Communications of the ACM. March 1999, Vol. 42 Issue 3, p71, 1 p.
Technology development, Research and development, Knowledge-based system, System design, New technique, Intelligent agent, University of Birmingham -- Research, Computer software industry -- Research, Knowledge-based systems -- Research, System design -- Innovations, and Intelligent devices -- Design and construction
The Cognition and Affect project at the University of Birmingham has created the SIM_AGENT software toolkit designed to support new approaches to rapid prototyping of multiagent systems. A rule-based system is used to implement the many reactive mechanisms agents require for various condition-action rule sets. A list-processing model is used for the internal databases and communication channels between agents and between components within agents. SIM_AGENT is implemented in the Pop-1 extendable Lisp-like language that uses Pascal-like syntax and runs in the Poplog environment. The toolkit is used to build sets of agents, each of which has several components operating in parallel and using various mechanisms, including neural nets and AI symbol manipulating mechanisms. SIM_AGENT shows that cognitively rich internal mechanisms and structures are required to model communication and planning within and among agents, but these may produce emotional states in intelligent systems under certain conditions.