Abstract: | Cassandra is a contingency planner that relaxes two of the perfect knowledge assumptions of classical planning. The planner is able to construct viable plans in cases where the agent's knowledge of the initial conditions is uncertain and the effects of some actions possibly non-deterministic. This feature makes it a potential planner for real world problems. Cassandra's original version in Lisp was too inefficient to be applicable to any real world problems. In this thesis we present a new version in Prolog and introduce some search control techniques that improve Cassandra's efficiency.
|