Architectural Foundations for Real-Time Performance in Intelligent Agents

Reference: Hayes-Roth, B. Architectural Foundations for Real-Time Performance in Intelligent Agents. 1990.

Abstract: Intelligent agents perform multiple loosely coupled tasks requiring both knowledge-based reasoning and interaction with dynamic entities in the environment. We characterize the problem of real-time performance in intelligent agents as a problem in intelligent control of perception, reasoning, and action, requiring: asynchrony, timeliness, selectivity, coherence, flexibility, responsiveness, and robustness. We propose an abstract architecture for intelligent agents, a blackboard architecture incorporating limited-capacity I/O buffers, dynamic control planning, and a satisfying control regime. We show how the proposed architecture is designed to meet real-time control requirement, briefly discuss current applications of the architecture, and show how it is instantiated in one of them, surgical intensive care monitoring. We also discuss the relationship between the proposed architecture and other approaches to intelligent agents. Finally, we characterize limitations of the proposed architecture as a necessary tradeoff between optimality of performance and management of complexity.

Full paper available as ps.

Jump to... [KSL] [SMI] [Reports by Author] [Reports by KSL Number] [Reports by Year]
Send mail to: ksl-info@ksl.stanford.edu to send a message to the maintainer of the KSL Reports.