Programming multi-agent systems

Viac o knihe

The earliest work on agents can be traced to the conceptualization of the actor model by Carl Hewitt, who, in a 1970s AI conference paper, described actors as entities with knowledge and goals. Research progressed with the Sprites model, where actors accessed a growing knowledge base, inspired by Hewitt’s “Scientific Computing Metaphor.” During the late 1970s and into the 1980s, a significant debate arose in AI between proponents of declarative and procedural languages. Actor researchers favored a procedural perspective, advocating for an open systems view over the closed world hypothesis inherent in logical, declarative approaches. This open systems perspective acknowledged that agents had arm's-length relationships, could not store consistent facts, and that system information could not be deemed complete (the “negation as failure” model). Subsequent work, including my own, explored using actors for general-purpose concurrent and distributed programming. By the late 1980s, several actor languages and frameworks emerged, such as Act++ (C++) by Dennis Kafura and Actalk (Smalltalk) by Jean-Pierre Briot. Recently, the Actor model has gained traction as parallel and distributed computing platforms and applications, including clusters, web services, P2P networks, multicore processors, and cloud computing, have become prevalent.