Corvus
|
|
The problem of using heterogeneous, distributed, and dynamic computing resources efficiently has potentential benefits for many applications including large networks of sensors and other huge data-gathering systems that require the filtering of usable, actionable knowledge from large streams of data. Our solution is the development of a cyber-infrastructure to work within such a distributed environment that can handle large flows of acquired data using artificial curiosity and other machine intelligence methods that can be “plugged-in” to the distributed architecture dynamically. Unlike other typical HPCC applications, our architecture is targeted for real-time acquisition and processing of large-scale data streams and can dynamically utilize heterogeneous computing resources and sensors to intelligently turn streams of data into usable knowledge.
Gathering and processing huge data sets from large Networks of Sensosr (NOS) has tremendous potential benefits in areas from environment and resource management, emergency response, national health monitoring to defense systems. There are several major challenges in using designing and implementing such systems. How do we effectively use distributed, dynamic, heterogeneous computing resources cooperatively to process and analyze huge data sets. How do we apply intelligent algorithms to collected data sets to extract insights. How can we synthesize conclusions from masses of, often conflicting and certainly overlapping, multi-sensory perceptions.
Our project combines research in grid and distributed computing with machine and collective intelligence research. Corvus is an infrastructure for distributed computing and NOS sensor gathering systems, with built-in plug-in distributed machine intelligence algorithms for analysis. We are interested in the artificial curiosity algorithm, among others, as cognitive models of multi-sensory synthesis.