Adaptive Radar Control Under Latent-Type Jamming via Game-Informed Reinforcement Learning
IEEE Access · 2026 · 14 · 102425-102453
DOI: 10.1109/access.2026.3709945 ↗Kurz gefasst
Fachartikel in der Zeitschrift IEEE Access, 2026. Verfasst von Abdulmuneem Alselwi und 5 weiteren Autor:innen. Die Arbeit nennt 53 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).
Abstract (Originalsprache)
Reliable radar operation in contested electromagnetic environments requires adaptive control under partial observability and adversarial interference. Classical game-theoretic methods provide interpretable strategic structure but may become brittle under model mismatch, whereas model-free reinforcement learning (RL) adapts through interaction but may learn inefficiently and generalize poorly under adversarial distribution shifts. We propose Game-Informed Reinforcement Learning (GIRL), a hybrid framework for adaptive radar control against latent-type jammers. GIRL integrates offline type-conditioned strategic priors derived from a reduced Stackelberg-style radar–jammer game, online belief inference over jammer type from radar-available features, and Kullback–Leibler (KL)-regularized proximal policy optimization that softly anchors policy updates toward the inferred-type prior. We evaluate GIRL in a physically motivated simulator that couples signal-to-interference-plus-noise ratio (SINR), probability of detection, tracking error, and loss-of-track events. Within the evaluated simulator regimes, GIRL improves sample efficiency, learning stability, and sensing and tracking performance relative to PPO, PPO+Belief, and DQN; achieves stronger average-performance trends than domain-randomized PPO and fictitious-play-style PPO under held-out deceptive and hybrid-switching scenarios; and degrades gracefully under prior misspecification. The results support GIRL as a structured learning framework for simulator-based adaptive radar control under latent-type jamming. They also highlight an important reliability challenge: under severe deceptive interference, track continuity can fail despite moderate average sensing metrics, motivating tail-aware evaluation and future risk-sensitive control objectives.
Zitieren
APA
Alselwi, A., Khaled, O., Khan, A. U., Abdullah, K., Abdulhadi, H., & Abdulhadi, H. (2026). Adaptive Radar Control Under Latent-Type Jamming via Game-Informed Reinforcement Learning. IEEE Access, 14, 102425-102453. https://doi.org/10.1109/access.2026.3709945
Vancouver
Alselwi A, Khaled O, Khan AU, Abdullah K, Abdulhadi H, Abdulhadi H. Adaptive Radar Control Under Latent-Type Jamming via Game-Informed Reinforcement Learning. IEEE Access. 2026;14:102425-102453. doi:10.1109/access.2026.3709945
BibTeX
@article{Alselwi2026,
author = {Alselwi, Abdulmuneem and Khaled, Osamah and Khan, Adnan Umar and Abdullah, Khaled and Abdulhadi, Hussein and Abdulhadi, Hasan},
title = {Adaptive Radar Control Under Latent-Type Jamming via Game-Informed Reinforcement Learning},
journal = {IEEE Access},
year = {2026},
volume = {14},
pages = {102425--102453},
doi = {10.1109/access.2026.3709945}
}
EUVIMED