Dec 31, 1983
Describes an initial version of TATR (tactical air target recommender), a prototype "expert system" being developed at RAND to assist tactical air targeteers in selecting and prioritizing targets. TATR applies a knowledge engineering problem-solving approach in which human domain knowledge is essential, and judgment plays a larger role than mathematical algorithms and stochastic formalisms. Under interactive user direction, TATR preferentially orders enemy airfields, determines targets on those airfields to attack, and identifies the most effective weapon systems against those targets. TATR is programmed in the ROSIE language, which has an English-like syntax that facilitates nonprogrammer comprehension and program verification. The program applies predetermined planning heuristics to generate an airfield attack plan. It then replans, incorporating user modifications, and projects a series of plans over several days. TATR interactively maintains databases by requesting and processing updates from the user and provides detailed information about plans, friendly force capability, and enemy force posture and status.