TE-DDPG: Temporal-Enhanced Deep Reinforcement Learning for Power Allocation in Cognitive Radar Networks
IEEE Transactions on Aerospace and Electronic Systems · 2026 · 62 · 13407-13422
DOI: 10.1109/taes.2026.3709256 ↗Kurz gefasst
Fachartikel in der Zeitschrift IEEE Transactions on Aerospace and Electronic Systems, 2026. Verfasst von Jiangtao Liu und 4 weiteren Autor:innen. Die Arbeit nennt 36 Literaturangaben. Gefördert von: National Natural Science Foundation of China, Zhenjiang Science and Technology Program, Key Research and Development Project of Henan Province.
Abstract (Originalsprache)
Cognitive radar networks have emerged as a transformative paradigm for collaborative target tracking, offering adaptive sensing and intelligent resource allocation in increasingly complex operational environments. This article investigates the problem of transmit power allocation for multitarget tracking and proposes a temporal-enhanced deep deterministic policy gradient (TE-DDPG) algorithm, a model-informed deep reinforcement learning framework that balances tracking accuracy with computational efficiency. Unlike conventional heuristic or iterative optimization methods that incur prohibitive online overhead, TE-DDPG leverages a data-driven policy to enable real-time decision-making while preserving tracking performance. We first derive the posterior Cramér–Rao lower bound for multiradar cooperative tracking, establishing a closed-form relationship between power allocation and estimation accuracy. To address the high-dimensional nature of dynamic resource management, we integrate a temporal convolutional network into the DDPG architecture as a structured feature extractor, enabling the agent to capture temporal dependencies and state interactions inherent in extended Kalman filter outputs. The experimental results demonstrate that the proposed approach significantly improves the tracking accuracy of the most challenging target, achieving performance close to the theoretical optimum. Furthermore, TE-DDPG exhibits superior real-time adaptability, strong generalization across diverse motion patterns, and seamless transferability—making it a promising solution for next-generation cognitive radar systems.
Förderung
- National Natural Science Foundation of China
- Zhenjiang Science and Technology Program
- Key Research and Development Project of Henan Province
Zitieren
APA
Liu, J., Liu, C., Jin, B., Zhang, Z., & Su, T. (2026). TE-DDPG: Temporal-Enhanced Deep Reinforcement Learning for Power Allocation in Cognitive Radar Networks. IEEE Transactions on Aerospace and Electronic Systems, 62, 13407-13422. https://doi.org/10.1109/taes.2026.3709256
Vancouver
Liu J, Liu C, Jin B, Zhang Z, Su T. TE-DDPG: Temporal-Enhanced Deep Reinforcement Learning for Power Allocation in Cognitive Radar Networks. IEEE Transactions on Aerospace and Electronic Systems. 2026;62:13407-13422. doi:10.1109/taes.2026.3709256
BibTeX
@article{Liu2026,
author = {Liu, Jiangtao and Liu, Chao and Jin, Biao and Zhang, Zhenkai and Su, Tao},
title = {TE-DDPG: Temporal-Enhanced Deep Reinforcement Learning for Power Allocation in Cognitive Radar Networks},
journal = {IEEE Transactions on Aerospace and Electronic Systems},
year = {2026},
volume = {62},
pages = {13407--13422},
doi = {10.1109/taes.2026.3709256}
}
EUVIMED