1. Indoor Location Sensing Techniques
Professor Lionel M. Ni
Dept. of Computer Science
Hong Kong Univ. of Science & Technology

2. Outline Location representation Location determination
Sensing technologies
Some prototype systems
RFID technology
Bluetooth technology
Conclusions

3. Location Representation
Affect data structure
Able to handle object mobility
Facilitate efficient query processing
Geometric model
Symbolic model
Location graph

4. Geometric Model: Coordinate
Coordinates: fine-grained representation
distance, angle, time, etc.
2D or 3D
Absolute coordinate
Based on a shared reference grid
GPS: latitude, longitude, elevation
Relative coordinate
Based on its own frame of reference
Readily available mathematical models
May be difficult to implement

5. Symbolic Model: Symbols
An abstract view of an object
Use a symbol or a name (e.g., he is in office 202)
Coarse-grained position
Applications may not need a low resolution
Fine-grained position may be difficult to obtain
Need more readers/sensors to increase accuracy

6. Which one to use? Application dependent
Output format of the underlying sensors
Cost and available sensing technologies
A combination of both
How to convert from one to another
How to bind one to another
Question: what’s the efficient path from location A to location B
Question: where this object is likely to move?

7. Location Graph Generalization of symbolic model Three important tasks
Selection of efficient navigation path
Not necessary shortest path due to security and domain regulation
Motion and path prediction
Weighted graph: grow and shrink with time
Nodes and weights of edges are dynamically updated
Hierarchical structure of abstraction
Facilitate scalability and data abstraction

8. Location Determination
Triangulation: use geometric properties of triangle to compute object locations
Scene analysis: use features of a scene observed from a certain reference point
Proximity: determine if an object is near a known location

9. Triangulation
Compute locations based on geometric properties of triangles
Lateration: distance measures (popular)
2D: 3 non-collinear reference points
3D: 4 non-coplanar reference points
Less reference points with domain specific knowledge
Angulation: angle measures
2D: two angles and one length
3D: two angles, one length, one azimuth

10. Distance Measurements
Time-of-flight
Distance = time x velocity
Sound wave: 344 m/s at 21oC
Light: higher lock resolution (six orders of mannitude)
Stationary or moving objects
Round-trip vs single-trip time: synchronization problem
Multipath problem: reflections

11. Distance Measurements (cont)
Signal attenuation
Signal strength is a function of distance from the emission source
E.g., free space radio signal attenuation is inverse proportional to distance square
Need a good signal propagation model
Less accurate due to signal reflection, refraction, and other interferences

12. Scene Analysis
Use features of a scene observed from a particular point to estimate the location of objects in the scene
Scenes: images, measurable physical phenomena (e.g., electromagnetic characteristics)
Simplify scenes to easily represent and compare features
Static scene analysis: lookup dataset
Differential scene analysis: track difference between successive scenes

13. Scene Analysis (cont’d)
Advantage: the location of objects can be inferred using passive observation and features that do not correspond to geometric angles or distances
Disadvantages
Need to access to the features of the environment to compare with
Change of environment has to create a new dataset
Usually require more processing time

14. Proximity Detect if an object is near a known reference location
Use the known reference location to proximate the object location
Methods
Objects discovery by access points
Registration through automatic identification systems

15. Others May use a combination of different techniques
Who performs location determination?
Mobile device
Centralized location service

16. Sensing Technologies Infrared Radio Frequency Ultrasound 802.11b
Bluetooth
Others

17. Some Research Prototypes
Microsoft RADAR
2D, IEEE b
HP Websigns
GPS (outdoor), b (indoor)
AT&T Cambridge Sentient
3D, b, ultrasound
Microsoft EasyLiving
3D, scene analysis (vision)
MIT Cricket

18. Electromagnetic Spectrum

19. Infrared (IR) Technology
Detection of optical signal power (not frequency and phase) leads to
Simpler system design
Economical, if sharing of components possible
No regulatory constraints on design
Technology perceived as safe

