Two AI languages for simulation

Modeling human decision making lies on the boundary between simulation and artificial intlligence. In the problem areas that the authors are investigating by simulation they model human decision makers as simulation agents, each expected to exhibit intelligent behavior. The agents must be intentional and they must anticipate future situations in generating proactive behavior. They must respond to the actions of other agents and to simulation events. To achieve this level of agent performance the authors have developed a declarative language of goals, achievement plans, and procedures. The language explicitly captures the procedural knowledge required to achieve agent goals. To build the object-oriented simulator, the authors have developed a frame language. The frame language provides separate concept and role hierarchies used to define simulation agents and objects. The problem-solving power of the present simulations is derived from the integration of the frame language, the procedural language, and the simulator. The focus here will be on the syntax and semantics of the frame language and the procedural language. To illustrate the use of these language tools, the authors describe an interactive wargame simulation designed to support military planning in the DARPA SIMNET project. The wargame, based on a frequently used and highly successful manual wargaming method, is deriven by familiar military domain heuristics. It provides the ability to rapidly explore many alternative event outcomes, significantly extending the depth and breadth of the problem space that the planning staff is able to explore.