Formalizing Complex Task Libraries in Golog

We present an approach to building libraries of tasks in complex action languages such as Golog, for query answering. Our formalization is based on a situation calculus framework that allows probabilistic, temporal actions. Once a knowledge base is built containing domain knowledge including type information and a library of tasks and the goals they can achieve, we are interested in queries about the achievability of goals. We consider cases where, using domain object type and goal information in the KB, a user is able to get specific answers to a query while leaving some of the specifics for the system to figure out. In some cases where the specifics are missing from the KB, the user is provided with the possible alternative answers that are compatible with the incomplete information in the KB. This approach is being explored in the context of a military operations planning domain for decision support.

• Type: Conference
• Year: 2006
• Where: ECAI
• Authors: Alfredo Gabaldon
• Keywords: Artificial Intelligence, Complex Action Languages, Domain Object Type, ECAI 2006, Situation Calculus Framework