The quasi-optimizer: An intelligent system for automated knowledge acquisition and strategy optimization
Expert systems have proven to be useful tools for problem solving in many areas. However, problems concerning their use outweigh many of their advantages. These problems include the slow and error-prone task of knowledge acquisition, the inability to handle incomplete and/or inconsistent information, the use of only static knowledge in an evolving world, and the inability to merge knowledge from several experts. The Quasi-Optimizer (QO) system attempts to make some steps toward the solution of these problems. The modeling modules of the QO system automate the task of knowledge acquisition as they build decision tree models of an expert's strategy. The system can passively observe the expert or, using statistical design techniques, can create a sequence of task environments for the expert to act. In either case, the system filters out inconsistencies and generates complete models. The QO system is tested in two challenging domains. The first is the optimization of simulation models. A simulation model, in general, generates situations to which it responds by computing some action or event. This situation-action mechanism in simulation models corresponds to that of a strategy. The QO system is able to model this behavior and, with appropriate information supplied by the user, is able to verify that there are no errors and, if so, validate the simulation model. The other domain is en-route Air Traffic Control (ATC). The QO system either passively observes situations and actions performed by an ATC operator or generates time sequences of scenarios for him/her to respond to. It then builds decision tree models of the observed operators and, from these, creates a quasi-optimum strategy.
- Research Organization:
- Arizona State Univ., Tempe, AZ (United States)
- OSTI ID:
- 5458207
- Resource Relation:
- Other Information: Thesis (Ph. D.)
- Country of Publication:
- United States
- Language:
- English
Similar Records
The acquisition of strategic knowledge
Post-event human decision errors: operator action tree/time reliability correlation