UAV-Assisted Search of Emitter with Dynamic Beam: A Reinforcement Learning-Based Method

Search and location of terrestrial emitters, especially those using spectrum illegally/improperly, can facilitate the utilization of spectrum resources and mitigate interference. However, the dynamic directional RF beams make search and geolocation of terrestrial terminals (such as the terminals for low-earth orbit satellite internet) intractable. This paper investigates the issue of searching and geolocating a radio frequency (RF) emitter with an unmanned aerial vehicle (UAV). To handle this issue, a novel UAV autonomous trajectory planning scheme is proposed by employing reinforcement learning (RL) and target probability map (TPM). Based on the detection probability and false alarm probability, a TPM is constructed to assist UAV to perform effective trajectory planning in the absence of knowledge about the uplink beam direction and the potential geographical location of the RF emitter. By employing the TPM information as a part of the state information, a deep-Q-network-based trajectory planing algorithm is proposed for searching. Simulation results confirm the advantages of the proposed algorithm over various baselines in time consumption and geolocation accuracy.