BRR-DQN: UAV path planning method for urban remote sensing images
Kexin Zheng, Xiaobo Liu, Jianfeng Yang, Zhihua Cai, Haoran Dai, Zhilang Zhou, Xiao Xiao, Xin Gong · 2021 China Automation Congress (CAC) · 2021
Unmanned aerial vehicle(UAV) urban path planning is a research hotspot in the construction of modern smart cities. We propose a novel DQN combine BRRNet which named BRR-DQN that can directly plan the urban remote sensing image path. First, the BRRNet is designed to model the environmental map, which provides a new solution for the environmental map modeling problem in the path planning field. Second, the reward function is introduced, we mainly optimizes the reward function of the DQN from two aspects: (a) Setting obstacles in the nearby dangerous area; (b) Designing the target area near the target point. Through the above optimization methods make up for the lack of building information extracted by BRRNet, the risk of UAV hitting the building is reduced, and the model convergence speed is accelerated. At the effiency of our method is demonstrated by urban remote sensing image.