Scholarlink

This study addresses the challenges of low efficiency and environmental adaptability in airport taxi scheduling by proposing a Multi-Agent Deep Deterministic Policy Gradient (MADDPG) framework. Using operational data from Xi'an Xianyang International Airport, we developed an environmental model incorporating aircraft status parameters, explicitly defining state-action spaces and designing an adaptive reward mechanism. A two-phase validation framework involving simulations and field tests demonstrated that our MADDPG-based solution achieves 97.20% scheduling efficiency, representing a 4.38% improvement over Multi-agent Proximal Policy Optimization (MAPPO) benchmarks. The algorithm exhibits robust performance across diverse airport configurations, effectively handling environmental complexities while enhancing operational throughput and taxiing safety. These findings highlight MADDPG's superior capability in dynamic resource coordination for aviation ground operations compared to conventional reinforcement learning approaches.

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

Intelligent Scheduling Strategies for Airport Aircraft Taxiing Based on Multi-agent Reinforcement Learning

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Man Zhang,
  • Zhi Wang

摘要

This study addresses the challenges of low efficiency and environmental adaptability in airport taxi scheduling by proposing a Multi-Agent Deep Deterministic Policy Gradient (MADDPG) framework. Using operational data from Xi'an Xianyang International Airport, we developed an environmental model incorporating aircraft status parameters, explicitly defining state-action spaces and designing an adaptive reward mechanism. A two-phase validation framework involving simulations and field tests demonstrated that our MADDPG-based solution achieves 97.20% scheduling efficiency, representing a 4.38% improvement over Multi-agent Proximal Policy Optimization (MAPPO) benchmarks. The algorithm exhibits robust performance across diverse airport configurations, effectively handling environmental complexities while enhancing operational throughput and taxiing safety. These findings highlight MADDPG's superior capability in dynamic resource coordination for aviation ground operations compared to conventional reinforcement learning approaches.