Analysis of GPS logs

1. Analysis of GPS logs Dagstuhl Wilko Quak

2. Overview • Introduction • GPS characteristics • System architecture • DBMS backend • Automatic rules • Challenges GPS Fun

3. Introduction • GPS-Monitored itinerary trackingof persons • 10 weeks * 100 respondents, each 7 days • Communication via Website GPS Fun

4. This is what we want: GPS Fun

5. This is what we have GPS Fun

6. This is how plan to get it GPS LOG Analyzed trip data Validated analysis Postgis Web server GPS Fun

7. GPS Characteristics • Accuracy 10 meters (approx.) • Battery life 18 hours • 10 000 trackpoints • Works only outdoors • Needs clear view of the sky • Returns innacurate logs • Has a start-up time GPS Fun

8. GPS tracklog characteristics • A tracklog is a list of points: • timestamp • location • elevation • connected GPS Fun

9. In Google Earth GPS Fun

10. GPS Characteristics (this is my life) GPS Fun

11. Flow Control + Architecture Web Server Internet User: Upload tracklog PostGIS DBMS Backend + Mapserver Server: Automatic Analysis User: hand-edit data Server: store data permanently GPS Fun

12. Automatic Analysis Automatic reasoning is needed! GPS Fun

13. Automatic reasoning is not enough. Keep the user in the loop! Hand Edits GPS Fun

14. DBMS Model GPS Fun

15. Automatic analysis • Process trackpoints to fill trip table: • Cleanup data • Split into trips • Derive modality • Good guess better than noting (user is in the loop) GPS Fun

16. Analysis Rules • Simple logic rules: • No Teleport: The endpoint of one trip is the startpoint of the next trip • Rules with external data: • If a track follows a railway line the user is travelling by train • Probabilistic rules: • The speed distribution on a track gives probabilities for: foot, bike, car. GPS Fun

17. Rule implementation I have different preffered languages for every rule: • PL/SQL • Java • XSLT • Probabilistic Language (???) • GIS system Challenge: How do I make an integrated system? GPS Fun

18. Discussion GPS Fun