【摘要】 大型公共建筑物结构和设计复杂，在火灾应急疏散中存在人群因火势扩散找不到有效疏散路径的难题。基于两种深度Q网络(DQN)算法，针对不同年龄段人员的紧急疏散速度，在烟雾扩散影响下的火灾仿真环境中寻找不同被困人员的有效疏散引导路径，从而得到不同被困人员在大型建筑物室内的有效疏散方案。实验结果表明：(1)DQN算法在火灾环境中的应用能更高效地获得低成本的室内火灾最优引导疏散路径；(2)Dueling DQN算法搜索最优疏散路径成功率和安全性高于DQN算法，更适合火灾最优引导疏散路径规划；(3)火灾最优疏散路径规划中，应该适当考虑不同年龄疏散人群的紧急疏散速度，为不同年龄群体提供合适的火灾最优疏散路径引导。更多还原

【Abstract】 Due to the complex structure and design of large public buildings, emergency evacuation becomes challenging for people in finding effective evacuation paths because of the spread of fire.This study utilizes two deep Q-network(DQN) algorithms to identify effective evacuation guidance paths for people trapped in a fire simulation environment, considering the emergency evacuation speeds of people from different age groups under different levels of smoke diffusion, so as to obtain effective evacuation schemes for people trapped in large buildings.The experimental results show that:(1)The application of the DQN algorithm in a fire environment can efficiently generate low-cost optimal guided evacuation paths for indoor fires;(2)The success rate and safety of Dueling DQN algorithm in finding optimal evacuation paths are higher than that of the DQN algorithm, making it more suitable for optimal fire evacuation guidance path planning;(3)In the planning of the optimal fire evacuation path, the emergency evacuation speeds of different age groups should be properly considered, so as to provide appropriate guidance on optimal fire evacuation paths for each group.更多还原