Automating Knowledge Acquisition in a Flightline Robot

摘要

Robots enjoy widespread use in industry on simplistic repetitive tasks in a controlled environment. Tasks in the military domain, particularly the flightline, require mobility and intelligence. While the mobility issue is being addressed, the intelligence issue is not. By giving the robot the ability to learn from a novice, the robot could be placed on the flightline and learn what it needs to know from the domain experts. Rather than deal with a toy problem, this work takes the actual refueling 'Job Guides' for two aircraft and shows how these can be used directly by a computer. The process involves three steps. First the text is transformed into a common natural language processor form. Second the forms are expanded by applying expert system and planning techniques to include missing domain and world knowledge. Finally the forms are examined to allow the commonalty between the two refueling tasks to be extracted.