TY - GEN
T1 - Multi-granularity penetration strategy optimization algorithm based on PER-SAC algorithm
AU - Zhang, Ruixin
AU - Qin, Jiahao
AU - Jia, Yubo
AU - Li, Yan
N1 - Publisher Copyright:
© 2025 SPIE.
PY - 2025/9/8
Y1 - 2025/9/8
N2 - With the development of artificial intelligence, intelligent algorithms have been applied to electronic countermeasures. In the typical penetration scenario, due to the complex working modes of current radars and advanced digital signal processing technology, traditional jamming methods cannot adaptively manage the jamming strategy. In this paper, a multi-granularity penetration strategy optimization method based on deep reinforcement learning is proposed. We model the penetration scenario using the Markov decision process, and design the reward function with the jamming signal ratio in order to jointly optimize the jamming power and penetration trajectory. We develop the PER-SAC algorithm, which is based on priority experience replay mechanism, to efficiently utilize experience samples and effectively optimize penetration strategy. Finally, the simulation results show the superiority of the PER-SAC in jamming success rate and learning speed.
AB - With the development of artificial intelligence, intelligent algorithms have been applied to electronic countermeasures. In the typical penetration scenario, due to the complex working modes of current radars and advanced digital signal processing technology, traditional jamming methods cannot adaptively manage the jamming strategy. In this paper, a multi-granularity penetration strategy optimization method based on deep reinforcement learning is proposed. We model the penetration scenario using the Markov decision process, and design the reward function with the jamming signal ratio in order to jointly optimize the jamming power and penetration trajectory. We develop the PER-SAC algorithm, which is based on priority experience replay mechanism, to efficiently utilize experience samples and effectively optimize penetration strategy. Finally, the simulation results show the superiority of the PER-SAC in jamming success rate and learning speed.
KW - Deep reinforcement learning
KW - Penetration strategy optimization
KW - Radar game confrontation
UR - https://www.scopus.com/pages/publications/105022285437
U2 - 10.1117/12.3077174
DO - 10.1117/12.3077174
M3 - Conference contribution
AN - SCOPUS:105022285437
T3 - Proceedings of SPIE - The International Society for Optical Engineering
BT - Fifth International Conference on Signal Image Processing and Communication, ICSIPC 2025
A2 - Feng, Shou
A2 - Zhang, Zhihao
PB - SPIE
T2 - 5th International Conference on Signal Image Processing and Communication, ICSIPC 2025
Y2 - 16 May 2025 through 18 May 2025
ER -