ABSTRACT Localisation of intermittent radio frequency (RF) interference sources is important for urban spectrum monitoring and the protection of safety‐critical navigation services. Among existing approaches, received signal strength indicator (RSSI)‐based localisation is attractive for unmanned aerial vehicle (UAV) platforms because it requires lightweight receivers and no antenna arrays or strict time synchronisation. However, most RSSI‐based localisation and UAV path‐planning methods rely on fixed sensing geometries, empirical dwell times or propagation models weakly coupled with online manoeuvring decisions. These limitations become critical in dense urban environments, where building‐induced line‐of‐sight/non‐line‐of‐sight transitions, intermittent emissions and unknown propagation parameters jointly reduce localisation reliability. In this paper, we propose a UAV‐based active localisation framework, termed VISTA, for intermittent urban RF interference sources. VISTA formulates each decision as a joint selection of flight direction and dwell time, maintains a joint belief over the source position, propagation parameters and emission behaviour and uses a unified action‐conditioned observation model for both planning and Bayesian update. A control‐aware Bayes‐risk criterion with three‐branch approximate scoring enables real‐time action selection. Unlike waypoint‐only informative path planning or estimator‐side NLOS mitigation, VISTA couples spatial motion, sensing time, intermittent detection and map‐aware propagation within one Bayesian action‐evaluation loop. Closed‐loop simulations using real Shanghai Lujiazui GIS data evaluate the proposed method, showing an reduction in per‐step planning time compared with a Monte Carlo reference while maintaining comparable localisation accuracy.
ABSTRACT Deep neural network‐based synthetic aperture radar automatic target recognition (SAR‐ATR) systems are vulnerable to adversarial perturbations. Existing physical‐domain adversarial attacks usually assume accurately known radar and platform parameters. In practical SAR jamming, reconnaissance errors can deform and defocus adversarial templates during signal modulation and image formation, reducing attack effectiveness. To address this problem, this paper proposes Hybrid Competition and Pre‐attenuation of Errors for Cross‐domain Adversarial Attacks (HCAD). Specifically, HCAD first uses class‐level mean saliency templates to guide one‐step universal adversarial perturbation generation towards class decision boundaries. Second, a dual‐path optimization framework combines direct attack and mixed‐image competition to preserve white‐box effectiveness and improve cross‐model transferability. Finally, an error pre‐attenuation module simulates parameter‐induced geometric deformation and spatially varying defocus during training, enhancing robustness after digital‐to‐physical mapping. Experiments on MSTAR and SEN1‐2 show that, under 20% reconnaissance parameter error, HCAD reduces correct recognition rates to 24.38% and 38.88% in non‐targeted attacks, compared with 45.76% and 56.35% for the strongest baseline. For targeted attacks, HCAD achieves targeted success rates of 54.12% and 67.10%, exceeding baseline results of 42.94% and 56.87%. These results demonstrate improved robustness against SAR parameter uncertainty.
ABSTRACT The rapid rise of the low‐altitude economy has significantly increased the demand for high‐performance positioning, navigation and timing (PNT) services. Meanwhile, the large‐scale construction of LEO satellite communications provides new opportunities for resilient navigation systems. In the future, space–air–ground integrated networks require an overall air‐interface architecture that is expected to jointly support high‐precision ranging, high‐rate communication, sensing and high‐dynamic adaptability. Motivated by the objective of transforming communication signals to achieve superior measurement capabilities, this paper proposes a novel modulation scheme integrating navigation and communication functions, namely orthogonal chirp division multiplexing with binary chirp subcarriers (OCDM‐BCS), which uses orthogonal binary offset chirp carrier (BOCC) signals as subcarriers. First, based on the proposed signal model, the analytical expressions for the ACF and PSD are derived, with the theoretical formulas validated through simulation results. Subsequently, the relationship between OCDM‐BCS parameters and their time–frequency characteristics is discussed. Compared with representative modulation schemes such as BOC(10,5), MBOC‐POS, OFDM, and OFDM‐BOC, the analysis shows that, due to its linear frequency modulation (LFM) characteristics, OCDM‐BCS exhibits superior code tracking accuracy, anti‐interference robustness and anti‐multipath capability. An NR‐inspired integrated communication‐navigation frame is further constructed, in which OCDM‐BCS is employed for low‐rate control/signalling transmission while conventional CP‐OFDM is retained for high‐rate payload data. Under ideal synchronisation and AWGN conditions, the BER performance of OCDM‐BCS is shown to be comparable to that of CP‐OFDM. With its excellent comprehensive performance, OCDM‐BCS provides a solid signal foundation for collaborative PNT services in future low‐altitude unmanned aerial vehicle (UAV) swarm applications.
ABSTRACT Aiming at the problems that the original Multi‐band Multi‐angle Feature Fusion Network (MMFFNet) is easily disturbed by noise and adopts a fixed multi‐band feature fusion mode under low signal‐to‐noise ratio (SNR), a UAV weak target micro‐Doppler signal recognition method based on improved MMFFNet is proposed. This method constructs a noise‐aware attention mechanism by integrating Convolutional Block Attention Module (CBAM) and Squeeze‐and‐Excitation (SE) attention, which suppresses noise‐induced misfocus and enhances effective target features. Meanwhile, an SNR‐adaptive Attentional Feature Fusion (AFF) module is designed to dynamically adjust the weights of multi‐band features according to the noise level. Experimental results demonstrate that the proposed method achieves a validation accuracy of 91.65% at −3 dB SNR, which outperforms the original MMFFNet by 6.59%. This work can serve as an effective technical support for robust radar identification of low, slow, small (LSS) UAV targets.
ABSTRACT Marine object detection is essential for maritime surveillance and marine scientific research. However, existing methods predominantly rely on single‐modality data, leaving effective cross‐modal information exchange mechanisms underexplored. This limitation makes detection models highly susceptible to dynamic environmental changes, often compromising detection accuracy. To address this, we research DM‐MODNet, a robust bio‐inspired network for marine object detection using dual‐modal remote sensing images (i.e., RGB and hyperspectral images). Inspired by the biological visual system (BVS), DM‐MODNet mirrors cognitive information processing through a cascaded structure comprising single‐modality extraction, dual‐modal fusion and output modules. Inspired by the bottom‐up and top‐down information‐flow principles of the BVS, the proposed network achieves highly efficient feature fusion and comprehensive interaction across diverse information sources, rendering the entire detection process biologically interpretable. Extensive experiments conducted on the M 2 SODAI and VEDAI datasets demonstrate that DM‐MODNet significantly outperforms state‐of‐the‐art methods in overall detection accuracy. Furthermore, it exhibits remarkable robustness against severe noise and strong adaptivity to complex, dynamic weather conditions.
