Techniques for Efficient Road-Network-Based Tracking of Moving Objects

1. Techniques for Efficient Road-Network-Based Tracking of Moving Objects Speaker : Jia-Hui Huang Date : 2006/10/23

2. Reference • [1] Alminas Civilis, Christian S. Jensen, Senior and stardas Pakalnis, “Techniques for Efficient Road-Network-Based Tracking of Moving Objects,” IEEE Transations on knowledge and data engineering. VOL.17,NO.5, MAY 2005 • [2] A. CCivilis, C.S. Jensen, J. Nenortaite, and S. Pakalnis, “Efficient Tracking of Moving Objects with Precision Guarantees,” Proc. Int’l Conf. Mobile and Ubiquitous Systems: Networking and Services, pp. 164-173, 2004, extended version available as DB-TR-5, Dept. of Computer Science, Aalborg Univ., Denmark, http:// www.cs.aau.dk/DBTR/DBPublications/DBTR-5.pdf.

3. Outline • Introduction • Existing Tracking Approaches • Road Network modification • Update reduction using routes • Update reduction using acceleration profile • Summary

4. Introduction (1/4) • Location-based services ( LBSs ) • Tracking geographical positions of moving object • Large volumes of location updates • How to reduce update within predefined accuracy ? • Use road network for predicting moving object’s future positions.

5. Introduction (2/4) • Definition • Road network composed of a set of segments. • A segment is defined as a sequence of coordinates. • Road network is partitioned into streets. • Each segment belongs to precisely one street. • Each segment identifies its street by referring to a street code.

6. Introduction (3/4) • Tracking scenario • GPS • Tracking approach • Threshold • Challenge • Predict future position • Minimize update

7. Introduction (4/4) • Scenario from UML view

8. Outline • Introduction • Existing Tracking Approaches • Road Network modification • Update reduction using routes • Update reduction using acceleration profile • Summary

9. Existing Tracking Approaches • Three existing tracking approaches • Point-based tracking • Represents object’s future position as the most recently reported position. • Update when position change exceed threshold. • Vector-based tracking • Future position compute by speed and heading of moving object. (velocity vector) • Zero-vector case of vector-based. (point) • Segment-based tracking

10. Existing Tracking Approaches • (a) update when exceed threshold • (b) update only when direction change (a) Point-based (b) Vector-based

11. Existing Tracking Approaches • Segment-based tracking • Utilize knowledge of the road network. • Digital representation of the road network. • Road network composed of a set of segments. • Segment is in-between a pair of consecutive intersections or an intersection and a dead end. • Map matching positions an object on a road network segment.

12. Existing Tracking Approaches • Future positions of a client are given by a movement at constant speed along the identified segment. • Update when deviates exceed threshold • Number of update correlated with the number of changes of segments. (c) Segment-based

13. Existing Tracking Approaches • Update time increases as the accuracy increases

14. Existing Tracking Approaches • Why improve segment-based approach ? • Average length of the road segment is short. • It may be a relative straight road is represented by several segments. • Road-related information can use.

15. Outline • Introduction • Existing Tracking Approaches • Road Network modification • Update reduction using routes • Update reduction using acceleration profile • Summary

16. Road Network modification • Number of update correlated with the number of changes of segments. • Connect the road segment that moving objects would change segment as few as possible. • Two types of modification • Street Code-based approach • Direction-based approach

17. Road Network modification • Two types of approach based onGeneral segment connection algorithm ( GSC ) • General segment connection algorithm ( GSC ) • Iterate through all segments according some specified ordering. • At each iteration, algorithm tries to extend the topmost or current segment.

18. Road Network modification • Some parameter • rn, (road network), a set of road network • stc, a Boolean valued variable that controls connection with different street codes. • rnPrioriization, canPrioritization sort order specifications • Sort order according segment’s property • Length, speed limit, ….

19. Road Network modification • Street code-based approach • Give priority to connecting with the same street code. • In cases where several candidates with the same street code, priority given to the shortest one. • Reduces the probability that unconnected segment will be short. • Among candidates segment, the longest are preferred.

20. Road Network modification • Directional-based approach • Moving objects are expected to be moving as directly as possible toward their destinations. • Extending segment with the same direction as much as possible. • Preference is given to segment with direction at the connection point that has a small angle with direction of the segment to be extended.

21. Road Network modification Property angleAvg denotes the average of the smallest angle values possible for both ends of the segment • Eg. A segment can be extended with three segment to one side with angles of 34, 22 and 90 degrees, then angleAvg = ( 22 +180) /2

22. Road Network modification • Comparison • Increase the performance of the segment-based policy and outperform the vector-based policy

23. Road Network modification • Comparison of Suburban versus city • Unmodified road network curve suggests that the majority of updates happens due to segment changes, not due to speed variation.

24. Outline • Introduction • Existing Tracking Approaches • Road Network modification • Update reduction using routes • Update reduction using acceleration profile • Summary

25. Update reduction using routes • Constant-speed optimal policy • Assume moving object travels on only one segment, and no updates occur due to segment change. • Assume constant moving speed. • Gives the lower bound of the number of update needed by segment-based policy.

26. Update reduction using routes • Use of Routes (traveler’s behavior) • Assume the travelers frequently use the same routes to their destinations. • Take advantage of knowledge of the routes used by a moving object. • When using the segment-based policy with routes, we know the future positions of an object. • If an object deviates from its route, this is treated simply as a segment change. • Like constant-speed optimal policy.

27. Update reduction using routes • Experiment

28. Outline • Introduction • Existing Tracking Approaches • Road Network modification • Update reduction using routes • Update reduction using acceleration profile • Summary

29. Update reduction using acceleration profile • Segment-based policy assumes that objects move at constant speed. • Model speed as a stair function. • More accurate modeling speed can help better predict the future position of moving object. • e.g. using averages of the speeds during past traversals of the route.

30. Update reduction using acceleration profile • Acceleration profile capture the average speed variation. • Dividing the average speed variation along the route into intervals. • Average acceleration is calculated for each interval. define profile apf as a sequence of n+1 measures and n accelerations ( , ……, , , ) acceleration is valid in interval [ , ) e.g. assume an object moves with speed and current location along the route is , belongs to the interval [ , ) then predicted position and speed of the object at time t is

31. Update reduction using acceleration profile • Experiment

32. Outline • Introduction • Existing Tracking Approaches • Road Network modification • Update reduction using routes • Update reduction using acceleration profile • Summary

33. summary • Performance of basic segment-based tracking is sensitive • Segmentation representation • Speed variation • Reduce number of update technique • Road network modification • Use of routes • Use of acceleration profiles

34. General segment connection algorithm

35. Speed model

36. Speed pattern