28 January 2025 in IEEE Transactions on Aerospace and Electronic Systems
In this paper, high-speed target detection in homogeneous clutter is investigated, where both blind speed sidelobe (BSSL) suppression and constant false alarm rate (CFAR) property are considered. The Radon-Fourier transform (RFT) and adaptive Radon-Fourier transform (ARFT) have been proposed for high-speed target detection in noise and clutter background, respectively, which can perform long-time coherent integration (LTCI) for high-speed target and therefore have excellent detectability potential. However, they do not possess the CFAR property and suffer from the BSSL problem, leading to the degradation of detection performance. To address these two problems, the iterative generalized adaptive matched filter (IGAMF) is proposed in this paper, which achieves target CFAR detection and BSSL suppression by detecting the targets one by one and eliminating the interference from the detected targets. The simulation results demonstrate that the IGAMF has a higher detection probability compared to RFT and ARFT, and can effectively address the BSSL problem.
27 May 2025 in IEEE Transactions on Aerospace and Electronic Systems
For the neural network (NN) detector in marine radar, its superior performance relies heavily on correctly annotated large datasets bringing expensive cost. The Automatic Identification System (AIS) can be used to automatically construct datasets by obtaining the location of other ships. However, the constructed dataset inevitably contains a large number of noisy labels, which can seriously affect the performance of the NN. Therefore, we propose a robust NN training method for marine radar ship target detection using noisy AIS labels. First, we propose the radar assisted AIS label construction strategy to automatically construct datasets with noisy AIS labels. Second, we analyze and find that the differentiable Neyman Pearson (DNP) loss is robust to noisy AIS labels due to its relaxing effect on incorrect labels and reinforcing effect on correct labels, and propose a dynamic adjustment strategy to make the DNP loss robust to noisy AIS labels with an unknown number of error labels. Third, to effectively train the NN detector with noisy AIS labels and the DNP loss, we propose the Logit-Peak-Distillation network training strategy, where we use the NN logit value processing strategy to quickly switch to lower probability of false alarm (PFA) by avoiding the information loss caused by the sigmoid activation function, use the peak detection strategy to accelerate the NN training by efficiently computing the PFA, and use the ensemble knowledge distillation strategy to reduce the effect of error labels by cleaning the noisy AIS labels into cleaner soft labels. In the experiment, compared with the benchmark methods, our proposed method has the better detection performance on the measured data of marine radar.
20 September 2024 in IEEE Transactions on Aerospace and Electronic Systems
Compared with the constant false alarm ratio (CFAR) detector, the dynamic programming algorithm for track-before-detect (DP-TBD) can utilize target motion information in large-scale interframe time to achieve better performance advantages in detecting targets with low signal-to-clutter ratio (SCR). To reduce computational complexity and minimize clutter interference, the DP-TBD algorithm directly processes the plot-list data, which is obtained by performing threshold preprocessing (or the CFAR detector) on the raw radar data. However, it is difficult to provide theoretical guidance on the detection threshold of the algorithm. Therefore, for the DP-TBD algorithm in processing the plot-list data, we theoretically derive and propose a numerical method to determine the detection threshold. First, we theoretically derive the probability density function (PDF) of the clutter merit function (MF), thereby providing a formula for calculating the probability of false alarm (PFA) of DP-TBD. Second, to mitigate the model error caused by the independent and identical distribution (IID) assumption, we propose a modified model with a coefficient to be determined from real data. Third, based on the theoretical formula, we propose a numerical calculation method implemented by Fast Fourier Transform (FFT) to calculate the theoretical detection threshold. The experimental results demonstrate that the proposed method can effectively and accurately describe the PFA of DP-TBD for different simulation conditions in a uniform clutter environment.
