Sensors

One of the challenges of spatial cognition, such as self-localization and navigation, is to develop an efficient learning approach capable of mimicking human ability. This paper proposes a novel approach for topological geolocalization on the map using motion trajectory and graph neural networks. Specifically, our learning method learns an embedding of the motion trajectory encoded as a path subgraph where the node and edge represent turning direction and relative distance information by training a graph neural network. We formulate the subgraph learning as a multi-class classification problem in which the output node IDs are interpreted as the object’s location on the map. After training using three map datasets with small, medium, and large sizes, the node localization tests on simulated trajectories generated from the map show 93.61%, 95.33%, and 87.50% accuracy, respectively. We also demonstrate similar accuracy for our approach on actual trajectories generated by visual-inertial odometry. The key benefits of our approach are as follows: (1) we take advantage of the powerful graph-modeling ability of neural graph networks, (2) it only requires a map in the form of a 2D graph, and (3) it only requires an affordable sensor that generates relative motion trajectory. Full article

Autonomous grounded vehicle-based social assistance/service robot parking in an indoor environment is an exciting challenge in urban cities. There are few efficient methods for parking multi-robot/agent teams in an unknown indoor environment. The primary objective of autonomous multi-robot/agent teams is to establish synchronization between them and to stay in behavioral control when static and when in motion. In this regard, the proposed hardware-efficient algorithm addresses the parking of a trailer (follower) robot in indoor environments by a truck (leader) robot with a rendezvous approach. In the process of parking, initial rendezvous behavioral control between the truck and trailer robots is established. Next, the parking space in the environment is estimated by the truck robot, and the trailer robot parks under the supervision of the truck robot. The proposed behavioral control mechanisms were executed between heterogenous-type computational-based robots. Optimized sensors were used for traversing and the execution of the parking methods. The truck robot leads, and the trailer robot mimics the actions in the execution of path planning and parking. The truck robot was integrated with FPGA (Xilinx Zynq XC7Z020-CLG484-1), and the trailer was integrated with Arduino UNO computing devices; this heterogenous modeling is adequate in the execution of trailer parking by a truck. The hardware schemes were developed using Verilog HDL for the FPGA (truck)-based robot and Python for the Arduino (trailer)-based robot. Full article

Using object detection techniques on immature fruits to find out their quantity and position is a crucial step for intelligent orchard management. A yellow peach target detection model (YOLOv7-Peach) based on the improved YOLOv7 was proposed to address the problem of immature yellow peach fruits in natural scenes that are similar in color to the leaves but have small sizes and are easily obscured, leading to low detection accuracy. First, the anchor frame information from the original YOLOv7 model was updated by the K-means clustering algorithm in order to generate anchor frame sizes and proportions suitable for the yellow peach dataset; second, the CA (coordinate attention) module was embedded into the backbone network of YOLOv7 so as to enhance the network’s feature extraction for yellow peaches and to improve the detection accuracy; then, we accelerated the regression convergence process of the prediction box by replacing the object detection regression loss function with EIoU. Finally, the head structure of YOLOv7 added the P2 module for shallow downsampling, and the P5 module for deep downsampling was removed, effectively improving the detection of small targets. Experiments showed that the YOLOv7-Peach model had a 3.5% improvement in mAp (mean average precision) over the original one, much higher than that of SSD, Objectbox, and other target detection models in the YOLO series, and achieved better results under different weather conditions and a detection speed of up to 21 fps, suitable for real-time detection of yellow peaches. This method could provide technical support for yield estimation in the intelligent management of yellow peach orchards and also provide ideas for the real-time and accurate detection of small fruits with near background colors. Full article

