|
|
Motivation
|
|
The goal of this project is to facilitate and ease the development of sophisticated applications using wireless networked sensors and actuators. Our vision is that an application designer need not be aware of the network topology or hierarchy, nor of the dynamic connectivity changes, nor even of the sensor details or networking protocols. Nonetheless, the specification of an application will be automatically transformed into software running in several parts of a system hierarchy; the details of this software distribution and the management of resources need not be visible. At the same time, the software can be highly tailored to the limited resources of a sensor network, and it can also be highly resilient to the slings and arrows of environmental interference. On the other hand, our network will also have complete transparency, so that network and security control operations can make intelligent decisions about reconfiguration or even reprogramming of the networking infrastructure. These security and control functions will be more aware of topology and resources, but their API's will be as easily used as the application API's. The key to simplifying the process of programming and managing sensor networks is the middleware that supplies the content networking abstraction. Our approach to sensor based pervasive environments is to create a virtual information space, and to manage the underlying network to achieve various optimization goals within that space. An essential component of this is an underlying ontology and semantic representation of information. For example, we would like to have a uniform understanding of “temperature” and “temperature measurement units”. |
|
Summarily, supporting new emerging applications requires a middleware infrastructure that: (a) is scalable and self-managing, (b) is based on content rather than names and/or addresses, (c) supports asynchronous and decoupled interactions rather than forcing synchronizations, and (d) provides some interaction guarantees. |
