The document summarizes research on representing spatio-temporal data from multiple sources in urban mobility analysis. It proposes a framework with concepts like observations, places, stays, and trajectories to integrate GPS, Wi-Fi, and GSM data. The concepts are validated by mapping real-world smartphone data to observations and places. Future work includes handling larger datasets, improving the place learning algorithm, and extending the analysis to groups.

1. ICCSA 2012
Dealing with multiple source spatio-
temporal data in urban dynamics analysis
João Peixoto and Adriano Moreira, Mobile and Ubiquitous Systems Group

2. MOTIVATION
The mobility of citizens in an urban area is the source of various
problems: traffic congestion, environmental
impacts, inadequacy of public transport, and spreading of
diseases…

3. MOTIVATION
It is important to understand the mobility behaviour of
individuals in space, understand space itself, and understand the
use people make of the urban space

4. MOTIVATION
The dynamics associated with the mobility in urban areas
always has two components, Time and Space, creating new
challenges

5. MOTIVATION
The current Geographic Information Systems are structured to
represent the spatial component of data but lack good support
for the temporal component (Yu and Shaw, 2004)

6. MOTIVATION
Detect the presence and mobility of people in urban spaces
requires the collection of data

7. MOTIVATION
The huge size of datasets being collected these days is creating
more challenges to representation and visualization rather than
solutions

8. RELATED WORK
Temporal snapshots of space occupation
Due the dynamics of the urban space, this approach may not be
the most effective for the analysis of pattern changes
(Hagen-Zanker and Timmermans 2008)
Reades, J., Calabrese, F., Sevtsuk, A., Ratti, C. (2007)

9. RELATED WORK
Trajectories with source-destination
• Large interval between samples  we lose intermediate movements
• To connect the source to destination we may have to affect the Time
component
Brockmann and Theis (2008)

10. BASIC CONCEPTS
TRAJECTORY
Our initial goal
TIME LEAP
SPACE LEAP
ELEMENTARY MOVEMENT
Create a flexible and comprehensive framework for the
spatio-temporal representation of movement data
STAY
OBSERVATION PLACE
RAW DATA

11. BASIC CONCEPTS
TRAJECTORY
Our initial goal
TIME LEAP
SPACE LEAP
ELEMENTARY MOVEMENT
To integrate different types of data from different sensors
To apply different scenarios of urban mobility
STAY
OBSERVATION PLACE
RAW DATA

12. BASIC CONCEPTS
It all starts with the Raw Data collected by a multitude of
sensors
RAW DATA

13. BASIC CONCEPTS
The observation of an artefact in a specific point of a spatio-
temporal space
(Id_Observation, Artefact, Location, Timestamp)
OBSERVATION
RAW DATA

14. BASIC CONCEPTS
Transformation process between Raw Data and Observation
OBSERVATION
RAW DATA

15. BASIC CONCEPTS
Based on Observations we extract the Places
OBSERVATION PLACE
RAW DATA

16. BASIC CONCEPTS
Time interval between the first and last observation of an
artefact in the same place
(Id_Stay, Artefact, Place, Timestamp_Initial, Timestamp_Final)
STAY
OBSERVATION PLACE
RAW DATA

17. BASIC CONCEPTS
A Change of Location of an artefact occurred over time
(Id_Movement, Artefact, Location_Start, Location_End, Timestap_Initial, Time
stap_Final)
ELEMENTARY MOVEMENT
STAY
OBSERVATION PLACE
RAW DATA

18. BASIC CONCEPTS
SPACE LEAP
A Change of Location of an artefact occurred over a long time
ELEMENTARY MOVEMENT
period
(Id_SpaceLeap, Artefact, Location _Start, Location _End, Timestap_Initial,
STAY
Timestap_Final)
OBSERVATION PLACE
RAW DATA

19. BASIC CONCEPTS
TIME LEAP
SPACE LEAP
Long time period between two sequential observations of an
ELEMENTARY MOVEMENT
artefact in the same place
STAY
(Id_TimeLeap, Artefact, Place, Timestamp_Initial, Timestamp_Final)
OBSERVATION PLACE
RAW DATA

