ICAPS 2025 Tutorial on Learning for Generalised Planning

Monday, 10 November 2025 | 13:30 - 15:00 AEDT

Dillon Z. Chen	Felipe Trevizan	Sylvie Thiébaux

Brief description

This tutorial will cover recent advances in Learning for Generalised Planning (L4P), as part of ICAPS 2025. Generalised Planning (GP) refers to the task of synthesising programs that solve families of related planning problems. We cover approaches that employ learning over symbolic models and training data for tackling GP. The aim of L4P is to scale up planning technology to solve problems of greater difficulty and number of objects. Indeed due to no free lunch, one cannot expect traditional PDDL planners which plan for each problem individually to handle all domains well. L4P is a rapidly growing subfield of AI Planning that has garnered the attention of both learning researchers and planning researchers alike. A semantic analysis of published ICAPS papers shows that the total number of planning papers using learning until 2024 ranks 3rd, up from 7th in 2019.

The scope of the tutorial involves covering the L4P problem setup, current approaches, and theoretical results. More specifically, the tutorial will provide a comprehensive overview of the state-of-the-art for L4P which ranges from purely symbolic and inductive approaches, to deep learning architectures, and lastly the usage of language models. The tutorial will include a hands-on lab component for participants to get familiar with L4P frameworks and architectures.

Useful Resources

• Slides (PDF) — Presentation slides.
• Some Generalised Planners
• BFGP — Generalised planner that performs heuristic search over generalised plans represented as assembly-like programs, or structured programs.
• GOOSE — Generalised planner that learns heuristics for search with graph learning algorithms.
• PG3 — Generalised planner that searches over generalised plans represented as decision lists.
• Sketch Learner — Generalised planner that learns policy sketches.
• Some Generalised Planning Tools
• DLPlan — Feature generator for planning problems using description logics.
• Mimir — PDDL parser and planning interface
• QNP to FOND Translator — Translator from QNP problems to FOND problems.
• WLPlan — Feature generator for planning problems using graph learning algorithms.
• IPC 2023 Learning Track — Generalised planning benchmarks from the 2023 International Planning Competition.