1 January 2024 in Journal of Advanced Transportation
Flight trajectory prediction is one of the key issues in ensuring the safety of air traffic, providing the air traffic controller with the foresight of flight conflicts so that control instructions for pilots can be preconceived. In a complicated mechanism, flight trajectories can be severely affected by convective weather, making accurately predicting trajectories challenging. To address this problem, we propose a boosted spatiotemporal deep learning ensemble for mining the law of how convective weather affects flight trajectory stretching. Instead of conventionally representing trajectory data in a geographic coordinate system, we design a relative coordinate system for gaining new trajectory features which tangibly reflect trajectory’s relations with planned route and convective weather. Besides, we raise a boosted ensemble framework of spatiotemporal deep learning models, trained by the samples pairing sequential trajectory with graphical weather, dedicating to strengthen the mining of the high‐value training samples that involve explicit flight deviations caused by convective weather. The experiments using actual flight and weather data demonstrate our method’s superiority in predicting flight trajectory affected by convective weather.
23 May 2019 in Science China Information Sciences
3 March 2025 in IEEE Transactions on Instrumentation and Measurement
Multivariate time series anomaly detection is crucial in the sensitive domains such as cybersecurity and grid monitoring, significantly contributing to reliability and safety of system operation. However, current methods suffer from inadequate utilization of decomposed time series, insufficient mining of contextual dependencies within the time series, and limited robustness against anomalies during training. To address these limitations, we propose ConFlow (Consistencyenhanced normalizing Flow model), which utilizes consistency of decomposed time series and contextual temporal embedding to enhance discriminative ability of flow model. First, to refine extraction of time series components, we propose a cascade decomposition and mixing module that iteratively decouple the time series. Second, these components are mapped to Gaussian distributions through the context-aware normalizing flow, incorporating both inter- and intra-series information into the density estimation. Third, the density consistency among decomposed time series is measured to reweight the estimation, while highly inconsistent series are viewed as anomalies and masked during training to improve model robustness. Finally, anomalies are detected using reweight density estimation. Experiments on five widely used datasets in the time series anomaly detection field demonstrate the superiority of our method over state-of-the-art approaches.
2 October 2025 in IEEE Transactions on Vehicular Technology
This paper explores the potential of dual-polarized (DP) technology with holographic multiple-input multi-output (HMIMO) surfaces. The primary goal is to mitigate the mutual interference between communication and sensing signals in integrated sensing and communication (ISAC) systems. By leveraging distinct polarization resources, DP enables the simultaneous transmission of both signal types within the same frequency and spatial domain. However, excessive transmission power will lead to increased polarization leakage. To address it, a robust power control scheme is proposed based on the DP ISAC channel model. In both perfect and imperfect channel state information (CSI) scenarios, our primary focus is minimizing power consumption while suppressing cross-polarization interference and ensuring optimal communication and sensing performance. Due to the massive number of antennas in HMIMO and the presence of multiple users and sensing targets, the problem becomes high-dimensional and highly non-convex, making it extremely challenging to solve. To address this challenge, a novel differential evolution-successive convex approximation (DE-SCA) algorithm is proposed. Through the deliberate integration of mutation and crossover mechanisms, our approach effectively navigates the high-dimensional precoding matrix space. Moreover, we devise a tailored convexification strategy to systematically reformulate the non-convex constraints, transforming the objective into a solvable convex optimization subproblem. Numerical results demonstrate that the synergy of global exploration and local refinement enhances both solution efficiency and convergence speed, particularly in large-scale HMIMO systems. The proposed algorithm achieves better communication and sensing performance with less power consumption. Additionally, it attains higher polarization diversity gain compared to conventional MIMO systems.
15 November 2024 in Multimedia Systems