Multi-agent Protocol Recognition during Simulation
|
Abstract:
Domain experts often express their knowledge in a different form than knowledge engineers would like. We implemented a plan recognition system within a simulation based tutor using a protocol formalism designed to closely resemble a form used by domain experts to communicate with each other. The system was responsible for using knowledge encoded in a linear formalism to implicitly recognize parallel activity. The recognition system supports recovery from incorrect user actions and accounts for synchronization of multiple agents and plans. The protocol mechanism was a component in a tutoring system based upon a real time simulation of cooperating agents following the user`s orders. A robust model of expert behavior was compared with the user's actions which were classified as correct, incorrect or partially correct for use in constructing a model of the user's understanding of the task being taught. The user model was indexed into the simulation model so that the state of the simulation could be biased towards optimal situations for learning. We have addressed several difficult issues in reasoning about multiple agents and using a real time simulation for training. This paper describes the semantics of the protocol mechanism. 1
Citations
| 38 | Student modeling – VanLehn - 1988 |
| 22 | Lies, Damned Lies and Stereotypes: Pragmatic Approximations Of Users – Kay - 1994 |
| 8 | Reasoning about the user within a simulation-based real-time training system – Eliot, Woolf - 1994 |
| 7 | Steps from explanation planning to model construction dialogues – Suthers, Woolf - 1992 |
| 3 | Constructive Interpretation of Human-Generated Exceptions During Plan Execution – Broverman - 1991 |
| 2 | Apprising the User of User Models: Interface Guidelines – Orwant - 1994 |
| 1 | Association. Advanced Cardiac Life Support – Heart - 1987 |
