Predictive Modeling of Frequency Hopping Signals via Deep Reinforcement Learning

  • Tuong, Van Dat; 
  • Nguyen-Trong, Khanh; 
  • Dao, Nhu-Ngoc; 
  • Cho, Sungrae
Citations

WEB OF SCIENCE

1
Citations

SCOPUS

1

초록

With the rapid expansion of legal and illegal activities using unmanned aerial vehicles (UAVs), the UAV incidents in the unarmed population are gradually increasing. To prevent the damage from such UAV incidents for the public facility, jamming over the communication between UAVs and their remote control is a cost-effective solution. However, modern UAVs are enabled to inherently reduce susceptibility to interference. Specifically, they are installed with frequency hopping signals that make a huge challenge for efficient jamming owing to the narrowband transmissions and rapid frequency changes. Therefore, it is necessary to develop an agile model that precisely and rapidly predicts parameters of frequency hopping signals to improve the jamming efficiency. This study presents a deep reinforcement learning framework capable of predicting future time-frequency locations of the signals. In particular, a deep deterministic policy gradient (DDPG) algorithm is developed and trained on empirical signal data collected from drone DJI Mavic 3 Pro. A novel reward function is introduced that evaluates prediction accuracy by combining the results of time and frequency predictions. Training results indicate a high fidelity of signal forecasting, achieving an excellent mean score that significantly converges with low standard deviation.

키워드

Frequency hopping; reinforcement learning; unmanned aerial vehicles
제목
Predictive Modeling of Frequency Hopping Signals via Deep Reinforcement Learning
저자
Tuong, Van Dat; Nguyen-Trong, Khanh; Dao, Nhu-Ngoc; Cho, Sungrae
DOI
10.1109/ICOIN68469.2026.11480657
발행일
2026
유형
Proceedings Paper
저널명
International Conference on Information Networking
페이지
297 ~ 302