Scholarlink

<p>Cross-border logistics planning faces challenges such as high planning complexity, conflicting constraints, and strong uncertainty. This paper addresses these challenges by proposing a model that combines genetic methods with reinforcement learning. This paper first models multi-objective cross-border logistics planning, emphasizing four key objectives: time cost, economic cost, environmental cost, and robustness. Logistics paths and nodes are modeled using a graph structure. A genetic algorithm (GA) is used to generate high-quality solutions for reinforcement learning. The policy network in the reinforcement learning framework then interacts with the environment to obtain rewards for implementation. Evaluation experiments on public datasets demonstrate that the proposed model demonstrates superior cost-effectiveness and accuracy compared to simulated annealing particle swarm optimization (SAP) and genetic algorithms. Compared to several other machine-based approaches, the proposed IGA-MORL algorithm reduces average transportation time by 28.8%, economic costs by 26.7%, and carbon emissions by 29.2%. The proposed model also exhibits the lowest robustness and fastest convergence speed. This model provides an effective solution for multi-objective cross-border logistics planning.</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

Cross-border logistics route planning method optimized by reinforcement learning and genetic algorithm

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年2月9日

查看全文
  • Fengwei Zhang,
  • Nana Liu

摘要

Cross-border logistics planning faces challenges such as high planning complexity, conflicting constraints, and strong uncertainty. This paper addresses these challenges by proposing a model that combines genetic methods with reinforcement learning. This paper first models multi-objective cross-border logistics planning, emphasizing four key objectives: time cost, economic cost, environmental cost, and robustness. Logistics paths and nodes are modeled using a graph structure. A genetic algorithm (GA) is used to generate high-quality solutions for reinforcement learning. The policy network in the reinforcement learning framework then interacts with the environment to obtain rewards for implementation. Evaluation experiments on public datasets demonstrate that the proposed model demonstrates superior cost-effectiveness and accuracy compared to simulated annealing particle swarm optimization (SAP) and genetic algorithms. Compared to several other machine-based approaches, the proposed IGA-MORL algorithm reduces average transportation time by 28.8%, economic costs by 26.7%, and carbon emissions by 29.2%. The proposed model also exhibits the lowest robustness and fastest convergence speed. This model provides an effective solution for multi-objective cross-border logistics planning.