Accurate tracking and predicting unmanned aerial vehicle (UAV) trajectories are essential to ensure mission success, equipment safety, and data accuracy. Maneuverable UAVs exhibit complex and dynamic motion, and conventional tracking algorithms that rely on predefined models perform poorly when unknown parameters are used. To address this issue, this paper introduces a hybrid dual-scale neural network model based on the generalized regression multi-model and cubature information filter (GRMM-CIF) framework. We have established the GRMM-CIF filtering structure to differentiate motion modes and reduce measurement noise. Furthermore, considering trajectory datasets and rates of motion change, a neural network at different scales will be designed. We propose the dual-scale bidirectional long short-term memory (DS-Bi-LSTM) algorithm to address prediction delays in a multi-model context. Additionally, we employ scale sliding windows and threshold-based decision-making to achieve dual-scale trajectory reconstruction, ultimately enhancing tracking accuracy. Simulation results confirm the effectiveness of our approach in handling the uncertainty of UAV motion and achieving precise estimations.

Direction of arrival (DOA) estimation plays a critical role in remote sensing, where it aids in identifying and tracking multiple targets across complex environments, from atmospheric monitoring to resource mapping. Leveraging difference covariance array (DCA) for DOA estimation has become prevalent, particularly with sparse arrays capable of resolving more targets than the number of sensors. This paper proposes a new hole-free sparse array configuration for remote sensing applications to achieve improved DOA estimation performance using DCA. By symmetrically placing a minimum redundancy array (MRA) and its complementary MRA on both sides of a sparse uniform linear array (ULA), this configuration maximizes degrees of freedom (DOFs) and minimizes mutual coupling effects. Expressions for calculating sensor positions and optimal element allocation methods to maximize DOFs are derived. Simulation experiments in various scenarios have shown the advantages of the proposed array in DOA estimation, including a strong ability to estimate multi-targets, high angular resolution, low estimation error, and strong robustness to mutual coupling.

Direct localization methods offer significant advantages in modern high-frequency (HF) radar systems for maritime target localization. However, the presence of amplitude–phase errors poses severe challenges to their accuracy and reliability. Traditional active calibration methods for direct localization often face high costs and operational complexities, while self-calibration techniques are limited to addressing only single or, at most, two types of error models. To address these issues, the Multi-Station Azimuth-based Position Error Calibration Algorithm (MAPEC) is proposed in this paper, which is an active calibration algorithm to solve mixed amplitude–phase errors in direct positioning. The proposed algorithm converts the estimated amplitude–phase errors on each angle of stations into the corresponding position on the localization search grid and adjusts the steering vectors based on the converted errors. By integrating MAPEC with direct localization algorithms, the method enables precise and robust localization under complex amplitude–phase error conditions involving multiple error sources. Simulation results demonstrate that this approach significantly improves localization accuracy across various error conditions, validating its applicability to direct localization applications.

Multi radar cooperative tracking leverages spatial diversity and measurement redundancy to enhance detection reliability and state estimation accuracy. Nevertheless, most existing fusion schemes adopt fixed or heuristically tuned weights and rely on simplified statistical assumptions, which limits robustness in non‐stationary electromagnetic environments where station wise SNR varies rapidly. This paper presents a deep reinforcement learning (DRL) driven adaptive fusion framework that jointly optimises measurement fusion weights and tracking‐filter parameters. A central feature is a radar‐physics‐guided inversion‐matching feedback loop that explicitly couples the fused Cartesian state with each radar's local measurement domain. Specifically, the fused estimate is back projected to local polar coordinates to compute spatio‐temporal residuals and Mahalanobis distance consistency statistics. Together with real‐time SNR and normalised innovation squared (NIS), these physically interpretable signals form the state representation for a Twin Delayed Deep Deterministic Policy Gradient (TD3) agent. We further design a multi objective reward to trade off position accuracy, velocity fidelity, tracking smoothness and uncertainty reduction. Simulations under heterogeneous sensing conditions verify that the proposed method improves adaptive weighting behaviour and preserves trajectory continuity compared with conventional baselines.

Stable tracking of maneuvering targets, such as Unmanned Aerial Vehicles (UAVs), in low Signal-to-Noise Ratio (SNR) environments remains a significant challenge. Weak target signals, often obscured by noise, frequently lead to failures in multi-frame energy accumulation, substantial state estimation errors, and degraded tracking performance. To address these limitations, this paper proposes the Adaptive Spatio-temporal Tracking Recurrent Attention Network (ASTRA-Net), an adaptive dual-task learning extension of the Track-Before-Detect (TBD) framework. ASTRA-Net integrates spatiotemporal feature fusion embedding with joint pre-detection and tracking. It captures temporal and spatial dependencies within noisy radar measurements and employs a multi-head self-attention mechanism to dynamically recalibrate feature weights based on varying motion patterns. This framework shares underlying features without stringent prior dynamic assumptions, enhancing overall robustness. It simultaneously learns trajectory prediction and state recognition, discriminates between model states using an information entropy threshold, and adaptively adjusts TBD model parameters during abrupt maneuvers. Experimental results demonstrate that ASTRA-Net excels in low-SNR, noise dominated environments, better than conventional methods in tracking accuracy. This approach offers a novel technical solution for the perception and tracking of low observable targets, particularly suitable for security surveillance and airspace management systems operating in noisy conditions.