Abstract
Orthogonal time frequency space (OTFS) modulation technology which provides reliable communication and precise sensing in high-mobility scenarios, has emerged as a potential solution for various unmanned aerial vehicle (UAV)-related applications. In this article, we consider an OTFS communication waveform-based UAV sensing situation. Due to random wind gusts and varying weather conditions, the sensing signals may experience sudden disturbances. To effectively address this challenge, we propose a deep-learning (DL)-based framework to compensate the impulse interference, which leverages empirical information and generates real-time predictions to achieve accurate UAV sensing. Specifically, we develop a prediction-assisted estimation network (PAEnet) to implement the proposed framework. The core component of PAEnet, the estimation network (ESnet), is capable to directly extract fractional delay and Doppler from the transmitted OTFS frame, thereby reducing the complexity of the sensing process. Through comprehensive simulation results, we demonstrate the effectiveness of the compensation mechanism in unreliable sensing scenarios, while showcasing the PAEnet's capability to achieve superior accuracy for OTFS-based UAV sensing.