mobility data management - infolab.cs.unipi.grinfolab.cs.unipi.gr/tutorials/handouts/trento... ·...

1

Mobility Data Management

Yannis Theodoridis & Nikos Pelekis

InfoLab, University of Piraeus, Greece infolab.cs.unipi.gr

Trento, November 2011

2

A guided tour on Mobility Data Mgmt

I. Mobility data storage and querying

Acquiring trajectories from raw data; Location-aware querying; Efficient trajectory indexing and storage in MODs

II. Mobility-aware applications and tools

Location-based services and tools; Algorithms and operations for LBS

III. Mobility data analysis and mining

Trajectory warehousing and OLAP; Mobility data mining and reasoning; Visual analytics for mobility data

IV. Privacy aspects

Preserving user traces’ anonymity

V. Outlook

Open issues – Future Challenges

2

3

Mobility-aware applications

From location-based services

to mobile social networking

4

Introduction

� Large diffusion of mobile devices, mobile services and location-

based services

3

5

Introduction (cont.)

� Definitions:

� (Swedberg, 1999) Mobile Location System (MLS) is a location system,

including applications, that

� determines the geographic position of mobile subscribers and

� provides them with relevant information and services

� (source: Wikipedia) A Location-Based Service (or LBS) in a cellular

telephone network is a service provided to the subscriber based on

her current geographic location

6

Introduction (cont.)

� Definitions:

� (Swedberg, 1999) Mobile Location System (MLS) is a location system,

including applications, that

� determines the geographic position of mobile subscribers and

� provides them with relevant information and services

� Geo-positioning technologies

� Satellite vs. mobile telephone network infrastructure (see Lecture1)

� Geo-related information & services (LBS); a great variety

� from Routing and Find-the-nearest …

� … to Google Latitude and Facebook Places (see slides that follow)

4

7

Location-based Services & Tools

� Navigation (vehicle or pedestrian) & Information services

� Routing, finding the nearest point-of-interest (POI)

� Location-based yellow pages (“what-is-around”, …)

� Resource management & Tracing services

� (taxi, truck, etc.) fleet management, administration of container

goods, location-based charging

� Tracing of a stolen car, locating persons in an emergency situation, …

� Social networking apps

� Google Latitude, Facebook places, Foursquare, etc.

8

Location-based Services & Tools (cont.)

� On-board navigation

� Example: Dash express (source: www.dash.net)

� Internet-connected automotive navigation system

� Up-to-minute information about traffic

� Yahoo! Local search for finding POIs

5

9


� Find-the-Nearest

� finds the nearest point of interest (POI) - restaurant, gas station, etc.

suggests the optimal way to move there (Router)

� ‘optimal’ in terms of minimum distance (driving mode)

or minimum time (walking mode)

� E-911 (US), E-112 (EU), E-119 (Japan) emergency

call numbers

� mandates require locating the caller within

a few meters (150 feet, by FCC regulation)

� (in US) E-911 call centers receive 15,000 calls from cellular phones per

day

10


� “Find my iPhone” (launched Jun.09 by Apple)

� Where is my iPhone?

source: apple.com

6

11


� “Volvo connectivity” (launched Jun.11 by Volvo car manufacturer)

� Where is my Volvo?

� Mobile app with a wide-cut area of trait, such as:

� Vehicle locater. The lodging of the

vehicle is shown on a map.

� …

12


�

See in real time where your friends

are! (launched Feb.09 by Google)

7

13


� Google places: Rate and share

places on Google

� Google places for business:

� (business perspective) Get your

business found on Google

� (end user’s perspective) rate

products, search for similar, compare

prices, etc.

14


� Tag yourselves and find

tagged friends

� Tell your friends where you are,

suggest places, etc.

8

15


� As well as adventure apps, just for

fitness and fun ☺

� Zombie, Run! for Android (by

www.zrli.org)

� The Zombie, Run! Zombie Finder is a

survival tool for people living in a

zombie-infested area. Zombie, Run!

displays the location of nearby zombies,

helping you safely navigate your urban

or rural area.

16

Back to technical stuff: LBS taxonomy

� According to (Gratsias et al. 2005; Frentzos et al. 2007)

Reference

object

(user)

Database

objects

stationary mobile

stationary

⇒ what-is-around

⇒ routing

⇒ find-the-nearest

⇒ guide-me

mobile ⇒ find-me ⇒ get-together

Almost every current commercial app is classified as stationary-stationary !!

9

17

Routing

� (“Get Directions” in Google maps)

� Find the optimal route from a departure to a destination point

� Goal to be optimized: the minimal network distance traveled or the

minimal traveled time or …

� The best path wrt. distance is not necessarily the best wrt. time!

