Scholarlink

<p>In the realm of Automated Guided Vehicle (AGV) systems, training Deep Reinforcement Learning (DRL) models presents significant challenges due to the complexity of tasks and environments, as well as the large state and action spaces involved. This paper introduces the Gradual Task Complexity Scaling (GTCS) approach, a novel learning procedure for DRL that effectively addresses these challenges. Unlike existing methods, which focus on directly achieving the final objective, GTCS incrementally upgrades the agent’s objectives while maintaining the same environmental context, enabling a more efficient balance between exploration and exploitation during training. The GTCS procedure features four key components: gradually expanding the reduced effective space size within the warehouse, increasing the number of products the agent must deliver, enhancing the capabilities of AGVs represented as DRL agents, and reducing the maximum number of steps allowed for task completion. GTCS outperforms previous approaches by improving the stability of the learning process, optimizing the delivery workflow, and achieving more efficient learning outcomes.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Gradual task complexity scaling (GTCS-DRL): a deep reinforcement learning approach for training automated guided vehicle system

  • 期刊论文

  • 发表日期：

    2025年2月20日

查看全文
  • Mohamed Rhazzaf,
  • Tawfik Masrour

摘要

In the realm of Automated Guided Vehicle (AGV) systems, training Deep Reinforcement Learning (DRL) models presents significant challenges due to the complexity of tasks and environments, as well as the large state and action spaces involved. This paper introduces the Gradual Task Complexity Scaling (GTCS) approach, a novel learning procedure for DRL that effectively addresses these challenges. Unlike existing methods, which focus on directly achieving the final objective, GTCS incrementally upgrades the agent’s objectives while maintaining the same environmental context, enabling a more efficient balance between exploration and exploitation during training. The GTCS procedure features four key components: gradually expanding the reduced effective space size within the warehouse, increasing the number of products the agent must deliver, enhancing the capabilities of AGVs represented as DRL agents, and reducing the maximum number of steps allowed for task completion. GTCS outperforms previous approaches by improving the stability of the learning process, optimizing the delivery workflow, and achieving more efficient learning outcomes.