The demand for indoor location services is growing rapidly. Various indoor positioning techniques have been proposed. As a supplement, public FM radio signal based methods can provide a possible solution for infrastructure-free indoor positioning which benefits from its wide coverage. However, existing research mainly focuses on algorithm development, with a notable absence of theoretical performance analysis. To bridge this gap, this paper firstly analyzed the FM signal indoor positioning performance by deriving the Fisher information. Our findings indicate that: FM RSS theoretically has difficulties in two-dimensional (2D) indoor positioning and exhibits directional differences in one-dimensional (1D) performance. Thus, when both directions are used for localization simultaneously, the direction with blurred features will significantly affect the overall accuracy of 2D positioning. Besides, there exists a problem of finding effective FM positioning region in large-scale scenes. To address these issues: First, the method for finding positioning region within effective FM signal environmental field with PDR aided is designed. Second, focusing on the problem of 2D positioning, a new strategy is suggested that takes the 1D characteristics of FM RSS by employing the split-direction accuracy estimation and parameter adjustment. To begin with, three improved lightweight accuracy estimation indicators are adopted to assess the FM signal positioning performance. Then, the split-direction adaptive FM fingerprinting method is proposed, which integrates the FM RSS signal and PDR-constrained space metrics by using the accuracy estimation results in different directions. Practical experiments are conducted in three large-scale scenes, and the results demonstrated an improvement of 86.6% and 33.4% compared to traditional 2D fingerprinting and fusion methods, respectively.

Existing integrated sensing and communication technology designed for monostatic settings segregates sensing and communication (S&C) at distinct terminals, impeding their adaptability to network paradigms. In this paper, we propose a general framework for joint target-localization and data-detection (JTD) in bistatic settings by using a combination of the deterministic known (DK) and random unknown (RU) symbols, e.g., pilot and data symbols. We first derive the performance limits for target localization-related parameters with combined symbols. We demonstrate that the Fisher information matrix (FIM) for target localization using the DK part can be expressed in a closed form through orthogonal projection, while that using the RU part aligns with the modified FIM in monostatic settings at an exponential rate with respect to the signal-to-noise ratio. Then, we propose a JTD scheme to exploit the reciprocal advantages for S&C, i.e., the detected data symbols are harnessed to enhance target localization in an alternating way. Finally, simulation results validate our theoretical analysis and the effectiveness of the proposed JTD scheme.

The integrated sensing and communication-enhanced (ISAC-enhanced) intelligent transportation system (ITS) has great development potential and market value in future urban vehicle-to-infrastructure (V2I) scenarios. With limited total power consumption, joint design of power distribution and beam steering of sensing and communication subsystems is required to achieve optimal system performance. This paper introduces an intelligent power allocation and beam-steering control system, where both the system actions are determined based on the radar sensing data cube. The adaptive power allocation module employs the Actor-Critic reinforcement learning agents to optimize power distribution based on vehicle GPS data, communication channel capacity, and historical power allocation, balancing the overall system performance including the sensing and communication functions. The automatic beam control module, utilizing the convolutional neural network (CNN) for dynamic beam control, intelligently adapts communication beams for vehicles navigating complex roadways, enhancing the beam tracking accuracy and quality of service. Validation on real-world datasets showcases the effectiveness of these modules, highlighting the potential to enhance vehicular communication performance.

Remote sensing scene zero-shot object detection aims to detect and recognize both seen and unseen catrgories of landscape elements with the guidance of the word embeddings. In this task, two primary challenges are identified. Firstly, there exists considerable variability within categories of landscape elements, causing a misalignment between visual features and word embeddings, particularly noticeable for unseen categories. Secondly, existing detection models struggle to provide accurate localization predictions, greatly impacting overall performance. To address these two issues, we propose WEA-DINO (Word Embedding Alignment-DINO). Based on the original DINO structure, our WEA-DINO-Head is specifically designed to align the hidden features of “matching queries” with word embedding features, effectively addressing the misalignment issue between visual features and word embeddings. Furthermore, aligning the hidden features of “denoising queries” with word embedding features enables the translation of localization capabilities from known categories to previously unseen ones. Through extensive experimentation on the DIOR benchmark dataset, our method demonstrates state-of-the-art performance. The code is available at https://github.com/cv516Buaa/WEA-DINO.

The advancement of deep neural network has propelled the widespread application of remote sensing target detection. However, compared to natural scenes, remote sensing targets possess inherent characteristics such as weak features and small scale, leading to a significant performance gap in traditional detection methods. To address these challenges, we undertake a systematic analysis of existing approaches, focusing on two key aspects: inadequate extraction of discriminative features and inappropriate regression measurement metrics. To tackle the first issue, an instance-level neighboring enhancement network (INEN) is proposed, enhancing the network’s feature extraction capability through inter-object feature aggregation. To address the second issue, a novel metric, smooth Wasserstein loss (SWL), is devised. Building upon these principles, a new tiny object detection (TOD) network for remote sensing images is developed. Extensive experiments on AI-TOD v1/v2 and DOTA v2 remote sensing tiny target detection datasets demonstrate that our approach achieves state-of-the-art (SOTA) performance. Codes are available athttps://github.com/sevenwgb/SWIN-TOD.