Adaptive Temporal Radio Maps for Location Estimation

1
Adaptive Temporal Radio Maps for Location
Estimation

• Jie Yin, Qiang Yang and Lionel M. Ni
• Computer Science
• Hong Kong University of Science and Technology
• Hong Kong, China

2
Introduction

• Location estimation based on IEEE 802.11b
  wireless networks has gained increasing attention
• GPS does not work indoor
• Radio Frequency (RF) wireless networks
• Challenges
• The signal strength is inherently uncertain due
  to multi-path fading effects
• Environmental dynamics affect the signal strength
  to a large extent

3
Static Localization

• Input

• Output
• Given a new signal-strength vector
• lt-63,-87,-79gt
• ? predict a mobile user to be in location2

4
Common Approaches

• Based on a two-phase algorithm
• Offline training phase
• Build a radio map by calibrating signal-strength
  samples from access points at selected locations
  
• Online localization phase
• Real-time signal-strength samples are used to
  search the radio map to estimate the locations

5
Related Work

• The RADAR system Infocom00
• Deterministic nearest neighbor
• Robotics based system Mobicom02
• The Bayesian inference method
• Post-processing used for refinement
• Joint clustering based system Percom03
• Maximum Likelihood (ML) method

6
Problem with Static Localization

• Assumption static radio map
• Once learned in the offline phase, a radio map is
  applied to estimate locations in any later time
  periods without adaptation
• ? Inaccurate location estimation
• The signal-strength samples collected in the
  online phase may significantly deviate from those
  stored in the radio map.

7
Problem with Static Localization

• Signal-strength distributions vary over different
  time periods
• Naïve approach repeatedly rebuild radio maps
• ? Too tedious manual effort
• ? Does not work day-to-day variations

8
Our Solution

• Key idea adapt the radio map based on reference
  points using a regression analysis

9
Definitions

• Location space L
• Each tuple represents a users location and
  orientation
• p access points and m reference points
• The signal strength received by a mobile client
  is defined as
• The signal strength received at the kth reference
  point is

10
Two-phase Algorithm (1)

• During the offline phase (time period t0),
• At each location, we learn a predictive function
  fij for the jth access point
• fij indicates the relationship between the
  signal-strength values received at reference
  points and the value received by the mobile client

11
Two-phase Algorithm (2)

• During the online phase (time period t),
• Based on the signal strength received at
  reference points, we compute the estimated
  signal-strength vector
  for each location using fij
• The signal strength received by the mobile client
  is referred as
• Compute the Euclidean distance Di and output the
  location with minimum distance

12
Critical Issue

• To learn the predictive function fij between the
  signal-strength values received by the mobile
  client and the reference points.
• Two algorithms via regression analysis
• A multiple-regression based algorithm
• A model-tree based algorithm

13
Multiple Regression

• The signal strength received by the mobile client
  is computed as a linear aggregate of the signal
  strength received at m reference points
• Regression coefficients represent the
  independent contributions of each reference point
• Perform least square estimation to compute
  regression coefficients in the offline phase

14
Model Tree

• Piecewise linear approximation
• Multiple regression vs model tree

15
Model Tree

• A model tree is a binary decision tree with
  linear regression functions at the leaf nodes

16
Experimental Setup
Access points Reference points
A section of office area in CS department
17
Data Collection

• Reference points on every other hour from early
  morning to midnight (800AM- 1200AM)
• Grid points
• Each position is 1.5 meters apart
• 450 samples per position with two directions
• Evaluation Data
• The group of data collected at midnight 1200AM
  for training
• The other eight independent groups for testing

18
Impact of Environmental Factors

• Comparison of overall accuracy over different
  time periods

19
Impact of Environmental Factors

• Comparison of accuracy within 1.5 meters over
  different time periods

20
Impact of Reference Points

• Comparison of average accuracy vs the number of
  reference points at 800AM

21
Impact of Access Points

• Comparison of average accuracy vs the number of
  access points at 1200PM

22
Conclusions

• We present a novel localization method based on
  temporal radio maps
• Reference points and regression analysis
• Our proposed method can adapt better to the
  variations of signal strength caused by
  environmental dynamics

23
Future Work

• Apply more effective probabilistic methods to
  build radio maps using the signal strength
  received at reference points
• Incorporate a users movement trajectories to
  improve accuracy
• Test the validity of our proposed algorithms in a
  larger-scale environment

24
Questions?

• Thank you