The need for power-efficient devices, such as smart sensor nodes, mobile devices, and portable digital gadgets, is markedly increasing and these devices are becoming commonly used in daily life. These devices continue to demand an energy-efficient cache memory designed on Static Random-Access Memory (SRAM) with enhanced speed, performance, and stability to perform on-chip data processing and faster computations. This paper presents an energy-efficient and variability-resilient 11T (E²VR11T) SRAM cell, which is designed with a novel Data-Aware Read–Write Assist (DARWA) technique. The E²VR11T cell comprises 11 transistors and operates with single-ended read and dynamic differential write circuits. The simulated results in a 45 nm CMOS technology exhibit 71.63% and 58.77% lower read energy than ST9T and LP10T and lower write energies of 28.25% and 51.79% against S8T and LP10T cells, respectively. The leakage power is reduced by 56.32% and 40.90% compared to ST9T and LP10T cells. The read static noise margin (RSNM) is improved by 1.94× and 0.18×, while the write noise margin (WNM) is improved by 19.57% and 8.70% against C6T and S8T cells. The variability investigation using the Monte Carlo simulation on 5000 samples highly validates the robustness and variability resilience of the proposed cell. The improved overall performance of the proposed E²VR11T cell makes it suitable for low-power applications. Full article

Deep learning algorithms have the advantages of a powerful time series prediction ability and the real-time processing of massive samples of big data. Herein, a new roller fault distance estimation method is proposed to address the problems of the simple structure and long conveying distance of belt conveyors. In this method, a diagonal double rectangular microphone array is used as the acquisition device, minimum variance distortionless response (MVDR) and long short-term memory network (LSTM) are used as the processing models, and the roller fault distance data are classified to complete the estimation of the idler fault distance. The experimental results showed that this method could achieve high-accuracy fault distance identification in a noisy environment and had better accuracy than the conventional beamforming algorithm (CBF)-LSTM and functional beamforming algorithm (FBF)-LSTM. In addition, this method could also be applied to other industrial testing fields and has a wide range of application prospects. Full article

Since introducing the Transformer model, it has dramatically influenced various fields of machine learning. The field of time series prediction has also been significantly impacted, where Transformer family models have flourished, and many variants have been differentiated. These Transformer models mainly use attention mechanisms to implement feature extraction and multi-head attention mechanisms to enhance the strength of feature extraction. However, multi-head attention is essentially a simple superposition of the same attention, so they do not guarantee that the model can capture different features. Conversely, multi-head attention mechanisms may lead to much information redundancy and computational resource waste. In order to ensure that the Transformer can capture information from multiple perspectives and increase the diversity of its captured features, this paper proposes a hierarchical attention mechanism, for the first time, to improve the shortcomings of insufficient information diversity captured by the traditional multi-head attention mechanisms and the lack of information interaction among the heads. Additionally, global feature aggregation using graph networks is used to mitigate inductive bias. Finally, we conducted experiments on four benchmark datasets, and the experimental results show that the proposed model can outperform the baseline model in several metrics. Full article

Changes in pig behavior are crucial information in the livestock breeding process, and automatic pig behavior recognition is a vital method for improving pig welfare. However, most methods for pig behavior recognition rely on human observation and deep learning. Human observation is often time-consuming and labor-intensive, while deep learning models with a large number of parameters can result in slow training times and low efficiency. To address these issues, this paper proposes a novel deep mutual learning enhanced two-stream pig behavior recognition approach. The proposed model consists of two mutual learning networks, which include the red–green–blue color model (RGB) and flow streams. Additionally, each branch contains two student networks that learn collaboratively to effectively achieve robust and rich appearance or motion features, ultimately leading to improved recognition performance of pig behaviors. Finally, the results of RGB and flow branches are weighted and fused to further improve the performance of pig behavior recognition. Experimental results demonstrate the effectiveness of the proposed model, which achieves state-of-the-art recognition performance with an accuracy of 96.52%, surpassing other models by 2.71%. Full article

