Learning and Revising Dynamic Temporal Theories in the Full Discrete Event Calculus

This paper presents the first automatic method for learning and revising dynamic temporal theories in the full-fledged Discrete Event Calculus (DEC), where fluents may be temporarily released from the law of inertia and subject to qualitative or quantitative domain laws. This is done by proposing a reformulation of the DEC, called the eXploratory Event Calculus (XEC), which can be more efficiently handled by state-of-the-art answer set solvers, and which supports a range of different logical semantics and policy options for resolving conflicts relating to the truth value or release status of fluents. The paper shows how XEC outperforms DEC on standard reasoning benchmarks, and how it can be used with an ILP system XHAIL to provide the first proof-of-principle demonstration of theory learning and revision in the full-featured DEC.