Object‐level semantic simultaneous localization and mapping (SLAM) plays a vital role in enabling autonomous systems to perceive, understand, and interact with complex environments. However, conventional methods often suffer from low geometric modeling accuracy, strong reliance on predefined object categories, and poor generalization to unseen objects. To address these challenges, we propose NS‐SLAM, a novel framework that integrates neural radiance fields (NeRFs) with a zero‐shot segmentation foundation model (SEEM). NS‐SLAM achieves high‐precision object‐level reconstruction and localization via a hash‐encoded NeRF representation while leveraging SEEM to provide dense, pixel‐level semantic segmentation without requiring prior object knowledge. The semantic information is extracted by feeding high‐fidelity NeRF‐rendered views into the SEEM model, followed by a back‐projection strategy to associate 2D semantics with 3D geometry. Extensive experiments on the TUM RGB‐D dataset and a newly collected office dataset demonstrate that NS‐SLAM significantly outperforms representative baselines such as QISO‐SLAM and QuadricSLAM. Specifically, NS‐SLAM reduces the absolute trajectory error (ATE) root mean squared error (RMSE) by 8.2% on average and improves 3D modeling quality metrics (depth L1, peak signal‐to‐noise ratio [PSNR], and SSIM) by over 10%, while demonstrating robust zero‐shot semantic recognition capabilities, successfully identifying and segmenting both known and unseen object categories. These results validate the effectiveness of NS‐SLAM in enhancing both the precision and generalization capabilities of object‐level semantic SLAM, offering a promising solution for robotic applications in complex environments with unknown objects.
Multiparameter fiber optic sensors face a fundamental challenge: cross‐sensitivity between measured parameters compromises accuracy and limits practical deployment. We present a triple‐cavity cascade architecture combining two fiber Bragg gratings (FBGs) and a Fabry–Pérot interferometer (FPI) for the simultaneous measurement of temperature, pressure, and humidity with exceptional numerically‐predicted discrimination performance. The system exploits complementary sensitivity models in a sensitivity matrix with determinant 2.59 and condition number 644, enabling robust parameter recovery through matrix inversion enhanced by a 3–8–8–3 neural network. A comprehensive numerical simulation over the full operating ranges (0–100°C, 100–500 kPa, 5–95%RH) demonstrates simulated accuracies of ±0.011°C in temperature, ±0.012 kPa in pressure, and ±0.021%RH in humidity, with all six cross‐sensitivity pairs suppressed below 1%. This represents a numerically predicted improvement potential of 5×–15× compared to existing two‐parameter systems and is, to our knowledge, the first numerically demonstrated three‐parameter design based on rigorous theoretical analysis. Error budget analysis identifies coating hysteresis (80%) as the primary accuracy limiter for humidity, while wavelength noise (72%) and phase noise (62%) dominate the temperature and pressure uncertainties, respectively. Dynamic response tests over 100 time steps confirm root mean square error (RMSE) values of 0.071°C, 0.313 kPa, and 0.244%RH under simultaneous sinusoidal, step, and random parameter variations. Although experimental validation remains pending, the rigorous theoretical framework provides a model for next‐generation environmental sensing with applications in aerospace monitoring, industrial process control, and precision instrumentation.
Automated pan evaporation (APEV) measurement based on Internet of Things (IoT)–enabled sensors and precision irrigation (PI) for crop production is not commonly used in crop fields, and its performance has not yet been studied in developing countries. Soil sensors are costly and inaccurate in variable soils. Similarly, traditional pan evaporimeters require labor‐intensive manual measurements. Therefore, the objective of this study was to develop and test the APEV measurement using an IoT‐enabled sensor and PI for crop production. The IoT‐based APEV measurement system for PI was developed, tested, and evaluated for eggplant production under field conditions. The field evaluation test was conducted in a randomized block design with two treatments and two replication blocks. The treatments consisted of APEV measurement and automated irrigation (IR) with a drip system ( T 1 ) and conventional surface flooding IR applied up to field capacity ( T 2 ). The treatment T 1 system was structured into three key components: (i) hardware, (ii) software and mobile app integration, and (iii) field validation of the developed system. The hardware system consisted of an ESP32, a solenoid valve, an ultrasonic sensor, a pan evaporimeter, a water flow sensor, and water supply‐related drip IR accessories. The T 1 system consisted of an ultrasonic sensor, a sensor node for collecting real‐time APEV values from the field, and an automation node for controlling IR. An IoT‐based mobile app (Blynk) was used for real‐time monitoring and operating the IR system’s switch on or off, while ThingSpeak enabled data storage, analysis, and real‐time data visualization and controlled the IR system. The T 1 system automatically measured the water level from the pan evaporimeter and supplied water to the crop field without manual intervention. The T 1 system produced a nearly identical yield, saved IR water by 50.6%, and achieved an IR water use efficiency (IWUE) and crop water use efficiency (WUE) by 17.9 and 16.6 kg/m 3 , compared with 9.0 and 8.3 kg/m 3 , respectively, under the traditional surface flooding IR system ( T 2 ). This newly developed IoT‐enabled PI based on the APEV IR method, using an ultrasonic sensor, needs further validation across diverse field conditions.
This paper introduces a secure workplace health safety system that prioritizes employee well-being. It leverages an artificial intelligence of things (AIoT) framework for smart attendance monitoring and workplace health safety, which is especially important in the post-COVID-19 era. Early detection of health issues can help prevent outbreaks. The system utilizes a fingerprint scanner for biometric attendance and a low-cost wireless sensor network (WSN) for data collection. It continuously monitors vital signs, such as blood oxygen level, heart rate, and body temperature, to identify potential contagious diseases. Data is accessible through the Blynk app (mobile) and website (desktop), enabling remote monitoring. Additionally, the system logs data to Google Sheets for easy access and provides security for system administrators by sending them notifications for new user data entries. A hardware prototype has been built and its performance has been successfully validated. Again, a machine learning (ML) model trained on the system's data can detect potential COVID-19 cases. The system tested several ML algorithms on its dataset, and a decision tree (DT) classifier achieved the highest accuracy (96%). This intelligent system effectively monitors attendance, and health data and detects potential COVID-19 cases with high accuracy. It is user-friendly, affordable, and accessible, offering significant benefits to both employees and employers in the post-pandemic world.
Accurate prediction of brake pad remaining useful life (RUL) represents a persistent bottleneck in automotive predictive maintenance, as current solutions remain trapped in an inherent trade-off among predictive accuracy, hardware cost, and material adaptability. Traditional monitoring schemes either depend on costly proprietary sensors or adopt static feature fusion strategies, which lack generalization across heterogeneous friction materials and exhibit poor robustness to industrial interferences. This work establishes a material-adaptive visual wear evaluation framework empowered by the SD-DA-TOC hybrid optimizer, which synergistically combines storm optimization, differential evolution (DE), and simulated annealing to realize physics-guided dynamic weight learning for multisource feature fusion. In contrast to mainstream vision-based methods that ignore material discrepancies, the proposed framework embeds material-specific properties directly into the fusion mechanism, allowing consistent performance across different brake pad formulations without manual tuning. Validated on four types of industrial brake pads, the proposed approach yields an R2 of 0.92 for RUL prediction, lowers prediction error by more than 40% relative to benchmark algorithms, and maintains stability under noise and uneven illumination. The proposed framework effectively overcomes the long-standing incompatibility among accuracy, affordability, and cross-material generalization, providing a technically sound and deployable solution for intelligent brake pad health management.
