• Introduction
Wireless sensor networks have received significant attention due to a number of potential applications based on sensing data observed from the physical world. Instead of waiting for periodic reports from dumb sensor nodes as proposed in conventional researches, users prefer querying the networks in a real-time fashion. In general, people are likely to search things related to their daily lives in the real world ranging from straight-forward queries to complex ones. Examples of such queries are weather forecast, a nearby Thai restaurant, route navigation, and so on. In order to expedite and realize the usage of sensor networks in ubiquitous society, sensor networks are expected to response to such real-world queries. However, a sensor network cannot have all kinds of information related to the users’ queries. It is impractical to install all kinds of sensors in the field. Even if the relevant kinds of information exist in the sensor network or the queries are limited to some specific types, the sensor network needs to be deployed in a wide area (city or country level) so as to answer the queries correctly. Because deploying a variety of sensors throughout a city or country is impractical and inefficient, a collaboration of independent sensor networks is a possible and promising solution.


• OSOITE Architecture
Overlay-network Search Oriented for Information about Town Events (OSOITE) provides an architecture for realizing real-world search by exploiting a collaboration of heterogeneous sensor networks. A sensor network is composed of simple sensor nodes and at least one Database Sensor Network (DBSN) peer that collects and manages sensing data in its own network. Each sensor network is an autonomous system and independent of each others, i.e., it may be deployed by different administrators. The sensor network connects to other sensor networks by constructing an overlay network amongst DBSN peers of each network. A query can be injected by any node and will be forwarded to relevant sensor networks using the constructed overlay network.
In summary, the OSOITE architecture provides the following functions.


・ Distributed storage.

Because distributed data storage is likely to be an efficient solution in comparison with a centric storage in which bottleneck always occurs, the OSOITE employs the distributed approach. Sensing data are stored in sensor nodes locally, or collected at DBSN peers of each sensor network.

・ Real-world search.

The OSOITE supports real-world search specified by users. For example, users may ask for a breezy (or calm, safe) path starting from a distinct point to a distinct train station. Main inputs for the OSOITE architecture are composed of three elements: location, time, and event. Only one element of a combination of multiple elements can be specified by the users.

・ Multi-resolution queries.

Each query asks for different resolutions of answer. Each DBSN peer also maintains different resolutions of sensing data which are managed in hierarchical overlay network for ease of processing. The OSOITE architecture forwards queries to such hierarchical structure, i.e., the queries are forwarded to the coarsest first, and they are forwarded to the node whose resolution satisfies the queries.

・ Interpolation.

A collaboration of multiple sensor networks may not cover an entire city or country, and users may ask for information that is not available. Even if information is available, resolution may not match that specified by users, i.e., users require finer resolution than that available in the network. In this case, interpolation technique is employed to calculate the possible answer.

・ Web API and query conversion.

Users submit human-readable queries through an intelligent web API. The web API converts the queries into SQL language which is convenient for query processing. The DBSN peers produce answers in SQL format, which is converted into human-readable language later.

・ Ease of deployment.

Each sensor network is completely independent of each others, but they work collaboratively to process real-world search. Each administrator can deploy its own network with its own policy. After installing the OSOITE software kit in the DBSN peers, real-world search are ready to use.


• Conclusions
Real-world search is a promising application for sensor networks. The OSOITE is an architecture for realizing such real-world search amongst heterogeneous sensor networks. Those sensor networks construct overlay network to work collaboratively. Sensing data are distributed through multiple networks with different levels of resolutions. The query is forwarded to the node whose resolution matches that specified in the query. If sensing data are not available or resolution is not fine enough, the interpolation technique is employed. The purpose of the OSOITE project is to provide a software kit to realize the above-mentioned functions.

International Symposium

25th Feb, 2010

Workshop

11th March, 2009

International Symposium

16th June, 2008