Learning-Based UAV Trajectory Optimization With Collision Avoidance and Connectivity Constraints

Unmanned aerial vehicles (UAVs) are expected to be an integral part of wireless networks, and determining collision-free trajectories for multiple UAVs while satisfying requirements connectivity with ground base stations (GBSs) is a challenging task. In this paper, we consider non-cooperative multi-UAV scenarios, in which need fly from initial locations destinations, collision avoidance, connectivity, kinematic constraints. We aim find the goal minimize their mission completion time. first formulate trajectory optimization problem as sequential decision making problem. We, then, propose decentralized deep reinforcement learning approach solve More specifically, value network developed obtain values given agent’s joint state (including information, nearby agents’ observable GBSs). A signal-to-interference-plus-noise ratio (SINR)-prediction neural also designed, using accumulated SINR measurements obtained when interacting cellular network, map GBSs’ into levels order predict UAV’s SINR. Numerical results show that SINR-prediction real-time navigation multi-UAVs can efficiently performed various environments high success rate.