New Directions in Modeling and Control of Progressive Processing

Abdel-illah Mouaddib, Shlomo Zilberstein, and Victor Danilchenko. New Directions in Modeling and Control of Progressive Processing. Proceedings of the ECAI Workshop on Monitoring and Control of Real-Time Intelligent Systems, Brighton, UK, 1998.

Abstract

Progressive processing is an approach to resource-bounded execution of a set of tasks under time pressure. It allows a system to limit the computation time allocated to each task by executing a subset of its components and by producing a sub-optimal result. Progressive processing is a useful model for a variety of real-time tasks such as diagnosis, planning, and intelligent information gathering. This paper describes recent results and new directions aimed at generalizing the applicability of progressive processing by addressing the issues of high duration uncertainty and quality uncertainty associated with each computational unit. We also examine new ways to model inter-task quality dependency and a richer topology of task structures.

Bibtex entry:

@inproceedings{MZDecai98ws,
  author	= {Abdel-illah Mouaddib and Shlomo Zilberstein and Victor Danilchenko},
  title		= {New Directions in Modeling and Control of Progressive Processing},
  booktitle     = {Proceedings of the {ECAI} Workshop on Monitoring and Control of
                   Real-Time Intelligent Systems},
  year		= {1998},
  pages		= {},
  address       = {Brighton, UK},
  url		= {http://rbr.cs.umass.edu/shlomo/papers/MZDecai98ws.html}
}

shlomo@cs.umass.edu