Reformulation Techniques for Automated Planning: A Systematic Review

TL;DR

This paper systematically surveys reformulation techniques for domain-independent, classical planning, organizing approaches such as entanglements, macro-operators, operators elimination, bagged representation, action schema splitting, and domain model configuration. It uses the Gripper domain as a running example to illustrate how each technique re-represents problems to improve planning efficiency, discusses their advantages and limitations, and compares their applicability. The authors also examine the synergy among techniques, the focus on ground representations, and gaps in lifted planning and cross-language reformulations, offering guidance for future research. Additionally, the review briefly surveys reformulation work beyond classical planning, including numeric, temporal, probabilistic, and multi-agent settings, and translation approaches to broaden solver compatibility. Overall, the work provides a qualitative, domain-agnostic map of reformulation strategies, highlighting promising combinations and identifying gaps for future exploration in both classical and non-classical planning contexts.

Abstract

Automated planning is a prominent area of Artificial Intelligence, and an important component for intelligent autonomous agents. A cornerstone of domain-independent planning is the separation between planning logic, i.e. the automated reasoning side, and the knowledge model, that encodes a formal representation of domain knowledge needed to reason upon a given problem to synthesise a solution plan. Such a separation enables the use of reformulation techniques, which transform how a model is represented in order to improve the efficiency of plan generation. Over the past decades, significant research effort has been devoted to the design of reformulation techniques. In this paper, we present a systematic review of the large body of work on reformulation techniques for classical planning, aiming to provide a holistic view of the field and to foster future research in the area. As a tangible outcome, we provide a qualitative comparison of the existing classes of techniques, that can help researchers gain an overview of their strengths and weaknesses.

Figures (7)

• Figure 1: The Gripper PDDL domain model.
• Figure 2: An example Gripper planning problem that consists of 4 balls to be carried from rooma to roomb.
• Figure 3: An overview of the use of a domain- and planner-independent reformulation technique.
• Figure 4: An excerpt of the modified Gripper planning task when the pick operator is entangled by init on the basis of the at predicate.
• Figure 5: An example macro-operator that encapsulates the sequence of operators pick(?obj ?from ?gripper), move(?from ?to), drop(?obj ?to ?gripper). For the sake of clarity, we use the same name for matched variables between operators.