Differentiable Learning of Lifted Action Schemas for Classical Planning

2026-05-13

AI Summary

This work addresses the problem of learning lifted action schemas from state trajectories in classical planning domains where states are fully observable but action parameters are unknown. To this end, the authors propose a differentiable neural architecture that, given only sequences of states, jointly infers action parameters and learns structured action models in an end-to-end trainable framework. This approach is the first to nearly perfectly recover ground-truth action structures under the challenging setting of implicit action parameterization. Empirical evaluation across multiple classical planning domains demonstrates its effectiveness, robustness to observation noise, and strong generalization capabilities—even in variants with slot-dynamic object models. The method thus provides a differentiable component for neuro-symbolic systems, bridging symbolic planning representations with gradient-based learning.

Abstract

Classical planners can effectively solve very large deterministic MDPs represented in STRIPS or PDDL where states are sets of atoms over objects and relations, and lifted action schemas add or delete these atoms. This compact representation yields strong search heuristics and provides an ideal setting for structural generalization, since lifted relations and action schemas give rise to infinitely many domain instances. A central challenge is to learn these relations and action schemas from data, and recent approaches have addressed this problem using different types of observations. In this work, we develop a novel neural network architecture for learning action schemas from traces where states are fully observed but action arguments are unobserved. The problem is a simplification but an important step towards learning planning domains from sequences of images and action labels, and we aim to solve this simplification in a nearly perfect manner. The challenge lies in learning the action schemas while simultaneously identifying the action arguments from observed state changes. Our approach yields a robust differentiable component that can then be integrated into larger neuro-symbolic models. We evaluate the architecture on various planning domains, where the learned lifted action schemas must recover the ground-truth structure. Additionally, we report experiments on robustness to observation noise and on a variation related to slot-based dynamics models.

Problem

• lifted action schemas
• classical planning
• state transitions
• action arguments
• differentiable learning

Innovation

• differentiable learning
• lifted action schemas
• classical planning
• neuro-symbolic integration
• action argument identification