交通运输工程与信息学报2025，Vol.23Issue(2)：189-206,18.DOI:10.19961/j.cnki.1672-4747.2024.12.007
基于强化学习的城市低空无人机路径规划
Urban low-altitude UAV path planning based on reinforcement learning
摘要
Abstract
[Background]Owing to the emergence of low-altitude economies,low-altitude airspace has become increasingly pivotal for the revolution of urban transportation.Unmanned aerial vehicles(UAVs)and other novel transportation modes are being used increasingly in urban spaces;however,they are affected by challenges such as path planning,dynamic obstacle avoidance,and multi-UAV coordination.[Objective]To optimize UAV path-planning algorithms,enhance the path-planning and obstacle-avoidance capabilities of UAVs in complex low-altitude environments,and ensure the efficiency and safety of multi-UAV collaborative operations.[Methods]An NP-MTDDQN algo-rithm based on reinforcement learning,which incorporates an N-step update strategy and an im-proved prioritized experience replay mechanism,is proposed.A three-dimensional low-altitude envi-ronment featuring diverse buildings and dynamic obstacles is constructed using wind-flow data from Shizhong District,Jinan City,Shandong Province,to conduct simulation experiments and validate the effectiveness of the algorithm.[Results]Three sets of controlled experiments indicate that the NP-MTDDQN algorithm can identify optimal paths,distinguish building priorities,as well as identify and avoid dynamic obstacles in complex environments with varying building densities,different pri-ority building distributions,and dynamic obstacles,thereby facilitating multi-UAV collaboration.Compared with the conventional DQN and DDQN algorithms,the NP-MTDDQN algorithm exhibits improved efficiency and accuracy in path planning.[Conclusions]The NP-MTDDQN algorithm pro-vides a novel solution to the path-planning problem for multi-UAVs in low-altitude intelligent trans-portation networks,thus potentially enhancing the management efficiency and response speed of ur-ban air traffic.关键词
智能交通/路径规划/无人机/深度强化学习/低空经济Key words
intelligent transportation/path planning/unmanned aerial vehicles/deep reinforcement learning/low-altitude economy分类
土木建筑引用本文复制引用
丁杰,王迪..基于强化学习的城市低空无人机路径规划[J].交通运输工程与信息学报,2025,23(2):189-206,18.基金项目
山东省自然科学基金项目(ZR2024QG174) （ZR2024QG174）
山东省城市更新学会重点研究专项项目(SURS240603) （SURS240603）