Traffic signs are updated quickly, and there image acquisition and labeling work requires a lot of manpower and material resources, so it is difficult to provide a large number of training samples for high-precision recognition. Aiming at this problem, a traffic sign recognition method based on FSOD (few-shot object learning) is proposed. This method adjusts the backbone network of the original model and introduces dropout, which improves the detection accuracy and reduces the risk of overfitting. Secondly, an RPN (region proposal network) with improved attention mechanism is proposed to generate more accurate target candidate boxes by selectively enhancing some features. Finally, the FPN (feature pyramid network) is introduced for multi-scale feature extraction, and the feature map with higher semantic information but lower resolution is merged with the feature map with higher resolution but weaker semantic information, which further improves the detection accuracy. Compared with the baseline model, the improved algorithm improves the 5-way 3-shot and 5-way 5-shot tasks by 4.27% and 1.64%, respectively. We apply the model structure to the PASCAL VOC dataset. The results show that this method is superior to some current few-shot object detection algorithms. Full article

The application of IoT (Internet of Things) technology to the health monitoring of expansion joints is of great importance in enhancing the efficiency of bridge expansion joint maintenance. In this study, a low-power, high-efficiency, end-to-cloud coordinated monitoring system analyzes acoustic signals to identify faults in bridge expansion joints. To address the issue of scarce authentic data related to bridge expansion joint failures, an expansion joint damage simulation data collection platform is established for well-annotated datasets. Based on this, a progressive two-level classifier mechanism is proposed, combining template matching based on AMPD (Automatic Peak Detection) and deep learning algorithms based on VMD (Variational Mode Decomposition), denoising, and utilizing edge and cloud computing power efficiently. The simulation-based datasets were used to test the two-level algorithm, with the first-level edge-end template matching algorithm achieving fault detection rates of 93.3% and the second-level cloud-based deep learning algorithm achieving classification accuracy of 98.4%. The proposed system in this paper has demonstrated efficient performance in monitoring the health of expansion joints, according to the aforementioned results. Full article

As a powerful tool in scientific research and industrial technologies, the cold atom absolute gravity sensor (CAGS) based on cold atom interferometry has been proven to be the most promising new generation high-precision absolute gravity sensor. However, large size, heavy weight, and high–power consumption are still the main restriction factors of CAGS being applied for practical applications on mobile platforms. Combined with cold atom chips, it is possible to drastically reduce the complexity, weight, and size of CAGS. In this review, we started from the basic theory of atom chips to chart a clear development path to related technologies. Several related technologies including micro-magnetic traps, micro magneto–optical traps, material selection, fabrication, and packaging methods have been discussed. This review gives an overview of the current developments in a variety of cold atom chips, and some actual CAGS systems based on atom chips are also discussed. We summarize by listing some of the challenges and possible directions for further development in this area. Full article

The current approach to connected and autonomous driving function development and evaluation uses model-in-the-loop simulation, hardware-in-the-loop simulation and limited proving ground use, followed by public road deployment of the beta version of software and technology. The rest of the road users are involuntarily forced into taking part in the development and evaluation of these connected and autonomous driving functions in this approach. This is an unsafe, costly and inefficient method. Motivated by these shortcomings, this paper introduces the Vehicle-in-Virtual-Environment (VVE) method of safe, efficient and low-cost connected and autonomous driving function development, evaluation and demonstration. The VVE method is compared to the existing state-of-the-art. Its basic implementation for a path-following task is used to explain the method where the actual autonomous vehicle operates in a large empty area with its sensor feeds being replaced by realistic sensor feeds corresponding to its location and pose in the virtual environment. It is possible to easily change the development virtual environment and inject rare and difficult events which can be tested very safely. Vehicle-to-Pedestrian (V2P) communication-based pedestrian safety is chosen as the application use case for the VVE in this paper, and corresponding experimental results are presented and discussed. A no-line-of-sight pedestrian and vehicle moving towards each other on intersecting paths with different speeds are used in the experiments. Their time-to-collision risk zone values are compared for determining severity levels. The severity levels are used to slow down or brake the vehicle. The results show that V2P communication of pedestrian location and heading can be used successfully to avoid possible collisions. It is noted that actual pedestrians and other vulnerable road users can be used very safely in this approach. Full article

