PDDL2.1: An Extension to PDDL for Expressing Temporal Planning Domains (1106.4561v1)

Abstract: In recent years research in the planning community has moved increasingly toward s application of planners to realistic problems involving both time and many typ es of resources. For example, interest in planning demonstrated by the space res earch community has inspired work in observation scheduling, planetary rover ex ploration and spacecraft control domains. Other temporal and resource-intensive domains including logistics planning, plant control and manufacturing have also helped to focus the community on the modelling and reasoning issues that must be confronted to make planning technology meet the challenges of application. The International Planning Competitions have acted as an important motivating fo rce behind the progress that has been made in planning since 1998. The third com petition (held in 2002) set the planning community the challenge of handling tim e and numeric resources. This necessitated the development of a modelling langua ge capable of expressing temporal and numeric properties of planning domains. In this paper we describe the language, PDDL2.1, that was used in the competition. We describe the syntax of the language, its formal semantics and the validation of concurrent plans. We observe that PDDL2.1 has considerable modelling power --- exceeding the capabilities of current planning technology --- and presents a number of important challenges to the research community.

• The paper demonstrates that pddl2.1 extends PDDL by integrating durative actions and numeric expressions to better represent time-dependent planning challenges.
• It introduces formal semantics that support concurrent plan validation through precise management of temporal constraints and numeric invariants.
• The advancements in pddl2.1 empower planners to generate higher-quality plans for real-world applications in domains like space mission planning and manufacturing.

pddl2.1: An Extension to pddl for Expressing Temporal Planning Domains

The paper "pddl2.1: An Extension to pddl for Expressing Temporal Planning Domains" by Maria Fox and Derek Long, published in the Journal of Artificial Intelligence Research in 2003, addresses pivotal advancements in the Planning Domain Definition Language (pddl). This essay provides an insightful overview of the paper, focusing on its contributions to temporal and numeric modeling capabilities in planning domains while examining both practical and theoretical implications.

Abstract

Fox and Long articulate a significant need within the AI planning community to extend pddl, accommodating realistic, application-driven problems involving time and various resource types. This consideration was primarily driven by the exigencies of the third International Planning Competition (IPC), which necessitated a modeling language capable of representing temporal and numeric properties. Consequently, pddl2.1 was developed and showcased, demonstrating enhanced modeling power yet highlighting new challenges for research.

Introduction

Launched in 1998 by Drew McDermott, the original pddl facilitated substantial progress in planning research. Over time, the community's focus shifted from propositional puzzle domains to more realistic problems necessitating reasoning about time and numeric quantities. The 2002 IPC, setting its sights on time and numeric resources, underscored the limitations of pddl and catalyzed the development of pddl2.1. This paper reflects on these developments, describes the syntax and semantics of pddl2.1, and discusses the validation of concurrent plans.

Enhancements in pddl2.1

Temporally Annotated Plans and Numeric Extensions

pddl2.1 introduces significant enhancements to model time and numeric resources, marking a departure from the purely logical constructs of pddl. The additions include:

1. Numeric Expressions: Allowing arithmetic operations on numeric quantities, pddl2.1 facilitates the inclusion of numeric functions in preconditions and effects. Functions can be updated using expressions such as increase, decrease, scale-up, and scale-down. This extension supports the modeling of complex resource constraints essential in application-driven domains.
2. Durative Actions: Durative actions replace instantaneous actions by encapsulating the temporal extent of actions. These actions allow conditions and effects to be temporally annotated, distinguishing between those required at the start, end, or over the entire duration of an action.

Plan Metrics

pddl2.1 allows domain designers to specify plan metrics. This extension is crucial for evaluating plans based on domain-specific criteria such as resource consumption or overall utility. Plan metrics empower planners to generate higher-quality plans tailored to real-world applications.

Formal Semantics and Plan Validation

The paper explores the formal semantics of pddl2.1, expanding the classical state-transition model to accommodate temporal and numeric aspects. Key constructs include:

1. Logical States and States: Extending states to include numerical values alongside logical propositions.
2. Durative Actions and Concurrent Plans: Handling concurrency and defining non-interference conditions for concurrent action execution.
3. Plan Execution and Validation: Specifying conditions for plan validity by ensuring that all temporal constraints and numeric invariants are respected throughout the plan's execution.

Validation of plans in the presence of numeric and temporal constraints is crucial for practical applications and has been adeptly addressed in the context of the IPC, underscoring the importance of thorough validation mechanisms.

Implications and Future Directions

Practical Implications

The enhanced modeling capabilities of pddl2.1 have practical implications in various domains, from space mission planning to manufacturing. The language's ability to handle complex temporal and numeric constraints makes it highly suitable for real-world problem solving, bridging the gap between theoretical research and practical applications.

Theoretical Implications

The introduction of continuous effects and duration inequalities raises interesting theoretical challenges. The formal semantics outlined in the paper provide a robust foundation for further exploration, although the undecidability brought by numeric expressions warrants caution and careful modeling in practical scenarios.

Conclusion

The paper by Fox and Long highlights critical advancements in extending pddl to meet the demands of realistic planning problems involving time and resources. pddl2.1 significantly enhances the expressive power of the language, addressing the needs of the planning community. While these advancements present new challenges, they also pave the way for future research and developments, ultimately pushing the boundaries of what AI planning systems can achieve in dynamic and complex environments. The work underscores the continuous evolution of planning languages, ensuring their alignment with real-world complexities and application-driven demands.