Interop97 - Noronha

Interoperability Issues in Intelligent Transportation Systems: Testing the Cross Streets Profile

Val Noronha
Vehicle Intelligence Testing & Analysis Laboratory
NCGIA - University of California
Santa Barbara CA 93106-4060, USA
Phone +1.805.893.8992 Fax +1.805.893.8993
noronha@ncgia.ucsb.edu

Intelligent Transportation Systems (ITS) is an emerging suite of technologies that promises to make highways safer and more efficient, by offering motorists tools to interact with other vehicles, and with ground-based information service providers. Some of the potential application areas are emergency assistance, real-time routing based on up-to-date freeway congestion measurements, traveller facility directories (e.g. motels and ATMs) and collision avoidance systems.

Vehicle and incident location are central data items in ITS. In a world of competing vendors of street network data - and inevitable discrepancies and errors in position, street naming, addressing and classification - the success of ITS hinges on the ability to communicate a location message unambiguously across dissimilar map bases. Currently this is not achievable within loosely specified industry-acceptable error tolerances. To be successful and acceptable to commercial navigation system vendors, a solution to this problem would need to minimize impact on vendor data bases and practices. For example, vendors should not be required to re-survey their national databases to more stringent standards; similarly, conflation to one vendor's database would imply that the other vendors' data were inferior.

Potential solutions to the location referencing and messaging problem are being proposed by Standards Development Organizations (SDOs) at the national and international levels. These solutions need to be field tested in simulated and real application conditions. Two current proposals under consideration are the Location Reference Messaging Specification (LRMS) and the ITS Datum (ITSD) from Oak Ridge National Laboratories.

The Vehicle Intelligence Testing & Analysis Laboratory (VITAL) at UC Santa Barbara is a testbed for interoperability issues related to ITS. Funded by the state and federal governments, we are pursuing a research agenda driven partly by the industrial needs of SDOs, and partly by academic questions. The initial task of the lab was to set up a testing infrastructure, consisting of a client system in a vehicle and a ground-based server. The client has a moving map display, differential GPS vehicle locator, dynamic routing and real-time wireless message exchange with a server. One of the first research tasks of the lab is to test the LRMS Cross-Streets Profile; this is the focus of this paper.

There are several ways to communicate a location on a network, e.g. coordinates, street addresses, grid references, routes-and-offsets. Each method has its advantages and drawbacks. Unfortunately it is not practical to use multiple methods for a single message, because message length is a constraint in wireless communications systems.

The Cross-Streets Profile is one of the message specifications in the LRMS, and currently the one most favoured by vendors. It specifies a location in the form of an offset along a street segment, where the segment is defined by a street name and bounding cross-streets. For example, "2473 metres down Hollister Avenue between Patterson Road and Walnut Street." Consider the potential for lack of interoperability when sending a message in this form between two data bases:

(a) Spelling errors and differences in vendor practices: Hollister is spelt Holister; Avenue is abbreviated Av in one database, Ave in another.
(b) Gross differences in street names: El Camino Real in database A, Highway 101 in database B.
(c) Topological and inclusion errors. A street that appears in database A does not appear in database B. Therefore an intersection referenced in the source database is not identifiable in the target.
(d) Positional differences at intersection nodes and shape points

In addition, there are problems inherent in the method itself, in particular the case where a street intersects another several times. Some of the above problems can be corrected using fuzzy search algorithms and reinforcement of clues. To test the effectiveness of this profile, we first define application scenarios, testing criteria, metrics and tolerances. The problem is broken down into components, e.g. (a) the probability of finding an accurate lexical match using simple and complex search techniques; (b) the mean positional error when relaying an offset from one database to another.

This paper discusses the ITS interoperability problem in general, and presents experimental design and results of work to date using popular commercial databases for the County of Santa Barbara, California.