A new architecture for cyber maps: An internet-oriented next-generation spatial information service System

A new architecture for cyber maps: An internet-oriented next-generation spatial information service System

Shigeru Shimada and Masaaki Tanizaki
Hitachi, Ltd,. Central Research Laboratory 1-280, Higashi-koigakubo
Kokubunji-city, Tokyo 185 Japan
EMail: {shimada,tanizaki] @crl.hitachi.co.jp

Abstract
To handle new information service environments, such as mobile or interoperable information infrastructures, we propose an architecture for a next-generation spatial information service system (NSIS). This provides vector-based spatial information using VRML, which enables vector-based interoperable communications, unified management of spatial and attribute databases by an ORDBMS, which brings effective management of multimedia databases, and an active retrieval mechanism by connecting the ORDBMS with WWW servers to match the distribution of spatial information with the capabilities of web clients. A prototype implementation, called Cyber Map, demonstrates promising performance at retrieving spatial information service via the Intemet.

Introduction
Communication infrastructures have made rapid progress recently, as illustrated by digital communication networks using ATM, and wireless communication using PHS or satellites. And using these communication infrastructures, Intemet-oriented communication techniques, such as Java and VRML (Virtual Reality Modeling Language), are also being developed very quickly. Through the integration of these sophisticated techniques, new information service systems that use mobile or interoperable communications are reaching the stage of practical use.

In Japan, utility companies, such as electricity, telephone, and gas companies, who are advanced users of geographic information systems (GIS), are strongly demanding a next-generation system. This should be able to handle spatial information, which means not only two-dimensional geographical data represented by maps, but also height information such as the number of floors in a building or the depth of cables in a facility. Such a next-generation spatial information service system (NSIS for short) will also be essential for the maintenance of city information infrastructures, which became an important issue after the great Hanshin earthquake. Many cyberspace plans have been announced by local governments in line with the new information highway policy of Japan. This policy needs a new information processing style in which all information is processed under intemetiintranet communication environments.

For this purpose, we are investigating a new system architecture for a system providing spatial information services suitable for business needs rather than a simple map browser. As the first step, in this study we propose a mobile and interoperable system architecture for an NSIS under intemet and intranet communication environments.

Fig. 1. System concept of NSIS: Next Generation Spatial Information Service System
New Functions of NSIS
Our concept of NSIS is shown in Fig. 1. An intemet-oriented information service environment will enable ~trieval from multiple databases at different WWW sites. For example, it will be possible to display in real time a mixture of results independently retrieved from a database of geographical maps (at site B) and from a database of underground construction (at site C). Trying to meet these demand by conventional methods would involve heavy data conversion, making the service impractical. Moreover, if wireless communication by PHS or satellite is supported, then retrieval from mobile communication terminals will also be possible. Web terminal such as Personal Digital Assistants (PDAs) will be able to retrieve data from geographical map/drawing databases via WWW servers. This will make possible a mobile information service system, which will be very effective for routine checking during normal operations or for recovery from emergencies.

Problems in provinding these new functions
Many kinds of intemet-oriented information service systems have already been proposed and they have passed through three generations of development, as shown in Table 1.[Kub 95][Mot 95][Sjn 94] For example, UCSB supports ADL (Alexandria Digital Library)[Alx 94], which can provide worldwide geographical maps on demand, and Fujitsu supports “Wild Bird” [Wld 96], which can provide mobile users in Japan with residential maps of urban areas. But almost all of these information services have become stuck at the fmt or second generation, because of the many problems, which are described below.

Problems of data type:
If spatial information is supplied in bit-map format, it is impossible to support hyper retrieval functions. Therefore, in order to support interoperable retrieval from information supplied from

Table 1 Progress in information service methods under internet communication environment

other WWW servers, information should be handled in standard formats like HTML or GIF. This Cm be summarized as:

Problem 1) How do we make a standard vector-based system under an intemet communication environment?
Problems of spatial information management method:
One characteristic of spatial information is that records are variable in length and have adjacency relationships with each other. Therefore, an RDBMS that supports a database server in a second-generation system cannot manage this kind of spatial data effectively.

Problems 2) How can we provide an effective unified management method for spatial databases and attribute databases?
What are the problems in making mobile retrieval practical. There are already many kinds of navigation systems, which are practical applications of a geographical information service. Almost all of these systems use off-line system composition, overlaying locations obtained from Global Positioning Satellites onto a map retrieved from a CD-ROM. Therefore, this type of information service is not suitable for various office-work systems that demand online database retrieval, so there are many problems as follows.