20. Problems with IR Signals
The Sun, fluorescent and incandescent lights share same electromagnetic spectrum with IR
High levels of IR radiation affects receiver sensitivity
Almost impossible to use outdoors during daytime
Detector Area Trade-offs
Material Opaqueness
Little Refraction, more reflection
Multipath propagation, intersymbol interference

21. Infrared in Everyday Life
Most obvious ones include
Remote controls of most electronic appliances (TV, audio systems, etc)
Most laptops nowadays are equipped with infrared port as another alternative of comm.
PDA (Personal Digital Assistant) like Palm and Pocket PC all have infrared as an easy form comm. between devices

22. Infrared in Everyday Life
Not so obvious as those devices mentioned before, infrared in fact exists even without our notice
Heat or thermal radiation
Sunlight, fire, radiator
Any object with a temperature above absolute zero radiates infrared

23. Infrared Comm. Protocol
Most popular ones include:
CIR, consumer IR, which consists of the different protocol schemes used by consumer electronics, like remote control
IrDA, infrared data association, which is the non-profit org. Responsible for maintaining the standardized IR data comm. Protocols and promoting interoperable, low cost infrared data interconnection standards that support a walk-up, point-to-point user model.

24. CIR, Consumer IR Frequency range: 30 – 40 kHz
Chosen such that these higher frequencies are less likely to be interfered by other light sources
Most are using some sorts of binary encoding, wither variances in length for both time and bit length

25. IrDA Benefit Cons “Wireless” Easy to implement and simple to us
Safe in any environment
No government regulatory issues
Open standardized protocols
Cons
Line of sight problem
Shorter range (up to 1m)

26. Infrared: Active Badge
Low power requirements
Low circuitry costs: $2-$5 for the entire coding/decoding circuitry
Simple circuitry
Higher security
Portable
High noise immunity
Line-of-sight
Coarse resolution
Short range
Blocked by common materials
Light, weather sensitive
Pollution can affect transmission

27. Radio Basics
Radio power fades as 1/distance^2 from where it’s transmitted.
Power needed is proportional to bandwidth.
Range and bandwidth limited by noise.
Signals can bounce causing multipath.
Low frequency signals can go around things.

28. RF Advantage Not line-of-sight
Not blocked by common materials: can penetrate most solids and pass through walls
Longer range
Not light sensitive
Not as sensitive to weather/environmental conditions

29. Issues Related to RF Regulation of use of allocated bands
Receivers need higher sensitivity: costly
Selectivity and high gain lead to complex design
Coexistence in crowded spectrum
Additional US cost burden to implement spectrum spreading for license-free operation in ISM bands

30. Issues Related to RF (cont)
Low cost systems achieved by high volume production
High volume wireless components benefits wireless LANs
Advances in fabrication technology leads to smaller, cheaper systems
Health concerns in usage of high-powered RF systems
Contend with atmospheric, galactic, human-made, intentional and unintentional noise
Intersymbol interference
Mobility of devices affects reception/transmission of information

31. Ultrasound Similar to IR
High precision for location aware computing (sub centimeter)
Low speed, but good enough for sub centimeter precision
Short range (up to 10m)

32. Ultrasonic Active Bat (AT&T) ultrasound time-of-flight measurement
can locate Bats to within 9cm of their true position for 95 percent of the measurements

33. Ultrasonic and RF: MIT Cricket
Ultrasonic time-of-flight and a radio frequency control signal
Lateration and proximity techniques
Decentralized scalability
4x4 square-foot regions
Beacon
Listener

34. High accuracy
High precision
Hard to deploy
High cost

35. Commercial Wireless Data Communications
Technology
Near Future
Digital Cellular
Bluetooth
Technology
Home RF
Proprietary
Systems
802.11b
Wireless LANs
Present
Cellular
19K
128K
384K
721K
11M
54M
Data Rate Bits per second