1 November 2024 in IEEE Transactions on Aerospace and Electronic Systems
Waveform-agile tracking performance of cognitive tracking radar (CTR) is significantly affected by the target maneuverability and increasing radio frequency interference (RFI) of other in-band systems. Therefore, a joint spectrum management and waveform optimization (JSMWO) policy is proposed to improve the maneuvering target tracking performance while mitigating RFI. Firstly, the JSMWO policy is mathematically formulated to simultaneously optimize the spectrum occupancy and waveform parameters by minimizing a JSMWO-oriented three-term objective function. Specifically, the first two terms consider the trade-off between interference avoidance and bandwidth utilization, and the impact of current spectrum occupancy action on future multistep rewards. The last sub-reward term incorporates the criterion-specific conditional predictive Bayesian risk, waveform library limitation, and real-time spectrum constraint generated by spectrum occupancy action. Secondly, to resolve the underlying mixed-integer, nonlinear, non-convex multivariable optimization problem, by proper coupling analysis, we develop a four-stage iterative method to decompose the original JSMWO policy into two sub-policies and tackle them sequentially, which encompasses the deep-Q-learning (DQL)-based spectrum occupancy optimization, particle swarm optimization (PSO)-based waveform optimization and feedback information-based online learning. Specifically, the dual-network architecture and experience replay mechanism equipped in DQL can effectively address the delay reward and future multistep effects during the spectrum occupancy sub-policy optimization. Waveform optimization sub-policy is subsequently reformulated as an equivalent constrained optimization problem, wherein the PSO can effectively optimize the waveform parameters through inter-particle interactions and iterations while satisfying the spectrum constraint and waveform limitation. Simulation results demonstrate the effectiveness and superiority of the above JSMWO policy.
4 April 2024 in IEEE Transactions on Aerospace and Electronic Systems
The increasing radio frequency interference (RFI) caused by other in-band radio frequency (RF) users, like communication systems, affects the waveform-agile tracking performance of cognitive tracking radar (CTR). To improve tracking performance while mitigating the RFI, both the spectrum occupancy and waveform parameters of CTR need to be jointly optimized based on real-time perceptual information. Therefore, we formulate the above tracking waveform optimization problem in the congested spectrum environment as waveform optimization for joint tracking and anti-interference (WO-JTAI) and propose a hybrid stochastic-deterministic waveform optimization (HSD-WO) method for WO-JTAI based on the JTAI-oriented multi-step objective function. Firstly, to establish the coupling relationship between spectrum occupancy, waveform parameters and the potential long-term performance of joint tracking and anti-interference, we propose a JTAI-oriented multi-step objective function based on the WO-JTAI-oriented partial observable Markov decision process (POMDP) model. This specific POMDP model can deal with the imperfect state information in the interference-limited tracking scenarios. Based on this, the objective function incorporates the tradeoff between bandwidth utilization and interference avoidance, and the predicted conditional (PC) Bayes risk based on tracking accuracy-oriented criterion. Secondly, to solve the optimal radar policy by minimizing the above objective function, based on a CTR framework with a dual controller structure, we propose the HSD-WO method for WO-JTAI. By decoupling this objective function and solving it sequentially, this two-step decoupling method iteratively performs the stochastic optimization (SO)-based spectrum occupancy optimization and deterministic optimization (DO)-based waveform parameters optimization until converging to the optimal policy. Simulation results demonstrate that effectiveness of the proposed method by comparing with other methods.
26 May 2025 in IEEE Transactions on Aerospace and Electronic Systems
To enhance maneuvering target tracking and mitigate noise suppression interference under low radar interception probability, by exploiting the inter-task reusable information and coupling relationship, the proposed cognitive radar multidimensional resource management (CRMRM) strategy collaboratively optimizes multi-dimensional signal domains to improve the linked multi-task processing capability. Firstly, to characterize comprehensive multi-task performance and diversified constraints, the CRMRM-oriented task priority-guided hierarchical criterion function constructs the predictive conditional Bayesian risk lower bound (PC-BRLB) and frequency-domain anti-jamming reward, while treating the radar interception probability as a secondary constraint, alongside real-time spectrum limitation and resource constraint, to form multicomponent constraint term. By constructing the subchannel selection action to formalize frequency-domain resource, this reward integrates the trade-off between bandwidth utilization and interference avoidance, and future multistep reward impact based on state transition. Secondly, to tackle the underlying high-dimensional, multi-constraint problem, via criterion characteristics analysis, a multiscale information-driven algorithm-level collaborative iterative solution method is proposed to address discrete and continuous variables separately through strategy decomposition and problem segmentation. The TD Actor-Critic (TDAC) iteratively performs the temporal difference (TD) error-based policy and value update, addressing the delayed reward and future multistep effect in frequency-domain resource optimization. Leveraging two propositions, the radiated power and dwell time are then jointly analytically optimized under interception threshold, reducing variable dimension and computational complexity. The improved artificial bee colony (IABC) handles the PC-BRLB by balancing global exploration and local exploitation, converging to the optimal pulse length and frequency modulation (FM) rate. Simulation results confirm CRMRM's superiority over alternative methods.