UMBC ebiquity

Agent Oriented Approaches to a Ubiquitous Grid

Status: Past project

Project Description:
A three year project funded by NSF (award number 0203958) directed by PI Anupam Joshi and CO-PIs Tim Finin, Hillol Kargupta and Elias Houstis and will be carried out 2002-2004 with $480,726 in funding from the NSF CISE Next Generation Software Program (NGS).

At the level of computing and networking hardware, we will see dramatic changes in the next few years. Computing will become pervasive – a large number of devices (e.g. phones, PDAs, household appliances) will become computationally enabled, and micro/nano sensors (the so called smart dust) will be widely embedded in most engineered artifacts. All of these devices will be (wirelessly) networked. More speci.cally, we will assume the emergence of (i) palmtop and wearable/embeddable computers, (ii) Bluetooth like systems which will provide short range, moderate bandwidth connections at extremely low costs, and (iii) widely deployed, easily accessible wireless LANs and satellite WANs. We assume that these will be a mixture of traditional low bandwidth systems and the next generation high speed ones. These developments will lead to wireless networks that will scale all the way from ad hoc body area networks to satellite WANs, and link together supercomputers, “palmstations” and embedded sensors & controllers.

Given this scenario, our proposed research will seek to extend the computational grid by making it ubiquitous and pervasive. In particular, we will develop agent based runtime systems where each component is autonomous, articulate, social and adaptive. Such a system will seamlessly partition computation across elements of the grid ranging from palmtops to supercomputers. Issues that we will investigate include (i) Component/ Service Discovery, (ii) Dynamic Composition of components, and (iii) computation partition across highly asymmetric and heterogeneous systems.

In order to demonstrate the e.cacy of our approach, we will develop applications relating to ubiquitous data mining. Ubiquitous data mining deals with large number of streams of data that have a very high data rate and are typically non-stored. These need to be analysed/mined on the .y to extract relevent information. Often such data come from wirelessly connected sources which have neither the computational resources to analyse them completely, nor enough bandwidth to transfer all the data to a central HPC site for analysis. Clearly, such systems will benefit from the ubiquitous grid. Ubiquitous data mining provides the framework to solve important problems in fields such as healthcare, intelligence, manufacturing and IVHS.

Start Date: October 2002

End Date: October 2005


There are 0 associated publications:  Hide the list...


There are 0 associated resources:  Hide the list...


Research Areas:
 Mobile Computing
 Multi-Agent Systems