A Method of Map Matching for Personal Positioning System

Center for Spatial Information ScienceThe University of Tokyo

Kay Kitazawa, Yusuke Konishi, Ryosuke Shibasaki

Needs for accurate positioning

Information services which depend on one’s location

Platform

Mobile phone (e.g i- mode ) PDA

GPS receiver

P

P

P

the nearest Parking

Full or any vacancy ?

accurate positioning technology is needed

You are here

Problems with current positioning services

Complementary method is necessary

In the valley between high-rise buildings Underground passage Inside of buildings

GPS system is not always available

GPS Point

We have to get trajectory between GPS points

GPS Point

-10.000

-5.000

0.000

5.000

10.000

15.000

20.000

25.000

30.000

35.000

40.000

-20.000

-10.000

0.000 10.000 20.000 30.000 40.000 50.000 60.000

y

Observed tracks

Actual walking path

Complementary method

AccelerometerGyrocompassMagnetic sensor Barometer

Personal Positioning System (PPS) developed by Mr.Yusuke Konishi

Accumulation of errors is inevitable

Map matching

Human can move more freely

Map Matching

・Map matching methods for car navigation

・Map matching methods for PPS (Personal Positioning System )

A new matching algorithm

3-D database

Road network : Nodes & Edges

cars always run

along with network

2-D

Objective

Development of a new method of “Map matching”

track human’s movement

3-D database

Position ( x, y, z )

Combination of algorithm Global matching

Local matching

Matching in 3-D space

room

entrance

corridor

and polygons

Local matching

Obstacle boundaries

Merit : Frequent modification

Intersect boundaries or not ?

Boundaries

Demerit : Hard to correct once mismatching

happened

modified points

??

Possible tracks

Global information is needed

90°turn in between long straight paths

Distinct characteristics

Global matching

Turn at the corner

Road Network

Candidates

Check the fitness

Distance

Orientation

Nodes and edges

Demerit : Long interval between processesMerit : Global consistency

Combine and “switch” two matching algorithms

“Switch” matching algorithm

display

Locally modified

θ> threshold ?

Globally modified

θ

Sharp change in angle

Switch to global matching process

6th floor

Ground floor

5th floor

4th floor

2nd floor

7th floor

8th floor

9th floor

10th floor

Subjective area3rd floor

Place of experiments

Institute of Industrial Science,The University of TokyoView from the East side

staircases

Entrance

Result : walk along corridor (2-D)

Raw data

Corrected data

Result : enter the room (2-D)

Corrected data

Raw data

Entrance

Result : go downstairs (3-D)

Corrected data

Raw data matching nodes of the entrance of staircases

Conclusions

• The combined algorithm is effective

• “Natural and free” trajectory can be reconstructed with least geometric constrains(e.g only obstacle boundaries )

Future works

• Adjustment of the parameter value(e.g threshold of sharp change in angle)

• Extend the matching algorithm for movements in staircases

Discrimination of “action mode”

(e.g walking, pausing,going up/down stairs etc. )

Always apply network in staircases

vector

Flow chartMap Matching

End

Add positional data to Database

Get coordinates of estimatedposition and time (x, y, z, t )

(x, y, z, t )

Calculation of tracks (vector)&

Modification of initial errors

Recent value of local modification

θ・・・　angle error

Ratio of length

GlobalmatchingY

Data of the objects around thecurrent point is obtained

N

Local matching

Y

Return the data of estimated point intact

N

Sharp angle change?

θ

The track crosses the outline of objects?

Flow chart : 2

Global matching

Return

ReturnN

The distance < threshold?

Get Nodes and Edges around the estimated position

Calculate the distance between the current position

and each nodes and edges

d

Choose the nearest node

Y

d

Measure the angle made byeach edge connected to the node

θSelect edges with

least difference in angle

Match the corner and pointwith angle change

Transform the shape of tracks

Start

Return

Y

Return thevalue intact

Measure the angle made by tracks and outlines

Angle < threshold?

N

Intersect with entrance?

Entrance ?

Flow chart : 3

Return

Record coefficients for correctionof cross angle and length

N

N

Initial errors

Return modified points

Get track of 5 more steps

N

Local matchingY

Rotate tracks aroundthe first point counter-

clockwise for thecross angle

Return

Y

Relocate intersectionto the entrance

θ

Distance between entrance< threshold ?

Angle < 90°?Y

＊

θ

Return

Start

-45.00000000

-40.00000000

-35.00000000

-30.00000000

-25.00000000

-20.00000000

-15.00000000

-10.00000000

-5.00000000

0.00000000

5.00000000

10.00000000

0.00000000

5.00000000

10.00000000

15.00000000

20.00000000

25.00000000

30.00000000

系列1

-45.00000000

-40.00000000

-35.00000000

-30.00000000

-25.00000000

-20.00000000

-15.00000000

-10.00000000

-5.00000000

0.00000000

5.00000000

-5.00000000

0.00000000

5.00000000

10.00000000

15.00000000

20.00000000

25.00000000

系列1

Data 1

Estimated dataMatched data

y

-45.00000000-40.00000000-35.00000000-30.00000000-25.00000000-20.00000000-15.00000000-10.00000000-5.000000000.000000005.00000000

10.00000000

-10.0000000

0

-5.00000000

0.00000000

5.00000000

10.00000000

15.00000000

20.00000000

25.00000000

30.00000000

y

-45.00000000

-40.00000000

-35.00000000

-30.00000000

-25.00000000

-20.00000000

-15.00000000

-10.00000000

-5.00000000

0.00000000

5.00000000

-10.00000000

0.00000000 10.00000000

20.00000000

30.00000000

系列1

Data 2

Estimated dataMatched data

-50

-40

-30

-20

-10

0

10

20

-30 -20 -10 0 10 20 30 40

系列1

101.49

101.5

101.51

101.52

101.53

101.54

101.55

101.56

0 50 100 150 200 250

系列1

Data 3

Estimated data

Fluctuation in air pressure