Balance assessment is crucial for health monitoring and rehabilitation evaluation of neurological diseases like Parkinson’s disease (PD) and stroke. The Berg Balance Scale (BBS) is a widely used clinical tool for balance evaluation. However, its dependence on trained therapists for subjective, time-consuming assessments limits its scalability. Current researchers have proposed several automated assessment systems. However, they suffer from difficulty in use in clinical settings and the need for feature engineering. The rapid advancement of wearable inertial measurement units (IMUs) provides an objective tool for motion analysis that is suitable for use in clinical environments. Thus, to address the limitations of manual scoring and complexities of capturing gait features, we proposed an automated BBS assessment system using an attention-based deep learning algorithm with IMU data, integrating convolutional neural networks (CNNs) for spatial feature extraction, bidirectional long short-term memory (Bi-LSTM) networks for temporal modeling, and attention mechanisms to emphasize informative features. Validated with 20 healthy subjects (young and elderly) and 20 patients (PD and stroke), the system achieved a mean absolute error (MAE) of 1.1627 and root mean squared error (RMSE) of 1.5333. Requiring only 5 min of walking data, this approach provided an efficient, objective solution for balance assessment to assist healthcare physicians as well as patients in their own health monitoring. The key limitations included: a limited generalizability to severely impaired patients who were unable to walk independently, and the inability to predict the score of individual tasks.

Intelligent operation and maintenance of urban rail transit systems is essential for improving train safety and efficiency. This study focuses on reducing time, physical effort, and safety risks in deploying intelligent metro inspection robots. This study introduces a design approach for an undercarriage robot simulation system and an offline teaching method. Gazebo and Isaac Sim are combined in this study. Gazebo is used for lightweight simulation in model development and algorithm testing. Isaac Sim is used for high-fidelity rendering and robust simulation in complex large-scale scenarios. This combined approach addresses critical aspects of system development. The research proposes environment data collection and processing methods for metro inspection scenarios. It also provides solutions for hole problems in point cloud mesh models and approaches for robot modeling and sensor configuration. Additionally, it involves developing a target vector labeling platform. Using these elements, an offline teaching system for undercarriage inspection robots has been designed with simulation tools. Offline teaching is unrestricted by on-site space and time. It reduces physical demands and boosts robot teaching efficiency. Experimental results indicate that it takes about 30 s to program a single manipulator motion offline. In contrast, manual on-site teaching takes about 5 min. This represents a significant efficiency improvement. While offline teaching results have some errors, high success rates can still be achieved through error correction. Despite challenges in modeling accuracy and sensor data precision, the simulation system and offline teaching approach decrease metro vehicle operation risks and enhance robot deployment efficiency. They offer a novel solution for intelligent rail transit operation and maintenance. Future research will focus on high-quality environmental point cloud data collection and processing, high-precision model development, and enhancing and expanding simulation system functionality.

Ultrasonic TOFD imaging, as an important non-destructive testing method, has a wide range of applications in pipeline girth weld inspection and testing. Due to the limited bandwidth of ultrasonic transducers, near-surface defects in the weld are masked and cannot be recognized, resulting in poor longitudinal resolution. Affected by the inherent diffraction effect of scattered acoustic waves, defect images have noticeable trailing, resulting in poor transverse resolution of TOFD imaging and making quantitative defect detection difficult. In this paper, based on the assumption of the sparseness of ultrasonic defect distribution, by constructing a convolutional model of the ultrasonic TOFD signal, the Orthogonal Matching Pursuit (OMP) sparse deconvolution algorithm is utilized to enhance the longitudinal resolution. Based on the synthetic aperture acoustic imaging model, in the wavenumber domain, backpropagation inference is implemented through phase transfer technology to eliminate the influence of diffraction effects and enhance transverse resolution. On this basis, the time-domain sparse deconvolution and frequency-domain synthetic aperture focusing methods mentioned above are combined to enhance the resolution of ultrasonic TOFD imaging. The simulation and experimental results indicate that this technique can outline the shape of defects with fine detail and improve image resolution by about 35%.

Autonomous underwater vehicle inspection in 3D environments presents significant challenges in spatial mapping for obstacle avoidance and motion control. Current solutions rely on either 2D forward-looking sonar or expensive 3D sonar systems. To address these limitations, this study proposes a cost-effective 3D reconstruction method using an oscillatory forward-looking sonar with a pan-tilt mechanism that extends perception from a 2D plane to a 75-degree spatial range. Additionally, a polar coordinate-based frontier extraction method for sequential sonar images is introduced that captures more complete contour frontiers. Through bridge pier scanning validation, the system shows a maximum measurement error of 0.203 m. Furthermore, the method is integrated with the Ego-Planner path planning algorithm and nonlinear Model Predictive Control (MPC) algorithm, creating a comprehensive underwater 3D perception, planning, and control system. Gazebo simulations confirm that generated 3D point clouds effectively support the Ego-Planner method. Under localisation errors of 0 m, 0.25 m, and 0.5 m, obstacle avoidance success rates are 100%, 60%, and 30%, respectively, demonstrating the method’s potential for autonomous operations in complex underwater environments.