20. BASIC CONCEPTS
TRAJECTORY
TIME LEAP
SPACE LEAP
ELEMENTARY MOVEMENT
Time-ordered list of ElementarySTAY
Movements of an artefact over
the space
OBSERVATION PLACE
(Id_Trajectory, Artefact, List of Elementary Movements)
RAW DATA

21. MAPPING DATA INTO THE FRAMEWORK
Goal: validate the concepts of our proposed framework for the
representation of spatio-temporal data

22. MAPPING DATA INTO THE FRAMEWORK
Our focus in this paper is only on three concepts:
Observation, Place and Stay

23. MAPPING DATA INTO THE FRAMEWORK
Android Smartphone Application that collects data from three
different types: GPS, Wi-Fi and GSM.

24. MAPPING DATA INTO THE FRAMEWORK
Raw Data
Timestamp Latitude Longitude Altitude Speed Accuracy Bearing
2011/06/29 15:25:07 1,297077 103,7808 93,5 0,75 17,88854 65
2011/06/29 15:25:18 1,297077 103,7808 108,2 0,75 26,83282 162,4
2011/06/29 15:25:31 1,297213 103,7806 134,4 1 40 283,8
Timestamp BSSID RSSI SSID
2011/06/29 15:25:08 00:27:0d:07:d6:c0 -90 NUS
2011/06/29 15:25:11 00:27:0d:07:d6:c0 -88 NUS
2011/06/29 15:25:12 00:27:0d:07:d6:c0 -88 NUS
Timestamp CID LAC MNC SIGNAL_STRENGTH
2011/06/29 15:25:08 962335 441 3 9
2011/06/29 15:25:10 962335 441 3 8
2011/06/29 15:25:11 962335 441 3 8

25. MAPPING DATA INTO THE FRAMEWORK
Observations
Timestamp Location Optional Attibutes
Position Symbolic Name Sensor_type
Latitude Longitude
15:25:07 1,297077 103,7808 GPS
15:25:08 00:27:0d:07:d6:c0 WIFI
15:25:08 962335 GSM
15:25:10 962335 GSM
15:25:11 00:27:0d:07:d6:c0 WIFI
15:25:11 962335 GSM
15:25:18 1,297077 103,7808 GPS

26. MAPPING DATA INTO THE FRAMEWORK
Place Learning
Psameplace(oi, oj)
Prob. function GPS Wi-Fi GSM
GPS P1 P2 P3
Wi-Fi P2 P4 P5
GSM P3 P5 P6

27. MAPPING DATA INTO THE FRAMEWORK
Results
Results - Places
– Place is described by its GPS part, Wi-Fi part, and GSM
part
– If the total accumulated time spent at that place is longer
than a minimum of two minutes  Place
– For a single person we detect 13 different Places
– If the time elapsed between consecutive observations in a
place do not exceed a given threshold (Tmax = 60
seconds)  Stay

28. MAPPING DATA INTO THE FRAMEWORK
Results

29. CONCLUSIONS AND FUTURE WORK
• The proposed concepts and framework are
appropriate to represent the three types of records
used.
• Additional concepts might also need to be defined
– Trajectory is only linked with Elementary Movement
• Include another's sensors to validate the concepts
(for example: ticketing data used in buses)

30. CONCLUSIONS AND FUTURE WORK
• Process massive datasets
– Space occupied at the level of storage
– Aggregate a large number of records
• Validate the place learning algorithm and try
different approaches
• Extend the study to groups of citizens
– Popular Places
– Popular Flows

31. THANK YOU !
joao.peixoto@algoritmi.uminho.pt
adriano.moreira@algoritmi.uminho.pt
Mobile and Ubiquitous Systems Group
Research group supported by FEDER Funds through the COMPETE and National Funds
through FCT – o para a Ciência e a Tecnologia under the Project: FCOMP-01-FEDER-
0124-022674.