Problems for information service style:
The quantity of information stored in a map database is very large, and wireless communication is not fast enough to handle it. Moreover, if PDAs are used as web terminals, their displays are not big enough and their graphics performance is very poor. They will not be able to handle the heavy processing or the resolution that ordinwy web terminals can. This can be summarized as: Problem 3) How can we provide the most effective information service to match the capabilities of each terminal?

Problems for nomadic Retrieval:
One drawback of wireless communication is that connections are frequently broken when a user moves, depending on the level of noise or location of repeaters. So when a user accesses the service, he/she should be offered the choice of canceling past database transactions or continuing suspended ones. This can be summarized as:
Problem 4) How can we achieve nomadic database retrieval?

New Architecture Proposals
Approach for Problem Solving
In this section, we propose anew system architecture for NSIS to solve these problems. We take the following approaches.

A. Spatial information service by VRML -> corresponding to problem 1)
B. Unified database management by ORDBMS -> corresponding to problem 2)
C. Dynamic information service by active retrieval mechanism
-> corresponding to problems 3) and 4)

The architecture is shown in Fig. 2 and details are explained in the following sections.

Spatlal Information Service by VRML
We use VRML for the spatial information service format because it can support standard vector-based communication in an internet environment. When VRML was originally standardized, it was defined as a language for describing 3-dimensional objects.[Pes 95] So some people might argue that VRML is not suitable for describing 2-dimensional maps. However, we believe that VRML is also suitable for representing 2-dimensional data because many maps and drawings

Fig.2 New System Architectwe for Spatial Information Service
also contain height information, such as the number of floors in a building on a residential map or the depth of cables in a facility drawing.

Unified Management of Databases by ORDBMS
As already mentioned, we propose a unified management method which manages spatial data and attribute data in a uniform manner by an Object Relational Database Management System (ORDBMS). In this case, in order to provide seamless treatment between ordinary clientlserver systems and internet-oriented ones, spatial data is stored in binary format, rather than in text format by VRML. And we use dynamic encoding using an active retrieval mechanisu which can encode text-based VRML from binary spatial data retrieved from the ORDBMS. We also assume that this ORDBMS supports SQL3 (Structured Query Language version 3, which is scheduled to be standardized by ISO in 1998), and we treat vector data as a set of coordinates represented in list form in order to achieve high-speed access by retaining adjacency relationships. And since high-speed coordinate operations are required for range searching etc, we use spatial indexing mechanisms such as R-Tree and K-D Tree.

Active Retrieval Mechanism
The new information service should be able to automatically change formats depending on the capabilities of the client web terminals even when they are using the same retrieval conditions. For example, it should send a summarized rough map to a PDA with a low-resolution display and a detailed map to a workstation with a high-resolution display. We propose an active retrieval mechanism that can convert the retrieval results into the format most suitable for each web terminal. We achieve this by taking a third-generation information service system and adding functions for identifying terminal capabilities and for media processing to the active retrieval mechanism as shown in Fig. 2. An overview of this active retrieval mechanism is given below.

[Step#l] Terminal capabilities and retrieval conditions are sent from the web terminal to the database server.
[Step#2] The terminal capabilities are set as the terminal OS’s environmental variable and transferred to the WWW server by the FORM function of the WWW browser.
[Step#3] In media processing functions such as image conversion function, VRML generation function and summarize function, the most suitable function for this terminal is determined by the terminal capability evaluation part.
[Step#4] The requesting web terminal is sent the results of database retrieval after they have been modified by the selected media processing function.

This active retrieval mechanism can be adapted to meet the demand for wireless nomadic database retrieval by adding process management of database transactions and by buffering the intermediate results,

Prototype implementation and performance evaluation
We implemented a prototype (nicknamed Cyber Map) of our proposed new architecture for NSIS and evaluated its performance.

Overview of Cyber Map Architecture
The system architecture of Cyber Map is based on the third-generation information service system architecture, as shown in Table 1. Basically, it consists of a method of connecting WWW servers with the ORDBMS. For this ORDBMS, we selected Illustra (which was purchased by Informix Softswm, Inc. in 1996) which progressively adopts SQL3 as the database retrieval language, and provides a “data blade” that can easily extend retrieval functions. Using Illustr~ we achieved unified management of spatial and attribute databases. And using the web data blade, we made an active retrieval mechanism that can supply information to each web terminal in the most suitable format by linking the WWW server and Illustra and by locating the media processing function in the data blade.

Details of Active Retrieval Mechanism
After the terminal capabilities (TID), which express display resolution and graphics processing speed of each web terminal together with the retrieval conditions, have been transferred to the WWW server, the retrieval mechanism provides information by applying the most suitable media processing, In particular, for high-resolution terminals, we created a function that provides a 3- dimensional virtual space in VRML format using 2-dimensional geographical map together with the number of floors, which substitutes for height information. Details of the implementation of the active retrieval mechanism, are given below.

