To address the problem of online generating transmit waveforms with both high precision and high stability for maneuvering extended targets, which requires real-time motion and shape state information obtained through interactive sensing, an intelligent waveform design method based on an Actor-Critic (AC) network is proposed. Firstly, Under the interactive multiple model (IMM) framework, the unscented Kalman filter (UKF) and joint probabilistic data association (JPDA) are integrated to achieve joint estimation of target motion and shape states. Secondly, a multi‑component reward function centered on the Gaussian Wasserstein distance (GWD) is constructed. Finally, the AC network is driven to interact with the environment in real time, enabling online waveform optimization design. The simulation results demonstrate that, compared with the mutual information (MI) criterion, fixed waveforms, and the deep Q‑network (DQN), the proposed method exhibits significant advantages in target shape estimation accuracy, tracking stability, and waveform parameter selection, verifying its effectiveness in high‑precision extended target tracking and highlighting its potential for practical engineering applications.
ZHUP K, LIANGJ, LUOZ H, et al. Cognitive radar target tracking using intelligent waveforms based on reinforcement learning[J]. IEEE Transactions on Geoscience and Remote Sensing, 2023,61:No.5107315.
[8]
熊芳勇.认知雷达导引头波形选择技术研究[D].郑州:信息工程大学,2023:1-72.
[9]
KOCHJ W. Bayesian approach to extended object and cluster tracking using random matrices[J]. IEEE Transactions on Aerospace and Electronic Systems, 2008,44(3):1042-1059.
[10]
DINGW J, ZHANGW W, ZHANGM, et al. Networked radar waveform design for detecting extended target in the presence of jamming[J]. Physical Communication, 2025,68:No.102539.
[11]
王文杰.面向认知雷达目标检测的波形设计研究[D].西安:西安电子科技大学,2023:1-90.
[12]
ZHANGH Y, CHENH, ZHANGW X, et al. Trajectory planning for airborne radar in extended target tracking based on deep reinforcement learning[J]. Digital Signal Processing, 2024,153:No.104603.
[13]
YANGZ R, CHENY X, HONGM Y, et al. Provably global convergence of actor-critic: a case for linear quadratic regulator with ergodic cost[C]//Proceedings of the 33rd International Conference on Neural Information Processing Systems. Red Hook, USA: Curran Associates In c., 2019:8353-8365.