36. Sources of Wireless Errors
Attenuation
loss of electromagnetic energy
Front end overload
transmitter’s overwhelming filters in the receiver
Narrowband interference
overlapping of a small frequency band
Spread spectrum interference
frequency hopping (FHSS) or Direct Sequence Spread Spectrum (DSSS)
Natural background noise
Multipath interference
interference due to multiple paths between the transmitter and the receiver
CIS 640

37. Spread Spectrum Techniques
FREQUENCY HOPPING (FHSS)
Strong signal sent over narrow bandwidth
Signal hops several times a second
Jump occasionally. 75 or more frequencies used. Up to 2 Mbps.
DIRECT SEQUENCE (DSSS)
Lower transmit output, signal is spread over wider bandwidth
Use of same frequencies consistently

38. IEEE b Technology
The standard is believed to be very good, will be widely accepted and will allow hardware prices to decrease.
Developed by consortium of major companies with focus on interoperability.
Optimized for wireless LANs.
Uses radio frequency signals in unlicensed 2.4GHz band to send and receive data.
Uses Direct Sequence Spread Spectrum (DSSS) RF method.
Equipment dynamically selects lower data rates as RF signal quality decreases: 11, 5.5, 2, 1 Mbits.
Allows roaming among radio access points.

39. IEEE b Technology
802.11b defines how the RF channel is used, allowing multiple devices to communicate on the channel as if it were a wire.
This in itself can form a standalone network as if wired devices were connected to a hub.
The Access Point sits on both the wireless network segment (space) and the wired segment, acting as a bridge from the wireless to the wired segments.
A bridge forwards data packets from one side to the other at the MAC layer.

40. Range from AP to Computer
100 Feet
Access Point
5.5Mb
11 Mb
Computers
11 Mbits
5.5Mb
2 Mb
1 Mb
0 Mb
Range = 0 to 700 feet

41. Signal Strength Microsoft RADAR: 3 meters with 50% probability
Washington SpotON: distance measurement
Wochester Polytech: based on received signal strength, angle of arrival, time of arrival

42. IEEE : RADAR
It is using a standard network adapter to measure signal strengths at multiple base stations positioned to provide overlapping coverage in a given area

43. Signal Strength: 802.11b Lab empty One student walking around
8:11am
9:00am
Lab empty
9:45am
One student walking around
Group meeting

44. Microsoft RADAR Strength Weakness Easy to set up
Requires few base stations
Uses the same infrastructure that provides general wireless networking in the building
Weakness
Poor overall accuracy:
scene-analysis: within 3 meters with 50 percent probability
signal strength: 4.3 meters at the same probability
Support Wave LAN NIC

45. LANDMARC Prototype Proposed at MSU (PerCom 2003) Selection criteria
Use commodity products or off-the-shelf components
Low cost
Resolution: no more than 2-3 meters
Decision: RFID technology

46. RFID: RF Identification
RFID is a means of storing and retrieving data through electromagnetic transmission to a RF compatible integrated circuit
3 basic components

47. RFID Advantages Non-line-of-sight nature
RF tags can be read despite the extreme environmental factors : snow, fog, ice, paint …
be read in less than 100 milliseconds
promising transmission range
cost-effectiveness

48. Passive RFID vs. Active RFID
Active tag System

49. Using RFID: First Attempt
How many readers are needed?
Build an array of readers: too expensive
How reliable is the tag detection?
Not very reliable due to signal attenuation
Placement of RF readers
Cannot measure distance directly

50. Difficulties

51. LANDMARC Approach
The system mainly consists of two physical components, the RF readers and RF tags

52. Active RFID RF Reader Active Tag system Range up to 150 feet
Identify 500 tags in 7.5 seconds with the collision avoidance
Support 8 power levels (function of distance)
Active Tag system
Emit signal, which consists of a unique 7-character ID, every 7.5 seconds for identification by the readers
Button-cell battery (2-5 years life)
Operate at the frequency of MHz

53. The Concept of Reference Tags

54. Known Reference Tags Distance estimation Placement of reference tags
Selection of k neighboring reference tags
Weight of each selected reference tags

55. Distance Estimation: Signal Strength
Signal Strength Vector of an unknown tag
Signal Strength Vector of a reference tag
Euclidian distance

