Location Systems for Ubiquitous Computing Jeffrey Hightower and Gaetano Borriello
Intro Ubiquitous computing a person wanting to know where he was when he did a particular task Help rescue teams Customize environment based on location of user devices have already been developed what they sense and how they go about achieving it physical attribute used size power usage type of results obtained
Physical Position and Symbolic Location Physical GPS - 47°39´17’’ N by 122°18´23’’ W Symbolic Abstract, relative to the position of a known object Provide coarse grained location information Derived from Physical-positioning systems Linking real-time train positions to the reservation and ticketing database can locate a person on a train
Absolute versus Relative positioning Absolute location systems – GPS uses a universal reference grid Two GPS receivers at the same position will show the same reading In Relative Systems, each receiver has its own frame of reference Devices that use a particular transmitter form a grid relative to that transmitter Absolute position can be transformed to a relative one – relative to another reference point
Localized Location Computation Object we are interested in computes its own location Ensures privacy Does not require the object to transmit information for external systems to locate it Burden on the object increases so it is better left to the external system
Accuracy and Precision Depend on the distribution of error and the density of elements Overlapping levels of positioning systems to obtain fine grained location information Coping dynamically with failures Suitability for application at hand
Scale Coverage of system, the number of objects the system can locate per unit area per unit time Communication bandwidth is important Increasing infrastructure
Recognition Recognition of located objects to carry out some action, like controlling the located device over the internet Assigning unique IDs to the located objects Combine contextual information
Limitations GPS does not work indoors Interference Characteristics of underlying technologies
Active Badge Active Bat Cricket RADAR Motionstar Magnetic Tracker Easy Living Smart Floor Enhanced 911
Active Badge Uses diffuse infrared technology - flooding an area with infra-red light Each badge emits signal with unique id every 10 seconds that is received by a network of sensors Location is symbolic – restricted area like a room Range of several meters Has difficulty in presence of sunlight
Active Bat Infers location based on time of flight of ultrasound pulse Each bat emits an ultrasound pulse with unique id to a grid of receivers At the same instant a controller resets the receiver Orientation is calculated by analysis Distance is computed from the time interval between the reset and receiving the pulse Accurate to within 3cm Paging Requires large sensor infrastructure
Cricket fixed ultrasound emitters and mobile receivers time gap to receive the signal is also set in the pulse to prevent reflected beams computation takes place at receiver decentralized architecture few centimeters of accuracy computational and power burden
RADAR Based purely in software, building on standard RF wireless LAN technology Uses signal strength and signal to noise ratio from wireless devices Employs multiple base stations with overlapping coverage Requires wireless LAN support on objects being tracked Generalization to multifloored buildings is a problem
Motionstar Magnetic Tracker Uses electromagnetic sensing Axial DC magnetic-field pulses are generated Position and orientation are found from by measuring the response on the three axes Less than 1mm spatial resolution and 0.1° orientation Must be within 1-3 meters of transmitter Motion capture for animation
Easy Living System to keep track of a room's occupants and devices Uses real-time 3D cameras to provide vision positioning measures location to roughly 10 cm on the ground plane, and it maintains the identity of people based on color histograms Difficult to maintain accuracy Aimed for a home environment
Smart Floor System for identifying people based on their footstep force profiles Does not need device or tag 93% overall user recognition High cost factor
Enhanced 911 Locates any phone that makes a 911 call reported in most instances with an accuracy of 100 meters or less Can be enhanced for use by cell phone users Identifying areas of traffic congestion
Future Work Integrating multiple systems Overlapping levels off sensing Increases accuracy Ad Hoc Location sensing Cluster of ad hoc objects Relative or absolute Correlation of multiple measurements High scalability
Choosing a System Accuracy based comparison Representing error distributions Evaluation Density of elements Prototyping using a simulator Quake iii