Scholarlink

<p>The pavement maintenance decision algorithm has emerged as a key research focus in the field of pavement management. Significant challenges remain in optimization algorithms designed to generate efficient maintenance plans. This paper analyzes the performance degradation patterns of pavements and the effectiveness of maintenance interventions, and proposes a reinforcement learning (RL)-based optimization algorithm for pavement maintenance decision-making. Specifically, a performance decay model is established to quantify the deterioration of pavement performance and capture the performance dispersion of concrete slabs, along with a maintenance effect analysis model. By integrating these two models, an operational environment for a reinforcement learning framework is constructed, within which the reinforcement learning algorithm is employed to optimize pavement maintenance plans. The experiment results show that reinforcement learning algorithm achieves a cost savings of up to 20% compared with the genetic algorithm, and the maintenance efficiency of medium or heavy repairs is improved by up to 28%. The reinforcement learning-based maintenance decision model is capable of generating more cost-effective maintenance plans and reducing overall pavement operational costs.</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

Maintenance decision-making model of pavement based on reinforcement learning

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年5月25日

查看全文
  • Yajun Huang,
  • Jin Wu,
  • Pengchao You,
  • Haiwei Sun,
  • Xiang Fang

摘要

The pavement maintenance decision algorithm has emerged as a key research focus in the field of pavement management. Significant challenges remain in optimization algorithms designed to generate efficient maintenance plans. This paper analyzes the performance degradation patterns of pavements and the effectiveness of maintenance interventions, and proposes a reinforcement learning (RL)-based optimization algorithm for pavement maintenance decision-making. Specifically, a performance decay model is established to quantify the deterioration of pavement performance and capture the performance dispersion of concrete slabs, along with a maintenance effect analysis model. By integrating these two models, an operational environment for a reinforcement learning framework is constructed, within which the reinforcement learning algorithm is employed to optimize pavement maintenance plans. The experiment results show that reinforcement learning algorithm achieves a cost savings of up to 20% compared with the genetic algorithm, and the maintenance efficiency of medium or heavy repairs is improved by up to 28%. The reinforcement learning-based maintenance decision model is capable of generating more cost-effective maintenance plans and reducing overall pavement operational costs.