Users Talk to their Model Trains: Interaction with a Speech-based Multi-Agent System

摘要

This paper presents an interactive multi agent system (MAS) for controlling a model train installation. User intentions can be communicated to the system via speech and are processed by hierarchically organized autonomous agents with distributed knowledge and responsibilities in the application domain. Agents on deeper levels in the hierarchy execute, observe, and refine plans constructed by higher level agents. Conflicts between requirements of tasks and the current application state are resolved by local planning and negotiating with the user. For this purpose a multi-modal natural language dialogue system is incorporated as the user interface.