Scholarlink

Efficient ambulance dispatching is critical for minimizing patient response time and improving survival outcomes in emergency medical services (EMS). This paper presents a deep reinforcement learning-based framework for intelligent ambulance allocation and routing in urban environments. The dispatch problem is modeled as a Markov Decision Process (MDP), where states incorporate patient severity, location, and traffic conditions. We implement and evaluate both Deep Q-Network (DQN) and Rainbow algorithms within a custom Gym environment, SmartAmbulanceEnv, designed to simulate realistic EMS scenarios. Experimental results demonstrate that Rainbow DQN significantly outperforms standard DQN in terms of reward convergence, decision stability, and average patient response time.

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

Optimization of Smart Ambulance Dispatching Based on Markov Decision Process and Deep Reinforcement Learning

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Hibat Eallah Mohtadi,
  • Abdellah Ouammou,
  • Mohamed Hanini

摘要

Efficient ambulance dispatching is critical for minimizing patient response time and improving survival outcomes in emergency medical services (EMS). This paper presents a deep reinforcement learning-based framework for intelligent ambulance allocation and routing in urban environments. The dispatch problem is modeled as a Markov Decision Process (MDP), where states incorporate patient severity, location, and traffic conditions. We implement and evaluate both Deep Q-Network (DQN) and Rainbow algorithms within a custom Gym environment, SmartAmbulanceEnv, designed to simulate realistic EMS scenarios. Experimental results demonstrate that Rainbow DQN significantly outperforms standard DQN in terms of reward convergence, decision stability, and average patient response time.