The aim is to enable multi-UAV ( Unmanned Aerial Vehicle) swarms to achieve comprehensive optimization of communication performance, task efficiency, and flight safety under specific network conditions,thereby better conducting patrol missions in urban areas. Based on double deep reinforcement learning technology, the spatial discretization processing on the airspace with known communication quality distribution is studied, spatial models, energy consumption models and communication models are established. A multi-dimensional reward function including data acquisition, flight safety, remaining power and path consumption is designed, and the training process is established through experience replay and target network mechanisms.Experiments show that the trained network model can generate optimal wireless network transmission strategies and safe flight trajectories in unforeseen environments. The research effectively solves the route planning problem under multi-objective constraints and verifies the applicability of double deep reinforcement learning in this field.

route planning, reinforcement learning, multi-unmanned aerial vehicle(UAV) systems