Hierarchical reinforcement learning based on macro actions

Abstract The large action space is a key challenge in reinforcement learning. Although hierarchical methods have been proven to be effective in addressing this issue, they are not fully explored. This paper combines domain knowledge with hierarchical concepts to propose a novel Hierarchical Reinforc...

Full description

Saved in:
Bibliographic Details
Main Authors: Hao Jiang, Gongju Wang, Shengze Li, Jieyuan Zhang, Long Yan, Xinhai Xu
Format: Article
Language:English
Published: Springer 2025-04-01
Series:Complex & Intelligent Systems
Subjects:
Online Access:https://doi.org/10.1007/s40747-025-01895-9
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1849326637739409408
author Hao Jiang
Gongju Wang
Shengze Li
Jieyuan Zhang
Long Yan
Xinhai Xu
author_facet Hao Jiang
Gongju Wang
Shengze Li
Jieyuan Zhang
Long Yan
Xinhai Xu
author_sort Hao Jiang
collection DOAJ
description Abstract The large action space is a key challenge in reinforcement learning. Although hierarchical methods have been proven to be effective in addressing this issue, they are not fully explored. This paper combines domain knowledge with hierarchical concepts to propose a novel Hierarchical Reinforcement Learning framework based on macro actions (HRL-MA). This framework includes a macro action mapping model that abstracts sequences of micro actions into macro actions, thereby simplifying the decision-making process. Macro actions are divided into two categories: combat macro actions (CMA) and non-combat macro actions (NO-CMA). NO-CMA are driven by decision tree-based logical rules and provide conditions for the execution of CMA. CMA form the action space of the reinforcement learning algorithm, which dynamically selects actions based on the current state. Comprehensive tests on the StarCraft II maps Simple64 and AbyssalReefLE demonstrate that the HRL-MA framework exhibits superior performance, achieving higher win rates compared to baseline algorithms. Furthermore, in mini-game scenarios, HRL-MA consistently outperforms baseline algorithms in terms of reward scores. The findings highlight the effectiveness of integrating hierarchical structures and macro actions in reinforcement learning to manage complex decision-making tasks in environments with large action spaces.
format Article
id doaj-art-cf6601ba17fc4a63b93486a40e461650
institution Kabale University
issn 2199-4536
2198-6053
language English
publishDate 2025-04-01
publisher Springer
record_format Article
series Complex & Intelligent Systems
spelling doaj-art-cf6601ba17fc4a63b93486a40e4616502025-08-20T03:48:06ZengSpringerComplex & Intelligent Systems2199-45362198-60532025-04-0111611710.1007/s40747-025-01895-9Hierarchical reinforcement learning based on macro actionsHao Jiang0Gongju Wang1Shengze Li2Jieyuan Zhang3Long Yan4Xinhai Xu5Chinese Academy of Military ScienceData Intelligence Division, China Unicom Digital Technology CoChinese Academy of Military ScienceChinese Academy of Military ScienceData Intelligence Division, China Unicom Digital Technology CoChinese Academy of Military ScienceAbstract The large action space is a key challenge in reinforcement learning. Although hierarchical methods have been proven to be effective in addressing this issue, they are not fully explored. This paper combines domain knowledge with hierarchical concepts to propose a novel Hierarchical Reinforcement Learning framework based on macro actions (HRL-MA). This framework includes a macro action mapping model that abstracts sequences of micro actions into macro actions, thereby simplifying the decision-making process. Macro actions are divided into two categories: combat macro actions (CMA) and non-combat macro actions (NO-CMA). NO-CMA are driven by decision tree-based logical rules and provide conditions for the execution of CMA. CMA form the action space of the reinforcement learning algorithm, which dynamically selects actions based on the current state. Comprehensive tests on the StarCraft II maps Simple64 and AbyssalReefLE demonstrate that the HRL-MA framework exhibits superior performance, achieving higher win rates compared to baseline algorithms. Furthermore, in mini-game scenarios, HRL-MA consistently outperforms baseline algorithms in terms of reward scores. The findings highlight the effectiveness of integrating hierarchical structures and macro actions in reinforcement learning to manage complex decision-making tasks in environments with large action spaces.https://doi.org/10.1007/s40747-025-01895-9Hierarchical reinforcement learningMacro action mapping modelCombat and non-combat macro actionsRule-based execution logic
spellingShingle Hao Jiang
Gongju Wang
Shengze Li
Jieyuan Zhang
Long Yan
Xinhai Xu
Hierarchical reinforcement learning based on macro actions
Complex & Intelligent Systems
Hierarchical reinforcement learning
Macro action mapping model
Combat and non-combat macro actions
Rule-based execution logic
title Hierarchical reinforcement learning based on macro actions
title_full Hierarchical reinforcement learning based on macro actions
title_fullStr Hierarchical reinforcement learning based on macro actions
title_full_unstemmed Hierarchical reinforcement learning based on macro actions
title_short Hierarchical reinforcement learning based on macro actions
title_sort hierarchical reinforcement learning based on macro actions
topic Hierarchical reinforcement learning
Macro action mapping model
Combat and non-combat macro actions
Rule-based execution logic
url https://doi.org/10.1007/s40747-025-01895-9
work_keys_str_mv AT haojiang hierarchicalreinforcementlearningbasedonmacroactions
AT gongjuwang hierarchicalreinforcementlearningbasedonmacroactions
AT shengzeli hierarchicalreinforcementlearningbasedonmacroactions
AT jieyuanzhang hierarchicalreinforcementlearningbasedonmacroactions
AT longyan hierarchicalreinforcementlearningbasedonmacroactions
AT xinhaixu hierarchicalreinforcementlearningbasedonmacroactions