� Relates to the well-known

shortest path (SP) problem from

graph theory and network analysis

� Several off-the-shelf solutions, e.g.

Dijkstra’s single-start SP algorithm

Live example:

Get from Athens (GR) to Aalborg (DK) http://g.co/maps/w9cmh

18

Routing (cont.)

� Relates to the well-known shortest path (SP) problem from graph

theory and network analysis

� Several off-the-shelf solutions, e.g. Dijkstra’s single-start SP algorithm

� Example of Dijkstra’s SP:

“find the SP from A to D”

10

19

Routing (cont.)

20

What-is-around

� (“Search nearby” in Google maps)

� Retrieve and display all POI located in the surrounding area

(according to user’s location or an arbitrary point)

� Area could by rectangular or circular or of any other shape

� Example: “Provide me all the gas-stations and ATMs within a distance

of 1km”

� It isn’t more complex than a typical range query!

� Recall the slides about range query processing using R-trees!

Live example:

Search for hotels around Acropolis, Athens (GR) http://g.co/maps/fbg3m

11

21

What-is-around

22

Find-the-nearest

� (not supported by Google maps!)

� Retrieve and display the nearest POI (restaurants, museums, etc.)

with respect to a reference location

� Example: “find the two restaurants that are closest to my current

location” or “find the nearest café to the railway station”

� Issue: distances are computed wrt. the network rather than the free

space (Euclidean distance)

� State-of-the-art algorithm (at the spatial database domain):

� “Euclidean Restriction” (Papadias et al. 2003)

Live example: find the 1-nearest hotel to Acropolis, Athens (GR)

12

23

Find-the-nearest

Closest wrt. Euclidean distance Closest wrt. network distance

24

Find-the-nearest (cont.)

� Euclidean Restriction

� 1st step: Find object P1 which is nearest to Q wrt Euclidean distance

� 2nd step: D1 is the network distance between P1 and Q

13

25

Find-the-nearest (cont.)

� Euclidean restriction

� 3rd step: Retrieve all points (Pi) with Euclidean Distance from Q less than D1

� 4th step: Among all Pi, find the one with the minimum network distance from

Q

26

Advanced LBS

� Examples

� Constrained routing (advanced wrt. geometry / complexity)

� alternative names: in-route find-the-nearest (Frentzos et al. 2007),

trip-planning-query (Li et al. 2005)

� “Get-together” or “Find-me” (motion is involved)

� See also Google Latitude, Zombie, Run!, …

14

27

Constrained routing

� Find the optimal route from a departure to a destination point …

� … with the extra constraint to pass through one among a specified set

of candidate points

� Example: “provide me the best route office to home constrained to

pass from a bank ATM ”

28

Constrained routing

� Any idea for pruning?

15

29

Constrained routing (cont.)

� Variation of Euclidean Restriction

� 1st step: Retrieve the optimal route R between P and Q

30



� 2nd step: Retrieve the Euclidean NN w.r.t. route R (point N1)

16

31



� 3rd step: Calculate network distances D1=Net_Dist(P,N1) and

D2=Net_Dist(N1,Q)

32



� 4th step: Retrieve all candidate objects (Ni) with total Euclidean distance from

both P and Q less than D = D1 + D2

Perpendicular bisector

Circumference

17

33



� 5th step: For each Ni, calculate D1i = Net_Dist(P,Ni) and D2i = Net_Dist(Ni,Q),

and report the Ni the one that minimizes Di = D1i + D2i

34

“Get-together” (or “Find me”)

� A number of moving objects try to meet a moving object

� Think of a live version of the popular Pacman game !!

� Could be an advanced “Google Latitude” or “Facebook Places” app

� Methodology:

� other objects are routed to a point where

the target object is estimated to arrive

after a time interval

� this “meeting point” is periodically

refreshed

18

35

Get-together (cont.)

� Challenge: future location prediction !!

36

Summarizing …

� So far, we’ve seen a variety of mobile-aware applications

� From ‘smart’ (updated in real-time) routing to mobile social networking (Google Latitude, Facebook Places, etc.)

� Internally (either at server or client side), what kinds of operationsshould be efficiently supported?

� Core database / computational geometry operations

� Graph-based operations

� Trajectory management operations

19

37

Operations to be supported


� Window search: selects points within a (circular or rectangular) window

� k-NN search: selects the k- nearest points to an object (point or region)

� Euclidean distance calculation: calculates the Euclidean distance between two points

� Path length routine: calculates the length of a path



38

Operations to be supported (cont.)



� Routing: finds the optimal route between two points (taking into consideration a number of constraints)

� Network distance calculation: calculates the network distance between two points

� Network k-Nearest Neighbor search: searches for the k-NN of a point with respect to network distances


20

39

Operations to be supported (cont.)




