In autonomous vehicles (AV), sensor fusion methods have proven to be effective in merging data from multiple sensors and enhancing their perception capabilities. In the context of sensor fusion, the distinct strengths of multi-sensors, such as LiDAR, RGB, Thermal sensors, etc., can be leveraged to mitigate the impact of challenges imposed by extreme weather conditions. In this paper, we address multi-sensor fusion in AVs and present a comprehensive integration of a thermal sensor aimed at enhancing the cognitive robustness of AVs. Thermal sensors possess an impressive capability to detect objects and hazards that may be imperceptible to traditional visible light sensors. When integrated with RGB and LiDAR sensors, the thermal sensor becomes highly beneficial for detecting and locating objects in adverse weather conditions. The proposed deep learning-assisted multi-sensor fusion technique consists of two parts: (1) visual information fusion and (2) object detection using LiDAR, RGB, and Thermal sensors. The visual fusion framework employs a CNN (convolutional neural network) inspired by a domain image fusion algorithm. The object detection framework uses the modified version of the YoloV8 model, which exhibits high accuracy in real-time detection. In the YoloV8 model, we adjusted the network architecture to incorporate additional convolutional layers and altered the loss function to enhance detection accuracy in foggy and rainy conditions. The proposed technique is effective and adaptable in challenging conditions, such as night or dark mode, smoke, and heavy rain. The experimental results of the proposed method demonstrate enhanced efficiency and cognitive robustness compared to state-of-the-art fusion and detection techniques. This is evident from tests conducted on two public datasets (FLIR and TarDAL) and one private dataset (CUHK).

To increase the reliability of the printed circuit board (PCB) manufacturing process, automated optical inspection is often employed for soldering defect detection. However, traditional approaches built on handcrafted features, predefined rules, or thresholds are often susceptible to the variation of the acquired images’ quality and give unstable performances. To solve this problem, a deep learning‐based soldering defect detection method is developed in this article. Like many real‐life deep learning applications, the number of available training samples is often limited. This creates a challenging low‐data scenario, as deep learning typically requires massive data to perform well. To address this issue, a multitask learning model is proposed, namely, PCBMTL, that can simultaneously learn the classification and segmentation tasks under low‐data regimes. By acquiring the segmentation knowledge, classification performance is substantially improved with few samples. To facilitate the study, a soldering defect image dataset, namely, PCBSPDefect, is built. It focuses on the dual in‐line packages (DIP) at the PCB back side, DIP at the PCB front side, and flat flexible cables. Experimental results show that the proposed PCBMTL outperforms the best existing approaches by over 5–17% of average accuracy for different datasets.

Indoor positioning system (IPS) technologies have widespread applications in logistics, intelligent manufacturing, healthcare monitoring, etc. The recently released Bluetooth low‐energy (BLE) 5.1 specification enables in‐phase and quadrature‐phase (I/Q) data measurements. It allows angle of arrival estimation and becomes a natural choice for IPS implementation. Conventional BLE 5.1 IPSs use multiple anchors to provide massive redundancy to improve system robustness. It however demands effective approaches to leverage redundancy. Besides, interference due to various environmental factors can introduce severe errors to I/Q data and affect positioning accuracy. Facing these challenges, herein, a novel deep learning‐based multianchor BLE 5.1 IPS is proposed. The system aggregates measurements from multiple anchors and makes them available at regular time steps. Then, a novel attentional filtering network tailored to infer high‐quality I/Q sample data is developed and a spatial regularization loss incorporating spatial location relationships to strengthen the feature embedding discrimination is proposed. Two multianchor BLE 5.1 I/Q sample datasets are developed and released for public download. Numerical experiments are carried out to compare the proposed method with previous BLE 5.1 IPS methods and methods utilizing other radio frequency data. Results indicate that the proposed method consistently achieves submeter accuracy and significantly outperforms the state‐of‐the‐art approaches.

In big data era, companies and organizations are keen to collect data from users and analyse their behaviour patterns to make decisions or predictions for profits. However, it undermines users’ privacy because the collected data can be quite sensitive and easy to leak. To address privacy problems, local differential privacy (LDP) has been proposed for untrusted data collectors to obtain statistical information without compromising user privacy. Most studies on LDP assume that all users fully cooperate and contribute to the data collection process and thus the collected dataset is complete. However, in practice, especially when user population is large, such assumption seldom holds due to communication loss, user unresponsiveness or unwillingness, and incomplete user-side data. Unfortunately, state-of-the-art LDP-based data collection schemes, such as GRR, OUE and OLH, cannot handle partial data collection effectively. In this paper, we propose collaborative sampling to address partial data collection in a multi-dimensional setting. Thanks to a two-phase mechanism, we can derive the optimal sampling rate for each dimension. The optimality is shown and proved with respect to the variance of estimated frequency. Besides that, collaborative sampling is general and can be used in GRR, OUE and OLH with minimal adaption. Through experimental results, we show collaborative sampling outperforms existing mainstream data collection schemes in partial multi-dimensional data collection.

Indoor positioning system (IPS) enables real-time tracking and positioning within indoor environments and supports various location-based services across different application settings. The growth of the Internet of Things has enabled fingerprint-based IPS to achieve sub-meter or centimeter-level accuracy. However, packet loss issues caused by attenuation, interference from other devices, and noise hinder the fingerprint-based IPS. Despite developments in IPS methods, such as fingerprint augmentation and the integration of variational inference, effectively using these techniques with fingerprints under packet loss to achieve robust positioning continues to pose a challenge. This paper develops a real-time positioning model named Packet Loss Indoor Positioning Net (PLIPNet). PLIPNet combines the variational inference process and encodes statistical means of fingerprints in each location as prior distributions of latent variables, making it free from prior parameter configurations. The location’s spatial information is incorporated into latent space and probability representation to improve latent distribution distinguishability and avoid significant positioning errors. Comparisons with state-of-the-art positioning models show that PLIPNet consistently performs the best under various packet loss settings. For instance, when the packet loss rate reaches 80%, PLIPNet achieves a localization error of only 25.7% of that achieved by the best existing model.

Rail flaws can be initiated either from manufacturing or service-related anomalies. These flaws may lead to train derailment while in service if not detected timely. Traditional non-destructive testing (NDT) techniques are mainly contact-type and have low inspection speeds. Laser ultrasonic testing is an advanced NDT that provides non-contact, broadband, and highly sensitive inspection. Meanwhile, Rayleigh waves suffer little attenuation during propagation and travel long distances on curved surfaces like rail. Therefore, this research proposes a design of a fully non-contact laser-based rail inspection system to detect surface and subsurface defects at different parts (head, web, and foot) of the inspected rail track. The laser-based thermoelastic generation of narrowband Rayleigh waves was investigated using the finite element method (FEM). The experiments were performed on defective rail samples with artificial surface and circular subsurface defects in the inspected rail track. The time and frequency analyses of both simulation and experimental results show that the line arrayed pattern (LAP) type of laser illumination is promising in generating narrowband Rayleigh waves that have great potential in detecting rail defects. In field measurements, finding the defect’s location is usually challenging because unwanted wave packets mostly surround defect echoes in a laser-generated ultrasonic signal. In this regard, a criterion named time-of-flight-based flaw detection (TOFFD) was proposed to identify defect echoes surrounded by high noise peaks automatically. TOFFD was successfully applied to laser-generated ultrasonic signals captured at the rail track to find the location of the flaw by identifying the defect echo.