

I1 I2

L1 0.0% 0.0%

L2 0.3% 0.4%

I1

I2

% readings due to interference of RF from I1

and I2 with ultrasound from beacons

Room B

Room C

Room A

Page 45: Tracking Prasun Dewan Department of Computer Science University of North Carolina dewan@unc.edu.

45

Inference Algorithm Error Rates

Error Rates Measured With Listener At L1

0

5

10

15

20

25

30

35

40

45

10 20 30 40 50 60 70 80 90 100

Number of readings

Err

or R

ate

(%)

MinMean

MinMode

Majority

Page 46: Tracking Prasun Dewan Department of Computer Science University of North Carolina dewan@unc.edu.

46

Mobile listener performance

Location Algorithm Error Rates

0

2

4

6

8

10

12

14

16

18

20

2 3 4 5 6

Sampling Interval

Erro

r Rat

e (%

)

MinMean

MinMode

Majority

Room A Room B

Room C

Page 47: Tracking Prasun Dewan Department of Computer Science University of North Carolina dewan@unc.edu.

47

ComparisonsBat Active

badgeRADAR Cricket

Track user location?

Yes Yes No, if client has signal map

No

Deploymentconsiderations

Centralized controller +matrix ofsensors

Centralized database + wired IR sensors

RF signal mapping and good radios

Spacenamingconvention

Position accuracy

Few cm Room-wide Room-wide ~2 feet forspatialresolution

Attribute

System

Page 48: Tracking Prasun Dewan Department of Computer Science University of North Carolina dewan@unc.edu.

48

Summary

Cricket provides information about geographic spaces to applications Location-support, not tracking Decentralized operation and administration

Passive listeners and no explicit beacon coordination Requires distributed algorithms for beacon

transmission and listener inference Implemented and works!

Page 49: Tracking Prasun Dewan Department of Computer Science University of North Carolina dewan@unc.edu.

49

Page 50: Tracking Prasun Dewan Department of Computer Science University of North Carolina dewan@unc.edu.

50

Page 51: Tracking Prasun Dewan Department of Computer Science University of North Carolina dewan@unc.edu.

51

Decentralized

Page 52: Tracking Prasun Dewan Department of Computer Science University of North Carolina dewan@unc.edu.

52

Page 53: Tracking Prasun Dewan Department of Computer Science University of North Carolina dewan@unc.edu.

53

Preserves user privacy Good granularity Component cost U.S. $10

Page 54: Tracking Prasun Dewan Department of Computer Science University of North Carolina dewan@unc.edu.

54

Beacon positioning

• Imaginary boundaries

• Multiple beacons per location

Location X

X1 X2

X3

ImaginaryBoundary

Page 55: Tracking Prasun Dewan Department of Computer Science University of North Carolina dewan@unc.edu.

55

Future work

Dynamic transmission rate with carrier-sense for collision avoidance.

Dynamic ultrasonic sensitivity. Improved location accuracy. Integration with other technologies such

as Blue Tooth.

Page 56: Tracking Prasun Dewan Department of Computer Science University of North Carolina dewan@unc.edu.

57

Inference algorithms

Compared three algorithms Minimum mode Minimum arithmetic mean Majority

Page 57: Tracking Prasun Dewan Department of Computer Science University of North Carolina dewan@unc.edu.

58

Minimizing errors.

Proper ultrasonic range ensures overlapping RF and ultrasonic signals RF data 7 bytes at 1 kb/s bit rate RF signal duration 49 ms Selected ultrasonic range = 30ft < 49 ft Signal separation < 49 ms

Page 58: Tracking Prasun Dewan Department of Computer Science University of North Carolina dewan@unc.edu.

59

Minimizing errors.

Interfering ultrasound causes RF signals to collide

Listener does a block parity error check The reading is discarded

Page 59: Tracking Prasun Dewan Department of Computer Science University of North Carolina dewan@unc.edu.

60