Comparing Functional and Imperative Programming Languages on Classical Planning Problems

1 SUMMARY In this project, we compare imperative (Python) and functional (Haskell) programming languages on solving the automated planning problems. Functional language is useful since the classical planning problem can be reduced to boolean satisfiability problem (SAT), where functional language and recursion provide a natural way of implementation. On the other hand, large-scale planning problems also suffer from the curse of dimensionality, where the efficiency of implementation plays a more important role. In this case, we might prefer imperative language. The goal of the project is thus to investigate the tradeoffs of functional and imperative programming on planning problems of different scales. Our results show that: (i) Recursion is a more natural way for classical planning, and thus functional programming is preferred qualitatively for implementation. (ii) On the other hand, iterative implementation is required for large-scale planning problem, where imperative language is preferred in terms of quantitative results. Furthermore, we explore the use of planning heuristics to alleviate the curse of dimensionality at scale and enable the functional language implementation to deal with large-scale planning.