GABAR: Graph Attention-Based Action Ranking for Relational Policy Learning

2024-12-06

AI Summary

Classical planning struggles with poor generalizability and scalability when addressing large-scale, high-complexity decision-making problems. Method: This paper proposes a relational action ordering learning framework that models action sequences as semantic dependency–encoded relational graphs, integrating graph attention mechanisms with gated recurrent units (GRUs) to jointly capture long-range action dependencies. It is the first work to formalize action ordering in planning as a relational graph learning problem—bypassing combinatorial explosion inherent in traditional search-based approaches. Contribution/Results: The method is trained solely on small-scale instances yet achieves zero-shot generalization to test problems up to an order of magnitude larger. On standard planning benchmarks, it significantly outperforms existing baselines while accelerating inference by over 10×, thereby overcoming key scalability bottlenecks of conventional planners.

Technology Category

Planning, Routing, and Scheduling: Learning for Planning and SchedulingSearch and Optimization: Learning to SearchHumans and AI: Human-Aware Planning and Behavior Prediction

Application Category

Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphsSearch and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for rankingSemantics and Knowledge: Scalable techniques for the creation, curation, publication, maintenance, and consumption of large, Web-based, structured, reusable, knowledge graphs and ontologies

Abstract

We propose a novel approach to learn relational policies for classical planning based on learning to rank actions. We introduce a new graph representation that explicitly captures action information and propose a Graph Neural Network architecture augmented with Gated Recurrent Units (GRUs) to learn action rankings. Our model is trained on small problem instances and generalizes to significantly larger instances where traditional planning becomes computationally expensive. Experimental results across standard planning benchmarks demonstrate that our action-ranking approach achieves generalization to significantly larger problems than those used in training.

Problem

• Complex Decision Problems
• Strategic Games
• Large-Scale Planning

Innovation

• GABAR method
• Graph Neural Networks
• Attention Mechanism