� Trajectory refresh: adds a new coming position to an existing

trajectory

� Trajectory projection: estimates the position of an object at a future

time (taking into consideration its history)

� Trajectory-based search: (spatial and/or temporal) window search, NN

search, trajectory similarity search, etc.

40

More on trajectory mgmt operations

� Trajectory refresh:

� Low complexity, straight-forward (?) operation

� Issue: location-update strategy

� Distance-based vs. movement-based vs. time-based update (Bar-Noy et al.

1995)

� Trajectory projection / future location prediction:

� Very interesting, open issue

� Not only consider the current position but the recent history as well !!

� Trajectory search:

� Other operations (window search, NN search) are straightforward

extensions of the respective spatial operations) while others (e.g.

similarity search) are by nature complex

21

41

Summary

42

Summary

� Mobility-aware applications are (or will be very soon) ubiquitous

� However, to be effective ( = high quality) and efficient ( = fast

response) they ask for robust infrastructure that will provide

� trajectory and motion pattern management

� modeling; physical representation; indexing; query processing and

optimization

� trajectory and motion pattern analysis / mining

� co-location patterns; outlier analysis; …

22

43

Questions

44

Reading list

� MLS / LBS infrastructure

� Bar-Noy, A. et al. (1995) Mobile Users: To Update or not to Update. Wireless

Networks, 1(2):175-185.

� Borriello, G. et al. (2005) Delivering Real-World Ubiquitous Location Systems.

Communications of the ACM, 48(3).

� Chen, Y. et al. (2004) LORE: An Infrastructure to Support Location-aware

Services. IBM Journal of Research and Development, 48(5/6):601-616.

� Davies N. et al. (2001) Using and Determining Location in a Context-Sensitive

Tour Guide. IEEE Computer, 34(8):35-41.

� Hjelm, J. (2002) Creating Location Services for the Wireless Web. John Wiley &

Sons.

� Imielinski, T. et al. (1997) Data on Air: Organization and Access. IEEE

Transactions on Knowledge and Data Engineering, 9(3): 353-372.

23

45

Reading list

� MLS / LBS infrastructure (cont.)

� Konig-Ries, B. et al. (2002) Report on the NSF Workshop on Building an

Infrastructure for Mobile and Wireless Systems. SIGMOD Record, 31(2).

� Spinney, J. and M. Mabrouk (2002) Using J2ME to Deliver Spatial Information

to Mobile Clients. Presented at JavaOne 2002. Retrieved from www.esri.com.

� Swedberg, G. (1999) Ericsson’s Mobile Location Solution. Ericsson Review, 4.

� Want, R. et al. (1992) The Active Badge Location System. ACM Transactions on

Information Systems, 10(1):91-102.

� Wong, V.W.-S. et al. (2000) Location Management for Next-generation

Personal Communications Networks. IEEE Network, 14(5): 18 -24.

46

Reading list

� LBS examples and taxonomies

� Amir, A. et al. (2002) Buddy Tracking – Efficient Proximity Detection Among Mobile Friends. IBM Research Report, RJ 10250.

� Frentzos, E. et al. (2007) Towards the Next Generation of Location Based Services. Proceedings of W2GIS Workshop.

� Gratsias, K. et al. (2005) Towards a Taxonomy of Location Based Services.Proceedings of W2GIS Workshop.

� Koshima, H. and J. Hoshen (2000) Personal Locator Services Emerge. IEEE Spectrum, Feb. 2000, pp. 41-48.

� Virrantaus, K. et al. (2001) Developing GIS-Supported Location-Based Services. WISE (2), 66-75

� Wong, V.W.-S., and V.C.M. Leung (2000) Location Management for Next Generation Personal Communication Services. IEEE Network, 14(5):18-24.

24

47

Reading list

� Location prediction for LBS purposes

� Elnekave, S., Last, M., Maimon, O. (2007) Predicting future locations using

clusters' centroids. Proceedings of GIS.

� Gowrisankar, H. and Nittel, S. (2002) Reducing Uncertainty in Location

Prediction Of Moving Objects in Road Networks. Proceedings of GIScience.

� Hassan, K. and Xiong, L. (2003) A Predictive Location Model for Location-Based

Services. Proceedings of GIS.

� Monreale, A., Pinelli, F., Trasarti, R., Giannotti, F. (2009) WhereNext: a location

predictor on trajectory pattern mining. Proceedings of KDD.

� Morzy, M. (2007) Mining Frequent Trajectories of Moving Objects for Location

Prediction. Proceedings of MLDM.

mobility data management - infolab.cs.unipi.grinfolab.cs.unipi.gr/tutorials/handouts/trento... ·...

Documents