Supervised machine learning technology has greatly improved the accuracy of radar target recognition based on HRRP signals. It relies on complete dataset labels, but the dataset is prone to noise labels due to the instability of data collection and the abstract nature of the HRRP signal itself. This problem will affect the model training robustness and testing accuracy. In this paper, we propose a noisy label learning method, "Contrastive Learning with or without Freeze, CLwF" to solve this problem. CLwF proposes a self-supervised learning algorithm to train models for efficient representation of HRRP signals. We also propose a clean rate estimation module to select the correct fine-tuning strategy for the model training with noisy labels. The experiment verified that CLwF can achieve excellent results under different noise rates.

In recent years, target recognition based on high-resolution range profile (HRRP) has been widely studied in the field of radar automatic target recognition. Compared with traditional methods, deep network model can automatically obtain the deep features of the target and is widely used, but expensive and time-consuming labeling is a difficult task. Inspired by the idea of contrastive learning, a self-supervised framework for learning target features based on double contrastive loss, namely contrastive learning with discriminant loss (CLDL) is proposed. Specifically, we minimize the absolute distance between positive sample pairs and maximize the absolute distance between negative sample pairs through constraints to ensure the proximity of two related samples and the discrimination of unrelated samples under the same HRRP signal. At the same time, the peak clipping strategy is proposed to preserve the useful information of the signal to the greatest extent during data augmentation. The experimental results show that the recognition performance of CLDL is better than the five mainstream unsupervised learning algorithms.

Radar sensor networks (RSNs) have more degrees of freedom than single radar system, and significantly improve the target angular resolution and parameter identifiability. This work proposes an intelligent waveform optimization strategy based on fuzzy Q learning (FQL) for multi-target tracking (MTT) of RSN in a cluttered environment. Specifically, this method adopts a distributed fusion architecture, each radar node independently detects and tracks multiple targets, and uses the covariance intersection (CI) fusion algorithm to solve the unknown correlation of each radar node. Combined with the target error covariance predicted by the Riccati equation, an FQL waveform optimization method is designed to select the best transmission from the waveform library. It integrates the radar and targets into a closed loop and updates the transmit waveform in real time according to the status of the targets to maximize global MTT performance. Simulation verifies that the proposed method’s tracking performance and CPU time are superior to the Q-learning.

The utilization of machine learning techniques has greatly improved the accuracy of micro-Doppler(m-D) signatures-based radar signal recognition. However, the "catastrophic forgetting" problem commonly exists in data-driven algorithms severely limits the adaptability of recognition algorithms in real-world applications, as models cannot incrementally train and learn new categories. In this paper, we propose an incremental learning method, "Boundary Transfer and Uncertainty Augmentation (BTUA)" for continuous learning of m-D signatures. BTUA utilizes boundary transfer(BT) to generate pseudo-decision boundaries for old categories and avoid the forgetting problem. It also employs an uncertainty augmentation(UA) algorithm to enhance the model’s generalization and improve the correctness of feature extraction. Finally, the validation demonstrates the advantages of our algorithm in terms of both accuracy and practicality.

Cognitive radar automatically adjusts its waveform via ceaseless interaction with the environment and learning from the experience. The waveform development of cognitive radar has been attracting much attention in improving tracking performance. In this paper, we propose an intelligent radar target tracking strategy based on variable nonlinear frequency modulated waveforms (NLFM). The strategy considers the combination of constant velocity (CV), constant acceleration (CA), and constant turning (CT) motion for high maneuvering targets. A library of NLFM is constructed and the entropy reward Q-Learning (ERQL) method is designed to perform joint waveform parameters selection. It merges the radar and target into a closed loop to provide the optimum target tracking performance, updating the waveform in real-time as the target state changes. Numerical results show that the tracking performance of our proposed method is much better than that of the linear frequency modulated waveform (LFM) pure parameter selection method.

Cognitive radar (CR) automatically improves itself via ceaseless interaction with the environment and learning from the experience. It continuously adjusts its waveform and parameters and illuminates strategies based on obtained knowledge to achieve robust target tracking despite complex and changing scenarios. Waveform development for CR has attracted sustaining attention in promoting tracking performance. In this article, we propose a novel framework of CR waveform selection for the tracking of high maneuvering targets in a cluttered environment with an interactive multimodel (IMM) probabilistic data association (PDA) algorithm. Based on this framework, criterion-based optimization (CBO) and entropy-rewarded Q-learning (ERQL) methods are designed to perform waveform selection, which is divided into pure parameters selection and joint selection of waveforms and parameters. This method integrates the radar target into a closed loop and realizes the real-time update of the transmitted waveform with the change of the target state, to achieve the best tracking performance of the target. The simulations performed on radar target tracking have demonstrated that the proposed ERQL method outperforms the existing method in both time complexity and tracking accuracy. Furthermore, field experiments have confirmed that the ERQL method is more effective in target tracking than the existing method.