Enabling Query Processing on Spatial Networks

Jagan Sankaranarayanan, Houman Alborzi, and Hanan Samet∗

Center for Automation Research, Institute for Advanced Computer StudiesDepartment of Computer Science

University of Maryland, College Park, MD 20742{jagan,houman,hjs}@cs.umd.edu

Abstract

A system that enables real time query processing on largespatial networks is demonstrated. The system provides func-tionality for processing a wide range of spatial queries suchas nearest neighbor searches and spatial joins on spatial net-works of sufficiently large sizes.

1 SILC Framework

Recently, we have witnessed the emergence of severalweb services1that enable users to pose simple spatial queries.However, the query result may not be of much practical useas the distances provided do not take into consideration theconstraints imposed by the underlying road network. Queryprocessing on spatial networks is considered computation-ally expensive and processing such queries in a realistic timeframe remains a challenge.

We recently proposed the SILC framework [2], a datastructure that enables fast and efficient retrieval of the short-est paths and distances between objects on a spatial network.We demonstrate a system (Figure 1) that augments a generalpurpose spatial database with the SILC framework, therebyproviding the database with the necessary functionalities toperform spatial queries on spatial networks. The followingare the unique contributions of our system.

1. To the best of our knowledge, we believe that our sys-tem is the first to allow efficient processing of most spa-tial queries on spatial networks using a general purposespatial database engine.

2. Our system is the first to decouple the spatial objectsfrom the underlying spatial network. Thereby, SILCcan be integrated into any spatial database with anychoice of spatial indices.

∗The support of the National Science Foundation under Grant EIA-00-91474 and Microsoft Research is gratefully acknowledged.

1Such as Mappoint, Mapquest, Yahoo!, and Google Maps

3. The system performs real-time processing of both ap-proximate and exact spatial queries on spatial networks,with little or no modifications to existing spatial queryprocessing techniques, thus reusing existing knowledgeand code base rather than developing new techniques tohandle queries on spatial networks.

4. The system design and the concepts we demonstrate aregeneral enough to be applied to any spatial databasesystem.

Spatial queryresponse encoding

SILC

C. Spatial relation and spatial index

D. S

ILC

mo

du

le

Embedding

A. S

and

Sp

atia

l Bro

wse

r

Spatial query

B. SANDspatial database

Path−

queriesdistance

spatial non−spatial

Figure 1. Block diagram representation of thedemonstrated system.

We have augmented SILC with the SAND [1] spatialdatabase system which is being developed at the Universityof Maryland. The demonstrated system performs shortestpath and distance queries, range and window queries, in-cremental nearest neighbor searches, incremental and non-incremental distance joins and semi-joins in real time on asuitably large road network dataset.

References

[1] H. Samet, H. Alborzi, F. Brabec, C. Esperanca, G. R. Hjalta-son, F. Morgan, and E. Tanin. Use of the SAND spatial browserfor digital government applications. Commun. ACM, 46(1):63–66, Jan. 2003.

[2] J. Sankaranarayanan, H. Alborzi, and H. Samet. Effi cientquery processing on spatial networks. In ACM GIS, pages 200–209, Bremen, Germany, Nov. 2005.

Proceedings of the 22nd International Conference on Data Engineering (ICDE’06) 8-7695-2570-9/06 $20.00 © 2006 IEEE