This thesis describes a course scheduling system that models planning as a satisfiability problem in relational logic. Given a set of course requirements for a degree program, our system can find a schedule of courses that will complete these requirements. It supports a flexible XML format for expressing course requirements and also handles additional user-specific constraints, such as requirements that certain courses be taken at particular times. Various optimizations were included in the translation to relational logic to improve the performance of our system and the quality of its results. We ran experiments on our system using degree programs from the Department of Electrical Engineering and Computer Science at MIT as input, and found that our approach is competitive with conventional planners.

In this work we describe a prototype system, Padok (Planning with Domain Knowledge), where the whole planning domain is modelled in Linear Time Logic (LTL) and planning is reduced to model search. The problem specification accepted by the system can include problem specific information, in a declarative format. The work briey resumes and extends (Cialdea Mayer et al. 2000), showing, through a complete example, how LTL is effectively used as a planning language.

Model checking is a widely used technique in verification of dynamic systems. Recently several papers have shown that model checking can be used to do planning. In the present paper we report the results of a set of experiments on using two wellknown model checkers, Spin and SMV, to do traditional planning (i.e, planning in a complete information setting for reaching a state where a certain goal is satisfied). In the experiments we have compared the performances of such model checkers with stateof-art planners, IPP, FF, and TLPLAN, on problems used in AIPS'98 and AIPS'00 competitions.