[Paper Review] Planimation

Gang Chen, Yi Ding+10 more|arXiv (Cornell University)|Aug 11, 2020

AI-based Problem Solving and Planning4 citations

TL;DR

Planimation is an open-source, web-based framework that visualizes PDDL planning solutions through an extensible animation profile, enabling intuitive understanding of plan execution via dynamic sprite-based animations. It decouples visualization from planning by extending PDDL with declarative animation rules, allowing modelers to debug plans and improve model-solver alignment through visual feedback.

ABSTRACT

Planimation is a modular and extensible open source framework to visualise sequential solutions of planning problems specified in PDDL. We introduce a preliminary declarative PDDL-like animation profile specification, expressive enough to synthesise animations of arbitrary initial states and goals of a benchmark with just a single profile.

Motivation & Objective

• To close the feedback loop in AI planning by enabling visual debugging of PDDL solutions.
• To reduce the cognitive load of understanding plan dynamics by transforming abstract plans into intuitive visual sequences.
• To support modelers in detecting misalignments between mental models and PDDL specifications through visual cues.
• To provide a modular, extensible framework that integrates with existing planners and visualization pipelines.
• To enhance explainability in AI planning by enabling human-in-the-loop validation of plan correctness and behavior.

Proposed method

• Extends PDDL with a declarative animation profile specifying visual properties (position, color, size, image) for each object or type.
• Uses effects in the animation profile to dynamically update object properties (e.g., x, y, color) based on predicate truth in each state.
• Introduces constraint functions like `distributex`, `distributey`, `distribute_grid_around_point`, and `align_middle` to automatically layout objects based on state conditions.
• Employs a two-part architecture: a Unity WebGL visualizer engine and a Python-based web API solver that interfaces with planners via the solver.planning.domains API.
• Generates JSON-based visualisation files (VFG) for persistent storage and offline playback, supporting export as GIF or WebM.
• Implements a web UI with interactive controls: step-by-step navigation, subgoal tracking, playback speed adjustment, and goal state visualization.

Experimental results

Research questions

• RQ1 How can visual feedback improve the detection of modeling errors in PDDL planning problems?
• RQ2 To what extent can a declarative animation profile decouple visualization from planning while remaining expressive enough for diverse domains?
• RQ3 Can automated layout functions (e.g., `distributex`) reduce manual configuration and improve consistency in plan visualizations?
• RQ4 How effective is visual plan animation in supporting human understanding and debugging of planning solutions?
• RQ5 What is the impact of visualizing plan execution on the explainability and usability of AI planning systems?

Key findings

• Planimation successfully visualizes sequential plans from PDDL problems across multiple domains, including Blocksworld, Grid, Logistics, and Towers of Hanoi.
• The use of constraint functions like `distributex` enables automatic, consistent layout of objects (e.g., blocks on a table) without manual coordinate assignment.
• The animation profile allows modelers to extend PDDL problems with minimal effort—just a single animation file—enabling rich visual feedback without modifying the core planner.
• The framework supports interactive exploration: users can click on actions to jump to specific states, inspect preconditions and effects, and track subgoal satisfaction in real time.
• The system enables export of visualizations as JSON, GIF, or WebM, facilitating reuse, sharing, and integration into educational or debugging workflows.
• Planimation has been deployed as a web application at planimation.planning.domains and is open-sourced on GitHub, with active community development and extensibility in mind.