1. Presenter: Mathias Jahnke Authors: M. Zhang, M. Mustafa, F. Schimandl*, and L. Meng Department of Cartography, TU München *Chair of Traffic Engineering and Control, TU München meng.zhang@bv.tum.de, www.carto-tum.de ICC 2009 A Grid-Based Spatial Index for Matching between Moving Vehicles and Road Network in a Real-Time Environment Santiago, Chile, November 15~22, 2009

2. - 2 - An efficient traffic management requires reliable knowledge that can be immediately derived from the current traffic state. The fact that better knowledge can be obtained from the larger data volume leads to challenges in data processing speeds. The more data being processed, the slower the current situation could be analyzed in high detail. A delay of analysis in turn may cause inaccurate and inconsistent results. Traffic state can be estimated with help of floating car data (FCD) solely or in addition to traditional techniques, e.g. by fusing FCD with local loop detector data or section travel time data. 1. Background However, most of the literatures basically concern about the accuracy of the matching process. There is still a need to make sure the algorithm can do the map-matching efficiently, especially for real-time applications. Part 1

3. - 3 - Real time application for traffic state estimation using FCD Part 1

4. - 4 - Find candidate links - Calculate the distance between a point and a line segment Part 1 (a) (b)

5. - 5 - Usually, the street network is represented by vectors of coordinate pairs corresponding to their start and endpoints, which are ordered by their connectivity. Given a random query point in space, it is very inefficient to find its nearest line segment without using indexing structures. 2. Spatial Index Part 2 A spatial index is typically utilized in large spatial databases to optimize spatial queries, whereas the spatial index reported in this paper serves the purpose of accelerating the matching process between points and line segments. Two special constraints should be satisfied: - it should be easily established and implemented; - it should lead to a maximum speed improvement of the matching process.

6. - 6 - Grid-Based Spatial Index for point data Part 2 dark grey square: base region; shallow grey squares: neighbouring blocks {Affected blocks} = {base region} U {neighbouring blocks}

7. - 7 - The index model for the organization of line segments Part 2 dark grey square: base region; shallow grey squares: neighbouring blocks Problem?

8. - 8 - The index model for the organization of line segments Part 2 dark grey square: base region; shallow grey squares: neighbouring blocks Worst case: l (p1, p2)

9. - 9 - The index model for the organization of line segments Part 2 dark grey square: base region; shallow grey squares: neighbouring blocks

10. - 10 - The index model for the organization of line segments Part 2 dark grey square: base region; shallow grey squares: neighbouring blocks Solution:

11. - 11 - The index model for the organization of line segments Part 2 Data structure of the proposed spatial index for linear data

12. - 12 - The index model for the organization of line segments Part 2 The established spatial index for the line segments l 56, l 78 and l 9,10

13. - 13 - 3. Experiments Part 3 Road network Spatial Index GPS-Points Map Matching & Traffic state estimation in a real time enviorment

14. - 14 - Part 3

15. - 15 - Part 3 Worthwhile to mention is that with the proposed spatial index, an increase of the network size will not significantly slow down the matching speed. X-axis: size of the road network (number of line segments); Y-axis: amount of moving vehicles; Z-axis: matching speed (Vehicles/ second)

16. - 16 - A tricky grid-based spatial index for linear data based on the decomposition principles. Reduced algorithmic complexity: O(n*n)  O(n), which enables the real-time traffic state estimation. The spatial index proved to have considerably improved the computing speed for the point-to-line matching. Besides Graz, Austria, two large cities in Germany and one mega city in China which covers an urban area of about 3700 km2 and involves about 176000 road objects with more than one million line segments. 4. Conclusions Part 4