Saved in:
Bibliographic Details
Main Author: Pastukhov, Sergey
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2504.04366
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866910905073139712
author Pastukhov, Sergey
author_facet Pastukhov, Sergey
contents We introduce a novel hierarchical reinforcement learning (HRL) framework that performs top-down recursive planning via learned subgoals, successfully applied to the complex combinatorial puzzle game Sokoban. Our approach constructs a six-level policy hierarchy, where each higher-level policy generates subgoals for the level below. All subgoals and policies are learned end-to-end from scratch, without any domain knowledge. Our results show that the agent can generate long action sequences from a single high-level call. While prior work has explored 2-3 level hierarchies and subgoal-based planning heuristics, we demonstrate that deep recursive goal decomposition can emerge purely from learning, and that such hierarchies can scale effectively to hard puzzle domains.
format Preprint
id arxiv_https___arxiv_org_abs_2504_04366
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Solving Sokoban using Hierarchical Reinforcement Learning with Landmarks
Pastukhov, Sergey
Artificial Intelligence
We introduce a novel hierarchical reinforcement learning (HRL) framework that performs top-down recursive planning via learned subgoals, successfully applied to the complex combinatorial puzzle game Sokoban. Our approach constructs a six-level policy hierarchy, where each higher-level policy generates subgoals for the level below. All subgoals and policies are learned end-to-end from scratch, without any domain knowledge. Our results show that the agent can generate long action sequences from a single high-level call. While prior work has explored 2-3 level hierarchies and subgoal-based planning heuristics, we demonstrate that deep recursive goal decomposition can emerge purely from learning, and that such hierarchies can scale effectively to hard puzzle domains.
title Solving Sokoban using Hierarchical Reinforcement Learning with Landmarks
topic Artificial Intelligence
url https://arxiv.org/abs/2504.04366