Representing Actions with an Assumption-Based Truth Maintenance System

Proceedings of the AAAI Conference on Artificial Intelligence, 5

Volume

Science

Automated Reasoning

The Assumption-based Truth Maintenance System, introduced by de Kleer, is a powerful new tool for organizing a search through a space of alternatives. However, the ATMS is oriented towards inferential problem solving, and provides no special mechanisms for modeling actions or state changes. We describe an approach to applying the ATMS to the task of representing the effects of actions. The approach extends traditional tree-structured context mechanisms to allow context merges. It also takes advantage of the underlying ATMS to detect inconsistent contexts and to maintain derived results. Some results are presented concerning possible approaches to the treatment of merges in questionable circumstances. Finally, the analysis of actions in terms of a truth maintenance system suggests the need for a more elaborate treatment of contradiction in such systems than exists at present.