UMBC ebiquity
Applied Artificial Intelligence
Characterizing Knowledge Depth in Intelligent Safety Systems
October 10, 1989
Intelligent process control may be viewed as encompassing four major tasks. An intelligent agent must monitor the target system to obtain the values of relevant stale variables in order to detect problems and ascertain the status of the components that may be employed in responding to those problems. An intelligent agent must determine plans for managing the current situation. An intelligent agent must select a response (the “best” one) through a process of plan evaluation. Finally, to carry out the chosen response, the agent must perform plan execution. While monitoring and execution are relatively straightforward operations, plan determination and plan evaluation may be accomplished in a number of ways that vary in their relative depth of reasoning. In this paper, we sketch an analysis of the reasoning underlying plan determination and evaluation tasks for a class of intelligent control systems that attempt to “provide a safety function.” This analysis has two objectives: to illustrate a domain-independent mode of analysis for examining progressively deeper models and to make the analysis available to those interested in building systems that provide safety functions.
Article
Taylor & Francis
3
2-3