58. Effect of the Value K
Cumulative Percentile Of Error Distance When K Value Is 2, 3, 4, 5

59. Influence of The Environmental Factors
Cumulative Percentile Of Error Distance in Daytime & Night

60. Effect of the Number of Readers
Cumulative Percentile Of Error Distance With 3 or 4 Readers Data

61. Placement of Reference Tags
With Partition
Without Partition

62. Placement of Reference Tags
Replacements of the Reference Tags with a Higher Density

63. Effect of Higher Density Reference Tags
Cumulative Percentile Of Error Distance With Higher Reference Tag Density

64. Lower Density of Reference Tags
Replacements of the Reference Tags with a Lower Density

65. Effect of Lower Density Reference Tags
Cumulative Percentile Of Error Distance With lower Reference Tag Density

66. Overall Accuracy
Using 4 RF readers in the lab, with one reference tag per square meter, it can accurately locate the objects within error distance such that the largest error is 2 meters and the average is about 1 meter.

67. RFID Conclusions
RFID can be a good candidate for building location-sensing systems
Able to handle dynamic environments
Suffer some problems
Difference of Tags’ Behavior
RFID does not provide the signal strength of tags directly
Unable to adjust emitting interval
Standardization

68. Bluetooth Short range radio link for electronic devices
Scenario 1: Imagine when you walk into your office with a Bluetooth-enabled computer and you’re automatically connected to all the peripherals and devices
Scenario 2: Connect a Bluetooth-enabled cell phone to access the Internet through your desktop computer or wirelessly share files during meetings
Bluetooth name?
From Harald Blatand (Bluetooth in English), a Danish Viking King who united Denmark and Norway
Bluetooth unifying computing and telecommunication industries

69. Bluetooth Frequency Band
Specified for low cost, single chip implementation
US, Europe and most countries
2400 MHz to MHz
RF channel
F = k MHz for k=0 to 78
Frequency hopping
Hop rate: 1600 hops/second
Hopping sequence determined by the device address of the master
Some countries, e.g, France, are different

70. Bluetooth network topologyM S P sb
Radio designation
Connected radios can be master or slave
Radios are symmetric (same radio can be master or slave)
Piconet
Master can connect to 7 simultaneous or 200+ inactive (parked) slaves per piconet
Each piconet has maximum capacity (1 MSps)
Unique hopping pattern/ID
Scatternet
Piconets can coexist in time and space

71. The Piconet All devices in a piconet hop togetherB E
All devices in a piconet hop together
To form a piconet: master gives slaves its clock and device ID
Hopping pattern determined by device ID (48-bit)
Phase in hopping pattern determined by Clock
Non-piconet devices are in standby
Piconet Addressing
Active Member Address (AMA, 3-bits)
Parked Member Address (PMA, 8-bits) or

72. The Scatternet Complex master slave master/slave Access Point LAN
Printer
Laptop
Mouse
Mobile Phone
Headset
LAN
Access Point
slave
master
master/slave
Source: Kris Fleming 20Mar01 [Bluetooth-BOF-at-50th-IETF-PAN-Talk.ppt]

73. Bluetooth Protocol Stack

74. Target Markets The first wave 1. PC, Notebooks
2. Organizers & Palm Computers
3. Headsets
4. Cellular/PCS
5. Cordless phones
6. Automotive cellular
7. Digital cameras
8. PBX

75. Target Markets The third wave The second wave 1. Home networking
1. Printers
2. Photo printers
3. Fax machines
4. Industrial, musical and vertical industries products
The third wave
1. Home networking
2. Office networks
3. Video projectors
4. Set top boxes

76. Convergence Scenario
Source: Marc de Courville, Motorola, May00 IEEE Plenary Meeting

77. Conclusions Need low-cost location sensing devices
Accuracy, granularity, latency issues
Other sensing devices (e.g., Berkeley MOTE and TinyOS)
Cell phone locations
Transition between indoor and outdoor environments

78. Questions?