Dust or condensed water present in harsh outdoor or high-humidity human breath samples are one of the key sources that cause false detection in Micro Electro-Mechanical System (MEMS) gas sensors. This paper proposes a novel packaging mechanism for MEMS gas sensors that utilizes a self-anchoring mechanism to embed a hydrophobic polytetrafluoroethylene (PTFE) filter into the upper cover of the gas sensor packaging. This approach is distinct from the current method of external pasting. The proposed packaging mechanism is successfully demonstrated in this study. The test results indicate that the innovative packaging with the PTFE filter reduced the average response value of the sensor to the humidity range of 75~95% RH by 60.6% compared to the packaging without the PTFE filter. Additionally, the packaging passed the High-Accelerated Temperature and Humidity Stress (HAST) reliability test. With a similar sensing mechanism, the proposed packaging embedded with a PTFE filter can be further employed for the application of exhalation-related, such as coronavirus disease 2019 (COVID-19), breath screening. Full article

Millions of commuters face congestion as a part of their daily routines. Mitigating traffic congestion requires effective transportation planning, design, and management. Accurate traffic data are needed for informed decision making. As such, operating agencies deploy fixed-location and often temporary detectors on public roads to count passing vehicles. This traffic flow measurement is key to estimating demand throughout the network. However, fixed-location detectors are spatially sparse and do not cover the entirety of the road network, and temporary detectors are temporally sparse, providing often only a few days of measurements every few years. Against this backdrop, previous studies proposed that public transit bus fleets could be used as surveillance agents if additional sensors were installed, and the viability and accuracy of this methodology was established by manually processing video imagery recorded by cameras mounted on transit buses. In this paper, we propose to operationalize this traffic surveillance methodology for practical applications, leveraging the perception and localization sensors already deployed on these vehicles. We present an automatic, vision-based vehicle counting method applied to the video imagery recorded by cameras mounted on transit buses. First, a state-of-the-art 2D deep learning model detects objects frame by frame. Then, detected objects are tracked with the commonly used SORT method. The proposed counting logic converts tracking results to vehicle counts and real-world bird’s-eye-view trajectories. Using multiple hours of real-world video imagery obtained from in-service transit buses, we demonstrate that the proposed system can detect and track vehicles, distinguish parked vehicles from traffic participants, and count vehicles bidirectionally. Through an exhaustive ablation study and analysis under various weather conditions, it is shown that the proposed method can achieve high-accuracy vehicle counts. Full article

The large mode field area fiber can raise the tolerance of power, and high requirements for the bending characteristics of optical fibers are needed. In this paper, a fiber composed of a comb-index core, gradient-refractive index ring, and multi-cladding is proposed. The performance of the proposed fiber is investigated by using a finite element method at a 1550 nm wavelength. When the bending radius is 20 cm, the mode field area of the fundamental mode can achieve 2010 μm², and the bending loss is reduced to 8.452 × 10⁻⁴ dB/m. Additionally, when the bending radius is smaller than 30 cm, there are two variations with low BL and leakage; one is a bending radius of 17 cm to 21 cm, and the other is from 24 cm to 28 cm (except for 27 cm). When the bending radius is between 17 cm and 38 cm, the highest bending loss is 1.131 × 10⁻¹ dB/m and the lowest mode field area is 1925 μm². It has a very important application prospect in the field of high-power fiber lasers and telecom applications. Full article

Light pollution is an ongoing problem for city populations. Large numbers of light sources at night negatively affect humans’ day–night cycle. It is important to measure the amount of light pollution in order to effectively ascertain the amount of light pollution in the city area and effectively reduce it where possible and necessary. In order to perform this task, a prototype wireless sensor network for automated, long-term measurement of light pollution was developed for the Torun (Poland) city area. The sensors use LoRa wireless technology to collect sensor data from an urban area by way of networked gateways. The article investigates the sensor module architecture and design challenges as well as network architecture. Example results of light pollution measurements are presented, which were obtained from the prototype network. Full article

