Symbolic Pattern Planning

16 February 2024 | Sala Stringa | 11:00 | Matteo Cardellini (Politecnico di Torino and Università degli Studi di Genova)   

Abstract

We present a novel approach for solving numeric planning problems, called Symbolic Pattern Planning. Given a planning problem Π, a bound n and a pattern≺ –defined as an arbitrary sequence of actions– we encode the problem of finding a plan for Π with bound n as an SMT formula Π≺n with fewer variables and clauses than all the other state-of-the-art encodings. Moreover, we prove that, for anybound n, the other encodings can never find a valid plan while ours does not. We argue that our approach provides a new starting point for symbolic planning,allowing to bridge the gap with search-based planning (this is joint work with Marco Maratea). To sustain such claim, we present some recent results (under submission) which show the benefits of integrating search-based techniques in Symbolic Pattern Planning. Further, we present how Symbolic Pattern Planning can be effectively extended to the temporal setting (under submission). All the claims are supported by extensive experimental comparisons involving the various available planning systems. 


Bio
Matteo Cardellini is a 3rd year Ph.D. student in the Italian National Ph.D. in Artificial Intelligence co-sponsored by Politecnico di Torino and Università degli Studi di Genova. His work is focused on planning and scheduling techniques for both domain-independent approaches and in applications like railway dispatching, urban-traffic and rehabilitation scheduling.