Luigi Bonassi, Francesco Percassi, Enrico Scala

BLAST: Bit-Blasting Numbers for Classical Planning (Extended Abstract)

Introduction

Blast: bit-blasting numbers for classical planning (extended abstract). Uncover how bit-blasting numbers makes classical planning practical. Optimized binary compilations and axioms enable classical planners to compete with numeric planners.

Abstract

It is well known that numeric planning can be made decidable if the domain of all numeric state variables is finite. This bounded formulation can be polynomially compiled into classical planning with Boolean conditions and conditional effects preserving the plan size exactly. However, it remains unclear whether this compilation has any practical utility. To explore this aspect, this work revisits the theoretical compilation framework from a practical perspective, focusing on the fragment of simple numeric planning. Specifically, we introduce three different compilations. The first, called one-hot, aims to systematise the current practice among planning practitioners of modelling numeric planning through classical planning. The other two, termed binary compilations, extend and specialise the logarithmic encoding introduced in previous literature. Our experimental analysis reveals that the overly complex logarithmic encoding can, surprisingly, be made practical with some representational expedients. Among these, the use of axioms is particularly crucial. Furthermore, we identify a class of mildly numeric planning problems where a classical planner, i.e., LAMA, when run on the compiled problem, is highly competitive with state-of-the-art numeric planners.

Review

This extended abstract, titled "BLAST: Bit-Blasting Numbers for Classical Planning," tackles a crucial gap between theoretical tractability and practical utility in numeric planning. The authors revisit the long-established concept of compiling finite-domain numeric planning into classical Boolean planning, a process known to preserve plan size and offer decidability. While this theoretical foundation has existed, its practical applicability has remained largely unexplored. The paper's primary contribution lies in its systematic investigation of this aspect, focusing specifically on the fragment of simple numeric planning and introducing novel and re-evaluated compilation strategies. The methodology centers around three distinct compilation schemes: a "one-hot" encoding that aims to formalize existing practitioner approaches, and two "binary compilations" which extend and specialize previous logarithmic encodings. A key and surprising finding from their experimental analysis is that the often-perceived complexity of logarithmic encoding can, in fact, be made practical through careful representational choices, with the judicious use of axioms proving particularly impactful. Furthermore, the work identifies a specific class of "mildly numeric" planning problems where a classical planner, LAMA, when applied to these compiled representations, demonstrates high competitiveness against specialized state-of-the-art numeric planners. Overall, this work presents a compelling argument for the practical viability of compiling numeric planning problems into classical ones, particularly in certain problem settings. The identification of effective representational expedients for logarithmic encodings and the demonstration of a classical planner's competitiveness in mildly numeric domains are significant contributions. While presented as an extended abstract, the results offer valuable insights for both researchers and practitioners, suggesting new avenues for leveraging mature classical planning techniques in numeric domains and highlighting the potential for bridging the divide between theoretical completeness and practical performance. Further work detailing the characteristics of "mildly numeric" problems and a deeper complexity analysis would undoubtedly enhance the full paper.

Full Text

Comments

You need to be logged in to post a comment.