(1) Web interface function (Web data blade)
The web data blade is equipped with the dynamic HTML generation function “WebExplodeo”. In Cyber Map, we utilize this function to dynamically generate and execute the spatial retrieval structural query language as “SR-SQL”. The process up to getting the SR-SQL suitable for the terminal’s capabilities is explained below.

[step#l] When it gets a ~trieval function from a web terminal, Web Explode gets a template for generating SR-SQL suitable for the retrieval type.
[step##l] After getting this template, Web Explode generates SR-SQL based on the transfer function obtained from the terminal capability table and the retrieval parameters supported from the web terminal. If it receives a retrieval request from a web terminal with a display resolution of 1152x864, for example, then VRML transfer will be selected.
[step#3] Parameters of the retrieval form, which is assigned from the web terminal, are inserted into variables of a where phrase and they generate SR-SQL.

(2) Spatial retrieval fimctions
The two-dimensional data blade supports spatial retrieval functions based on R-Tree spatial indexing theory[Gut 84]. In Cyber Map, spatial retrieval functions such as retrievals in a circular area around a specified spatial location (circular retrieval) are described in a template position in a table storing SR-SQL, and they are used by substituting variables in the SR-SQL production part of the web data blade. For example, in order to execute a circular retrieval centered on a position (25, 50) in the Shinjuku area and within a 100-meter radius, spatial retrieval is done by generating the following SR-SQL.

select transfer function(tuples)
from figure table of Shinjukuarea
where Intersectsln (Circ (25, 50, 100), tuples);
(3) VRML generating functions
There is a mechanism for generating a virtual 3-dimensional space from a 2-dimensional space plus height attributes. As already mentioned, we took a method that calls a function by SR-SQL which can generate one object from one tuple stored in a 2dimensional figm table, in order to generate VRML. For this VRML generating function, we produced “vrml-funco” which can generate a 3-dimensional object from Poly & Path typed datA which describes closed polygon line figures and the number of floors in a building. This function is described in SR-SQL selected from the terminal capability table, and is executed after spatial retrieval.

Performance Evaluation
To evaluate the performance of Cyber Map, we used a residential map CD-ROM (Zmap-Town II ’95, Tarna district Vol. 1.1) published by Zenrin Co., Ltd.. We produced a function that could generate a virtual 3-dimensional space as VRML using this database. An example of the display

Fig. 3 Display exmaple of generated 3-dimensionalvirtual space by the active retrieval mechanism

Fig 4 Relation between generated VRML capacity and processingtime
of this function is shown in Fig. 3. The left window is an example of a GIF-type image map (Kokubunji-city area 600x600 m) on NetSpace Navigator. When the demanded position for retrieval is specified on this image map by setting the radius at the bottom of this window, the system mt.rieves buildings and roads within the circular area defined by that radius and displays them as a 3-dimensional virtual space. The right window in Fig. 3 shows a Web Space (VRML browser from Silicon Graphics, Inc.) display of VRML generated when the radius was set to 100 m. The response time measured from the demand for circular retrieval to the display of the virtual 3-dimensional space using an Indigo2 (Silicon Graphics Inc.’s graphics workstation) is shown in Fig. 4. The average response time for circular retrieval within a 150-m radius was 10 seconds, indicating that Cyber Map has good enough performance for practical use.

Conclusion
The new architecture we propose provides unified management of spatial data and attribute data and an active retrieval mechanism for an internet-oriented next-generation spatial information service system. It offers a clear approach to making mobile and interoperable information service systems. A prototype implementation, called Cyber Map, demonstrated good enough performance for practical use. The interactive link between the web terminal and the supplied spatial information remains as a topic of future study.

References

[Alx 94] http:hlexandria.sdc.ucsb.edu
[Gut 84] GuttmanA.: (1984), “A Dynamic Index Structurefor SpatialSearching”,Pm. of the 1984SIGMODInt. Conf. on Managemmt of Data.47-58
[Kub95] Kubota, et. al.: (1995), ‘Test of connecting between database and intemet applications”, Technical report of ieice of Japan, VO1.DE95-42,pp.33-40
[Mot 95] Motoda, et. al.: (1995), “Inspection of connecting mechanism of WWW and database services”, Technical report of EWE of Japan, Vol. KBSE 95, pp.47-54
[Pes95] Pesce M.: (1995), “Browsing & Building Cyberspace”, New Riders [Sjn 94] Sjoin M.: (1994), “A WWW Front End to an 00DBMS”, Proc. 2nd International WWWConf. 1994 [Wld 96] http/www.wildbird.or.jp