To overcome the temperature effect of NaI(Tl) detectors for energy spectrometry without additional hardware, a new correction method was put forward based on pulse deconvolution, trapezoidal shaping and amplitude correction, named DTSAC. To verify this method, actual pulses from a NaI(Tl)-PMT detector were measured at various temperatures from −20 °C to 50 °C. Pulse processing and spectrum synthesis showed that the position drift of the ¹³⁷Cs 662 keV peak was less than 3 keV, and the corresponding resolution at 662 keV of the sum spectra ranged from 6.91% to 10.60% with the trapezoidal width set from 1000 ns to 100 ns. The DTSAC method corrects the temperature effect via pulse processing, and needs no reference peak, reference spectrum or additional circuits. The method solves the problem of correction of pulse shape and pulse amplitude at the same time, and can be used even at a high counting rate. Full article

The intelligent fault diagnosis of main circulation pumps is crucial for ensuring their safe and stable operation. However, limited research has been conducted on this topic, and applying existing fault diagnosis methods designed for other equipment may not yield optimal results when directly used for main circulation pump fault diagnosis. To address this issue, we propose a novel ensemble fault diagnosis model for the main circulation pumps of converter valves in voltage source converter-based high voltage direct current transmission (VSG-HVDC) systems. The proposed model employs a set of base learners already able to achieve satisfying fault diagnosis performance and a weighting model based on deep reinforcement learning that synthesizes the outputs of these base learners and assigns different weights to obtain the final fault diagnosis results. The experimental results demonstrate that the proposed model outperforms alternative approaches, achieving an accuracy of 95.00% and an F₁ score of 90.48%. Compared to the widely used long and short-term memory artificial neural network (LSTM), the proposed model exhibits improvements of 4.06% in accuracy and 7.85% in F₁ score. Furthermore, it surpasses the latest existing ensemble model based on the improved sparrow algorithm, with enhancements of 1.56% in accuracy and 2.91% in F₁ score. This work presents a data-driven tool with high accuracy for the fault diagnosis of main circulation pumps, which plays a critical role in maintaining the operational stability of VSG-HVDC systems and satisfying the unmanned requirements of offshore flexible platform cooling systems. Full article

Fifth-generation (5G) networks offer high-speed data transmission with low latency, increased base station volume, improved quality of service (QoS), and massive multiple-input–multiple-output (M-MIMO) channels compared to 4G long-term evolution (LTE) networks. However, the COVID-19 pandemic has disrupted the achievement of mobility and handover (HO) in 5G networks due to significant changes in intelligent devices and high-definition (HD) multimedia applications. Consequently, the current cellular network faces challenges in propagating high-capacity data with improved speed, QoS, latency, and efficient HO and mobility management. This comprehensive survey paper specifically focuses on HO and mobility management issues within 5G heterogeneous networks (HetNets). The paper thoroughly examines the existing literature and investigates key performance indicators (KPIs) and solutions for HO and mobility-related challenges while considering applied standards. Additionally, it evaluates the performance of current models in addressing HO and mobility management issues, taking into account factors such as energy efficiency, reliability, latency, and scalability. Finally, this paper identifies significant challenges associated with HO and mobility management in existing research models and provides detailed evaluations of their solutions along with recommendations for future research. Full article

Rock climbing has evolved from a method for alpine mountaineering into a popular recreational activity and competitive sport. Advances in safety equipment and the rapid growth of indoor climbing facilities has enabled climbers to focus on the physical and technical movements needed to elevate performance. Through improved training methods, climbers can now achieve ascents of extreme difficulty. A critical aspect to further improve performance is the ability to continuously measure body movement and physiologic responses while ascending the climbing wall. However, traditional measurement devices (e.g., dynamometer) limit data collection during climbing. Advances in wearable and non-invasive sensor technologies have enabled new applications for climbing. This paper presents an overview and critical analysis of the scientific literature on sensors used during climbing. We focus on the several highlighted sensors with the ability to provide continuous measurements during climbing. These selected sensors consist of five main types (body movement, respiration, heart activity, eye gazing, skeletal muscle characterization) that demonstrate their capabilities and potential climbing applications. This review will facilitate the selection of these types of sensors in support of climbing training and strategies. Full article