You are on page 1of 94

EasyTracker

Automatic Transit Tracking, Mapping, and


Arrival Time Prediction Using Smartphones
James Biagioni, Tomas Gerlich,
Timothy Merrifield and Jakob Eriksson

We love bus trackers!

slide 2

Winter in Chicago

slide 3

Our shuttle web (before)

slide 4

Our shuttle web (before)

slide 4

Our shuttle web (after)

slide 5

One service for everyone

slide 6

Our shuttle web

slide 7

Our shuttle web

slide 7

This paper in a nutshell

Automatic generation of
route shapes
- stop locations
- schedules
-

Online processing for


vehicle-to-route classification
- arrival-time prediction
-

slide 8

EasyTracker installation

slide 9

EasyTracker installation
1. Obtain smartphone

slide 9

EasyTracker installation
1. Obtain smartphone
2. Install EasyTracker app

slide 9

EasyTracker installation
1. Obtain smartphone
2. Install EasyTracker app
3. Stick phone in bus

slide 9

4. Relax

slide 10

System overview
S
P
G

GP

slide 11

System overview
S
P
G

GP

slide 11

Batch processing

slide 12

Batch processing

slide 12

Raw GPS traces

slide 13

Route map

slide 14

Raw GPS traces

slide 15

Kernel Density Estimation

slide 16

Kernel Density Estimation

slide 16

Kernel Density Estimation

slide 16

Kernel Density Estimation

slide 16

Kernel density estimation


1
f(x) =
n

n
X

K(x

xi )

i=1

K(x) = p

1
2 2

slide 17

x2
2 2

2-D histogram

slide 18

Trajectory density estimate

slide 19

Thresholded image

slide 20

Map extraction

Davies et al., 2006


slide 21

Route extraction

Map match GPS traces


Viterbi-based map matching
- based on Thiagarajan, et al. 2009
-

Extract common routes


edge subsequence matching
- statistical test removes spurious results
-

slide 22

Route extraction

slide 23

Route extraction

slide 23

Route extraction

slide 23

Route extraction

slide 24

Route extraction

slide 24

Route extraction

slide 24

Route extraction

slide 24

Route extraction

slide 24

Route extraction

slide 24

Route extraction

slide 24

Route extraction

slide 24

Route extraction

slide 24

Route extraction

slide 24

Route extraction

slide 24

Route extraction

slide 24

Route extraction results

slide 25

Welchs t-Test
0.030

0.023

0.008

slide 26

P-value

0.015

Routes separated
Real Routes

Spurious Routes

slide 27

Stop extraction

slide 28

Route-labeled GPS traces

slide 29

2-D histogram

slide 30

Point density estimate

slide 31

Thresholded binary image

slide 32

Noise in binary image

slide 33

Noise reduced binary image

slide 34

Stop extraction

slide 35

Stop extraction

slide 36

Stop extraction performance

slide 37

Schedule extraction

slide 38

Bus stop arrival times

slide 39

Bus stop arrival times

slide 39

Bus stop arrival times

slide 39

Bus stop arrival times

slide 39

Bus stop arrival times

slide 39

Bus stop arrival times

slide 40

First stop schedule

slide 41

Travel time variance

slide 42

Last stop arrival times

slide 43

Compute mean travel times

1 X t
aj
travel time(1, j, t) =
|D|
D

slide 44

t
a1

Compute downstream
schedules

j
kt

1
kt

+ travel time(1, j, t)

slide 45

Last stop arrival times

slide 46

Last stop schedule

slide 47

Schedule accuracy

slide 48

Schedule accuracy

slide 48

System architecture

slide 49

Online processing

slide 50

Online processing

slide 50

Un-classified buses

slide 51

Classified buses

slide 52

Hidden Markov model

slide 53

Classification accuracy
1.0
0.8
0.6
0.4
0.2
0

Correct

Incorrect

slide 54

Unclassified

Classification delay

slide 55

Classification delay

slide 55

Arrival time prediction

slide 56

Predicting arrival times

time until arrival(si ) = travel time(sprev+1 , si )+


(1

slide 57

)travel time(sprev , si )

Arrival time predictions

slide 58

Schedule vs. real-time

slide 59

Schedule vs. real-time

slide 59

Schedule vs. real-time

slide 59

System overview
S
P
G

GP

slide 60

Come and see our demo!

Thursday, 3:30p-7:30p
slide 61

Thanks!
Questions?