Improving Planning Performance in PDDL+ Domains via Automated Predicate Reformulation

Abstract

In the last decade, planning with domains modelled in the hybrid PDDL+ formalism has been gaining significant research interest. A number of approaches have been proposed that can handle PDDL+, and their exploitation fostered the use of planning in complex scenarios. In this paper we introduce a PDDL+ reformulation method that reduces the size of the grounded problem, by reducing the arity of sparse predicates, i.e. predicates with a very large number of possible groundings, out of which very few are actually exploited in the planning problems. We include an empirical evaluation which demonstrates that these methods can substantially improve performance of domain-independent planners on PDDL+ domains.

A Logical Semantics for PDDL+

“…However, if one notes (or derives via automated theorem-proving) that only a small fraction of road segments may ever be affected by a particular grounding of F lowGreen (i.e., the maximum number of roads that meet at any one intersection), the SEA is reduced to a manageable size. This appears to be a case of a sparse predicate [17] which presents a serious difficulty to PDDL+ planners as well, except it manifests in the explosion of search spaces rather than axioms.…”

“…All this work points to the fact that PDDL+ planning is significantly more involved than the purely discrete case and might benefit from moving further away from grounding. Recent research focuses on ways of reducing the difficulty by improving modelling techniques [11] or automatically reducing the arity of predicates with large numbers of possible groundings [17]. Although generally less tractable than the PDDL family of languages, situation calculus has been used successfully for planning purposes in [13].…”

Effective grounding for hybrid planning problems represented in PDDL+

“…We want to assess the importance of grounding on realistic and complex hybrid problems. For this reason, following the approach exploited by Franco et al (2019), the experimental evaluation is performed by considering four benchmark domains: two that have been used to models real-world applications, and two that are derived from well-known benchmarks exploited in past International Planning Competitions (Vallati et al, 2018).…”

“…Limited work has been carried out on reformulation of non-classical PDDL models. Chrpa et al (2015) extended the notion of entanglements to numerical planning, and Franco et al (2019) focused on PDDL+ reformulation to reduce the arity of predicates and fluents to limit the exponential explosion of the ground problem size. A different line of work on reformulation investigated techniques to reduce the performance of automated solvers, to identify aspects of the models to which existing planning engines are sensitive to (Vallati and Chrpa, 2019).…”

Collaborative Robotic Manipulation: A Use Case of Articulated Objects in Three-dimensions with Gravity

“…We adopt ENHSP as a planning engine [22], [23], and the well-known MoveIt framework for motion planning and execution. ENHSP has been selected on the basis of a comparative analysis involving different planners supporting PDDL+ [19], and of its good performance on PDDL+ benchmarks [24]. An extensive analysis of the overall performance of ENHSP is out of the scope for our discussion.…”