Services on Sensor Networks
From SSRG Annoki
People Involved
Pawel Gburzynski Ioanis Nikolaidis Eleni Stroulia
The real potential of sensor networks is that, while each node is meager in terms of capabilities, the collective capabilities of a constellation of such devices are extremely valuable, indeed even indispensable, for example when sensors are used for monitor and control of critical infrastructure (pipelines, utilities, highways, etc.). An immediate observation that can be made is that for a device costing a few dollars to be "in charge of" information of potential critical importance, the added value cannot be on the platform as such but on the control and services it can provide.
Sensor networks are the "poor relatives" of data communication networks. Despite intense academic interest on sensor networks and the properties of protocols and/or data collection and distribution services of such networks, certain elementary problems remain when it comes to development of applications. In particular, it is not uncommon for sensor applications to be one-of implementations, with little reuse capabilities. At best, a core functionality, which usually is termed the "sensor OS" can be retained across applications (and ported across platforms), but each application (having its own necessities in terms of routing of data, data processing, latency requirements etc.) is left to ad-hoc design principles. Hence, once a sensor network is deployed, the re-tasking and re-use becomes prohibitively expensive, or just impossible without building from scratch. In light of such concerns, the danger is that investment in sensors will be ineffective and, as a result, integration with the rest of the cyberinfrastructure will be delayed.
Our objective is to take the sensors and elevate them to first class entities for tomorrow's cyberinfrastructure. What we are advocating is to avoid "inflating" sensor platforms with application primitives and abstractions that are influenced by the mainstream operating systems community. The pragmatics of the situation are that, ideally, sensors ought to be cheap (to the point that they can be disposable if need be) and as a result the tiniest platforms are usually used (e.g., a few kilobytes of RAM) which is in agreement also with the need to retain operation for multiple years if possible using a single battery cell.
To maximize code reuse potential, we are working on aspects of composition of middleware components and corresponding sensor-resident components to address issues of quality of service (as it relates to reliability and robustness) while permitting adaptable, configurable and re-deployable software components. In a sense the research addresses how middleware requirements should be "compiled" into per-node code and flexible protocols and the corresponding gateway-resident or server-resident stubs. Our prior experience with one-of application deployment as well as with the development of a fairly portable sensor OS provide us the rare mix of experience that is necessary to undertake the task.
As part of that activity, we have developed SensorGIS a web-based architecture (relying on REST web services) that visualizes sensor network data using a geographic information system (GIS). Unlike most of the sensor network visualization tools, SensorGIS is generic and thus can be applied to various types of sensor networks.
A video demonstration SensorGIS can be fount at SensorGIS-video.