Abstract
Bernard P. Zeigler established a rigorous systems foundation for model-based intelligent agents. With his collaborators, he developed architectures in which agents use internal representations to plan actions, monitor consequences, coordinate with other agents, and recover when execution departs from expectations. The 1988 robot-organization architecture and the 1990 monograph on intelligent agents and endomorphic systems articulated these capabilities 38 and 36 years before 2026. His research program connects the Discrete Event System Specification (DEVS) formalism, System Entity Structure (SES), and experimental frames into complementary foundations for behavior, architecture, and validity. These foundations address central engineering demands of contemporary artificial intelligence (AI): explicit state, controlled tool use, event timing, component interaction, and reproducible evaluation. This paper develops their application to agents built around large language models (LLMs), presents a Parallel DEVS execution architecture, and verifies a deadline-resolution contract. Zeigler’s enduring contribution is a disciplined way to turn intelligent decision capability into a specified, executable, and testable system.