Planning in the Fluent Calculus Using Binary Decision Diagrams

As a part of my group’s activities on the extension of the fluent calculus (Hölldobler and Schneeberger 1990; Thielscher 1998), I’ve been working to provide an efficient inference engine for the fluent calculus by using the efficiency of binary decision diagrams (BDDs) (Bryant 1986). In the past, BDDs have significantly improved the performance of algorithms and enabled the solution of new classes of problems in areas such as formal verification and logic synthesis (see, for example, Burch et al. [1992]). Surprisingly, BDDs have only recently been introduced to implement the solution of planning problems (Cimatti et al. 1997). The goal of our project was to investigate whether BDDs might also help to increase the efficiency of algorithms solving problems in the field of reasoning about action and change. For a start, I have implemented the solution of fluent calculus planning problems restricted to deterministic actions and propositional fluents (Hölldobler and Störr 2000; Störr 2001).