Breaking the Grid: Distance-Guided Reinforcement Learning in Large Discrete and Hybrid Action Spaces

2026-02-09

AI Summary

This work addresses the challenge of scaling reinforcement learning to large discrete or hybrid action spaces, where traditional approaches suffer from the curse of dimensionality due to reliance on grid structures or expensive nearest-neighbor searches. The authors propose Distance-Guided Reinforcement Learning (DGRL), which leverages a random-body exploration mechanism in a semantic embedding space to achieve full coverage of local trust regions. DGRL introduces Sampling-based Dynamic Neighborhoods (SDN) and Distance-Based Policy Updates (DBU), transforming policy optimization into a stable regression task that decouples gradient variance from action space size. This enables effective non-hierarchical modeling of hybrid actions. Experiments demonstrate that DGRL outperforms state-of-the-art methods by up to 66% in both structured and unstructured environments, while significantly accelerating convergence and reducing computational complexity.

Technology Category

Machine Learning: Reinforcement LearningSearch and Optimization: Learning to SearchMultiagent Systems: Adversarial Agents

Application Category

Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for rankingGraph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphsSemantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions

Abstract

Reinforcement Learning is increasingly applied to logistics, scheduling, and recommender systems, but standard algorithms struggle with the curse of dimensionality in such large discrete action spaces. Existing algorithms typically rely on restrictive grid-based structures or computationally expensive nearest-neighbor searches, limiting their effectiveness in high-dimensional or irregularly structured domains. We propose Distance-Guided Reinforcement Learning (DGRL), combining Sampled Dynamic Neighborhoods (SDN) and Distance-Based Updates (DBU) to enable efficient RL in spaces with up to 10$^\text{20}$ actions. Unlike prior methods, SDN leverages a semantic embedding space to perform stochastic volumetric exploration, provably providing full support over a local trust region. Complementing this, DBU transforms policy optimization into a stable regression task, decoupling gradient variance from action space cardinality and guaranteeing monotonic policy improvement. DGRL naturally generalizes to hybrid continuous-discrete action spaces without requiring hierarchical dependencies. We demonstrate performance improvements of up to 66% against state-of-the-art benchmarks across regularly and irregularly structured environments, while simultaneously improving convergence speed and computational complexity.

Problem

• large discrete action spaces
• curse of dimensionality
• reinforcement learning
• hybrid action spaces
• irregularly structured domains

Innovation

• Distance-Guided Reinforcement Learning
• Sampled Dynamic Neighborhoods
• Distance-Based Updates
• Large Discrete Action Spaces
• Hybrid Action Spaces