The Epistemic Planning Domain Definition Language: Official Guideline

TL;DR

The paper presents EPDDL, a standardized, PDDL‑like language for Dynamic Epistemic Logic (DEL) based epistemic planning, addressing fragmentation across DEL fragments by providing a uniform syntax and DEL‑ground semantics. It introduces abstract epistemic actions and action type libraries to capture common observability patterns, while preserving full expressivity through a formal DEL foundation and a semantics that supports global and local perspectives, multiple designated worlds, and common knowledge. The guideline details both syntax (problems, domains, action types, events, and observability conditions) and semantics (mapping EPDDL constructs to DEL models, product updates, and translations), and demonstrates how useful DEL fragments can be represented in EPDDL for interoperability, reproducible evaluation, and cross‑planner comparisons. By enabling compact fragments and reusable libraries, the approach supports a broad range of epistemic planning benchmarks and facilitates systematic benchmarking and future advances in the field.

Abstract

Epistemic planning extends (multi-agent) automated planning by making agents' knowledge and beliefs first-class aspects of the planning formalism. One of the most well-known frameworks for epistemic planning is Dynamic Epistemic Logic (DEL), which offers an rich and natural semantics for modelling problems in this setting. The high expressive power provided by DEL make DEL-based epistemic planning a challenging problem to tackle both theoretically, and in practical implementations. As a result, existing epistemic planners often target different DEL fragments, and typically rely on ad hoc languages to represent benchmarks, and sometimes no language at all. This fragmentation hampers comparison, reuse, and systematic benchmark development. We address these issues by introducing the Epistemic Planning Domain Definition Language (EPDDL). EPDDL provides a unique PDDL-like representation that captures the entire DEL semantics, enabling uniform specification of epistemic planning tasks. Our contributions are threefold: 1. A formal development of abstract event models, a novel representation for epistemic actions used to define the semantics of our language; 2. A formal specification of EPDDL's syntax and semantics grounded in DEL with abstract event models; 3. A demonstration of EPDDL's practical applicability: we identify useful fragments amenable to current planners and show how they can be represented in EPDDL. Through examples of representative benchmarks, we illustrate how EPDDL facilitates interoperability, reproducible evaluation, and future advances in epistemic planning.

Figures (5)

• Figure 1: Initial state $I$ of Example \ref{['ex:classical-bw']}. Each block is represented by a rounded square and is labeled by its name. Column names are placed below the gray line.
• Figure 2: Epistemic state of Example \ref{['ex:single-agent-bw']}. Rectangles represent worlds, labelled by their name (the actual world has a double line). Edges denote the accessibility relation. We represent labels graphically, similarly to Example \ref{['ex:classical-bw']}, $L(w_1) = \{ on(b_1, c_1), on(b_2, b_1), on(b_3, c_2), on(b_4, c_3) \}$.
• Figure 3: Epistemic state of Example \ref{['ex:multi-agent-bw']}. Since we now have multiple agents, each edge is labeled by the corresponding agent(s).
• Figure 4: Non-deterministic epistemic state of Example \ref{['ex:non-deterministic-state']}.
• Figure 5: Public announcement $a = ann(\Box_{r} \neg on(b_1, c_3))$ (left) and epistemic state $s_r \otimes a$ (right) of Example \ref{['ex:public-announcement']}. Events are graphically represented by squares (designated events are boxed). Each event $e$ is labeled by a pre-/postconditions pair $\langle \mathit{pre}(e), \mathit{post}(e)\rangle$ (if $e$ has trivial postconditions, we simply label it with $\mathit{pre}(e)$).

Theorems & Definitions (44)

• Definition 1: Classical Planning Task
• Example 1: Blocks World
• Definition 2: Single-Agent Epistemic Model and State
• Definition 3: Truth in a Single-Agent Epistemic Model
• Example 2: Single agent Epistemic Blocks World
• Definition 4: Multi-Agent Epistemic Model and State
• Definition 5: Truth in a Multi-Agent Epistemic Model
• Example 3: Multi-agent Epistemic Blocks World
• Definition 6: Multi-Pointed Epistemic Model
• Definition 7: Truth in a Multi-Pointed Epistemic Model