Abstract
In this paper, we consider a formation control problem for leader-follower unmanned aerial vehicles (UAVs) in a GPS-denied environment. The distance and the azimuth and elevation angles, defined in a local spherical coordinate frame, are used to describe the relative motion between two UAVs. A novel deep reinforcement learning (DRL) technique is leveraged to generate the required control policies that maneuver a follower UAV in a desired formation with respect to the leader. The effectiveness of the proposed DRL-based leader-follower formation is demonstrated in a simulated environment.