Transformer has shown impressive performance on global feature modeling in many applications. However, two drawbacks induced by its intrinsic architecture limit its application, especially in fault diagnosis. First, the quadratic complexity of its self-attention scheme extremely increases the computation cost, which poses a challenge to apply Transformer to a computationally limited platform like an industry system. In addition, the sequence-based modeling in the Transformer increases the training difficulty and requires a large-scale training dataset. This drawback becomes serious when Transformer is applied in fault diagnosis where only limited data is available. To mitigate these issues, we rethink this common approach and propose a new Transformer, which is more suitable for fault diagnosis. In this article, we first show that the attention module can be actually replaced with or even surpassed by a convolution layer under some conditions in mathematics and experiments. Then, we adopt the convolutions into the Transformer, where the computation burden issue is alleviated and the fault classification accuracy is significantly improved. Furthermore, to increase the computation efficiency, a lightweight Transformer called LiteFormer, is developed by utilizing the depth-wise convolutional layer. Extensive experiments are carried out on four datasets: Case Western Reserve University dataset; Paderborn University dataset; and two gearbox datasets of drivetrain dynamic simulator. Through our experiments, our LiteFormer not only reduces the computation cost in model training, but also sets new state-of-the-art results, surpassing other counterparts in both fault classification accuracy and model robustness.

With the increasing demand for location-based services (LBS) in daily production and life, indoor positioning has gradually become one of the current research hotspots. The most pressing challenge in indoor positioning is achieving high-precision location in complex indoor environments. Aiming at the problem that severe non-line-of-sight (NLOS) interference with UWB positioning accuracy in complex indoor environments, this article proposes a novel indoor UWB NLOS correction positioning method based on anchor LOS/NLOS map, which utilizes prior environmental information. After anchors are deployed, the LOS/NLOS map is generated using the prior environmental information. Then the line-of-sight (LOS) anchors are selected for position solution, and the observations under the influence of the NLOS anchors are corrected. Then the trajectory is optimized using a Map-Modified Kalman filtering algorithm, and it adopts a method to solve the position of two LOS anchors by using trajectory constraints, which can deal with more complex indoor environment. The experimental results show that the NLOS correction method in this article has better accuracy than the traditional filtering algorithm, especially in NLOS environment with few anchors. Compared with the Standard Kalman filtering algorithm, the average positioning accuracy is improved by 69.30%, and compared with the robust Kalman filtering algorithm, the average positioning accuracy is improved by 33.77%.

The driving behavior of bus drivers is related to the safety of all passengers and regulation of urban traffic. In order to analyze the relevant characteristics of speed and acceleration, accurate bus trajectories and patterns are essential for driver behavior analysis and development of effective intelligent public transportation. Exploiting real-time vehicle tracking, this paper develops a platform with vehicle-mounted terminals using differential global navigation satellite system (DGNSS) modules for driver behavior analysis. The DGNSS traces were used to derive the vehicle trajectories, which were then linked to road information to produce speed and acceleration matrices. Comprehensive field tests were undertaken on multiple bus routes in urban environments. The spatiotemporal results indicate that the platform can automatically and accurately extract the driving behavior characteristics. Furthermore, the platform’s visual function can be used to effectively monitor driving risks, such as speeding and fierce acceleration, in multiple bus routes. The details of the platform’s features are provided for intelligent transport system (ITS) design and applications.

This paper carried out a series of designs, simulations and implementations by using the physical-like mechanism of a bionic quadruped robot dog as a vehicle. Through an investigation of the walking mechanisms of quadrupeds, a bionic structure is proposed that is capable of omnidirectional movements and smooth motions. Furthermore, the kinematic and inverse kinematic solutions based on the DH method are explored to lay the foundation for the gait algorithm. Afterward, a classical compound pendulum equation is applied as the foot-end trajectory and inverse kinematic solutions are combined to complete the gait planning. With appropriate foot–ground contact modeling, MATLAB and ADAMS are used to simulate the dynamic behavior and the diagonal trot gait of the quadruped robot. Finally, the physical prototype is constructed, designed and debugged, and its performance is measured through real-world experiments. Results show that the quadruped robot is able to balance itself during trot motion, for both its pitch and roll attitude. The goal of this work is to provide an affordable yet comprehensive platform for novice researchers in the field to study the dynamics, contact modeling, gait planning and attitude control of quadruped robots.

Over the past few decades, wearable exoskeletons of various forms have been developed to assist human activities or for rehabilitation of movement disorders. However, sustainable exoskeletons with efficient energy harvesting devices still have not been fully explored. In this paper, we propose the design of a lightweight wearable Bowden-cable-actuated soft exoskeleton robot with energy harvesting capability. Unlike previous wearable exoskeletons, the presented exoskeleton uses an electromagnetic generator to both harvest biomechanical energy and to output mechanical torque by controlling an operation mode relay switch based on a human’s gait. Moreover, the energy-harvesting module also acts as a knee impact absorber for the human, where the effective damping level can be modulated in a controlled manner. The harvested energy is regulated and stored in super capacitors for powering wireless sensory devices when needed. The experimental results show an average of a 7.91% reduction in thigh muscle activity, with a maximum of 3.2 W of electric power being generated during movement downstairs. The proposed design offers important prospects for the realization of lightweight wearable exoskeletons with improved efficiency and long-term sustainability.

Device-free localization (DFL) is a technique of locating targets via wireless sensor networks (WSNs) without carrying any electronic devices. This article introduces a device-free vehicle localization system that can determine the position and 2-D size of a dynamic vehicle in a monitoring region via ultra wide band (UWB) networks. Specifically, our system involves an obstructed recognition model for detecting the state of links, a state-input radio tomography imaging (SI-RTI) algorithm for reconstructing a high-resolution image, a contour threshold method for predicting the size of a vehicle target, and a gradient-advance-retreat (GAR) optimization algorithm for improving accuracy. Field experiments show that our system can estimate the position and size of a moving vehicle with a location root-mean-square error (RMSE) of less than 1.1 m and a root-mean-square overlap degree of more than 75% between the prediction and the real target.