Learning the meaning of places

savagelizardAI and Robotics

Nov 25, 2013 (3 years and 8 months ago)

106 views

Learning the meaning of
places

IfGi

Location based Services

SS 06


Milad Sabersamandari

Inhalt


Introduction


Existing place learning algorithms


Extracting Places from traces of
locations


Application with Bluetooth


Advantages and disadvantages


References

Introduction


Location learning systems


Locations are expressed in 2
principal ways


Coordinates


Landmarks


Intrested in „places“ (e.g. home,
work, cinema)


Introduction


Define „places“


Manually by hand


Rectangular region around an office
represented in coordinates


Automatically


Spends a significant amout of time
or/and visits frequently


-
> Place learning algorithms

Introduction


Locations based services


Location based reminder


Location based to
-
do list application


„Location based intelligent desicions
service“



Existing place learning algorithms



Ashbrook and Starner
´
s GPS
Dropout Hierachical Clustering
Algorithm (A&S)


The comMotion Recurring GPS
Dropout Algorithm


The BeaconPrint Algorithm

Ashbrook and Starner
´
s Clustering
Algorithm (A&S)



Loss of GPS signal of at least
t

minutes


Indicates a speed of continuilly below
1 mile per hour


Positions are merged (variant k
-
means clustering algorithm)


The comMotion Recurring GPS
Dropout Algorithm


GPS is lost three or more times
within a given radius


Merge the points to places


The BeaconPrint Algorithm



Fingerprint algorithm


Input: sensor log from mobile device


List of places the device went
(waypointlist)


GSM and 802.11

The BeaconPrint Algorithm



1. Segment a sensor log into times when
the device was in a stable place and
assign a waypoint.


2. Merge waypoints which are captured
from repeat visits to the same place.


Likewise, an effective recognition
algorithm has two capabilities:


1. Recognize when the device returns to a
known place using a waypoint list.


2. Recognize when the device is not in a place
We refer to this state as mobile.


Extracting Places from traces of
locations



Uses Place Lab to collect traces of
locations


In many cities and towns available


Place Lab works in urban areas
aswell as indoors


Location recorded once per second


Places appear as clusters of locations


Extracting Places from traces of
locations



Place Lab


Uses that each WiFi access point
broadcasts its unique MAC address


A database maps these addresses to
longitude and latidute coordinates


Existing clustering Algorithm


k
-
means Algorithm


Gaussian mixture model (GMM)


Require the number of clusters as a
parameter


Require a significant amout of
computation

Time based clustering


Eliminate the intermediate locations
between important places


Determine the number of clusters
(important places) autonomously


Simple enough to run on a simple
low battery mobile device

Time based clustering


Basic idea is to cluster along the time
axis


New measured location is compared
with previous locations


Decide if the mobile device is moving


Parameter:distance
d
between the
locations and a cluster
´
s time
duration
t

Time based clustering


Parameter: distance
d
, time
t


Current cluster
cl


Pending location
ploc


Significant places
Places

Time based clustering

Time based clustering


Unlike other clustering algorithms
this algorithm computes the clusters
incrementally


The computation is simple


Easily supported on small battery
mobile devices

Application with Bluetooth



Bluetoothcell with radius
r


Bool value for each cell


Short distance


Time duration of 11 seconds

Application with Bluetooth


Application with Bluetooth



Application with Bluetooth


Replace


Measured location
loc


measured BTcell
cell


Pending location

ploc


pending BTcell

pcell


Current cluster

cl


as a set of BTcells


Advantages and disadvantages



GPS (Advantages)


Standardized


Covers most of the earth
´
s surface


Continually decreasing in cost


GPS (Disadvantages)


Inability to function indoors


Occasional lack of geometry accuracy


Loss of signal in urban canyons and
other „shadowed“ areas




Advantages and disadvantages



Bluetooth (Advantages)


Standardized


3 classes (different ranges)


Everywhere available (indoor)


Bluetooth (Disadvantages)


Short distance


Long time duration


Accuracy = 1 Bluetoothcell


Bad java support


References

1.
Jong Hee Kang, William Webourne, Benjamin Stewart, Gaetano
Borrielo.
Extracting Places from Traces of Locations

2.
Jeffrey Hightower, Sunny Consolvo, Anthony LaMarca, Ian
Smith, Jeff Hughes†.
Learning and Recognizing the Places We
Go

3.
John Krumm, Ken Hinckley.
The NearMe Wireless Proximity
Server

Vielen Dank für Ihre
Aufmerksamkeit