A Positioning System Using Distributed Mobility Pattern Speculation

1
A Positioning System Using Distributed Mobility
Pattern Speculation

• Takamichi Ishihara, Nobuhiko Nishio
• Department of Computer Science, Ritsumeikan
  University

2
Outdoor Location SystemOur Past Approach

• We want to obtain the location
• Outdoors
• With the smallest device possible
• WiFi-Based
• Users have mobile devices with WiFi.
• WiFi-MAP is prepared from which position is
  estimated.
• Location Estimation Algorithm
• Simple rectangle are used.(Cell-ID)
• Use of signal strength to improve accuracy
• Signal Strength is classified into three
  categories weak, normal, and strong
• Depending on strength the size of the rectangle
  is varied

3
Initial Location Estimation Algorithm
Rectangle B
Logged point
Rectangle A
Max Lat
Min Lon
Max Lon
Min Lat
Rectangle C
The smallest rectangle that encloses all logged
points
Presumed range in present location
Estimated location
4
Experimental Result in the SuburbsAlgorithm
Performance Comparison
5
WOWnet Project

• Aim Community-Based Security System
  Industry-Academia-Civilian Collaboration Project
• Transmit an SOS from a childs wireless device
• Individual nearby the Base-station receives the
  SOS and arrives on the scene

Size of device assumed in the future
6
Application

• Table-sized demo of context-aware services for
  cities

Smart Advertising - Sending advertisements to
users based on their own movement history
Personal Navigation - Pointing users in the
direction of their destination in real-time
7
Applying Mobility Pattern Matching

• In algorithms up to now, location accuracy isnt
  high in places where the density of APs is low.
• Using Mobility Pattern
• Improve current algorithm precision
• Estimating location where APs are not accessible
• Recognition of irregular movement

8
Preliminary Experiment With Mobility Pattern
Matching

• Initially, a mobility pattern was defined as
• The temporal sequence of APs encountered while
  walking along a certain route.
• i.e. the route between your home and school
• In this preliminary experiment, the mobility
  pattern was made in an approx. 300m section of
  Ritsumeikan University.

9
Experimental Result for Campus PrototypeMobility
Pattern Usage
10
Proposed System

• From this experiment, the effectiveness of
  mobility pattern usage is confirmed.
• But, handheld devices cant hold all mobility
  patterns.
• Therefore,
• We propose A Positioning System Using
  Distributed Mobility Pattern Speculation.

11
Now, how is mobility pattern data defined?

• Defined as the smallest unit of a mobility
  pattern
• A location on the line segment between two
  points
• Method for selecting end-points
• Eligible points
• Strongest Signal Point Point within a Rectangle
  where the signal is strongest
• AP Boundary Points Points where the AP is first
  and last heard
• The path that connects the selected end-points si
  one pattern.

12
Generation of Mobility Pattern Data

• Mobility pattern data is generated in parallel
  with the WiFi-MAP Generation stage (War Driving).
• WiFi-MAP is made with an existing method.
• When generating mobility pattern data, the values
  returned by GPS along the path between eligible
  points are logged (latitude, longitude, time).
• Two or more patterns are generated.

13
Generation of Mobility Pattern Data
WiFi-MAP
1 mobility pattern
GPS data point
There are 12 eligible points, so 11 patterns are
generated.
Path of movement
Strongest Signal Point
AP Boundary Point
14
How is location estimated by using mobility
patterns?

• A handheld device has a WiFi-MAP and a part of
  the mobility pattern data.
• Location is not estimated by the measurements
  momentary signal condition only.
• The speed of the device is obtained from the
  travel time between the edge points of the
  mobility pattern.
• Location between the edge points is then
  estimated based on the speed and mobility pattern
  data.

15
Decentralization of mobility pattern data

• How is mobility pattern data grouped for the
  purpose of decentralization?
• Presently the following criteria are used
• Time( i.e. all points that can be walked to
  within 5 min.)
• Distance ( i.e. 1km2)

focused set of patterns
16
Revision of WiFi-MAP and Mobility Pattern Data

• Mobility pattern data usage is limited to routes
  along which WarDriving(WiFi-MAP Generation) was
  done.
• In practice, both the WiFi-MAP and mobility
  pattern data renewal architecture are necessary.
• Design of suitable data structure for revision

17
Challenge

• Lets assume that there is WiFi-MAP in an area
  where there is no mobility pattern data.
• Locky.jp etc.
• Using that WiFi-MAP it may be possible to
  generate imaginary mobility pattern data.

Area without neither WiFi-MAP nor mobility
pattern
18
Current Research Plan

• Implement a prototype of the newly defined
  movement pattern usage.
• Initially leaving out the considerations for map
  revision.
• Verify the effectiveness of the new proposal.
• Is the eligible point selection method
  appropriate?
• Mobility pattern location estimation algorithm
• etc