In the industrial environment, Autonomous Guided Vehicles (AGVs) generally run on a planned route. Among trajectory-tracking algorithms for unmanned vehicles, the Pure Pursuit (PP) algorithm is prevalent in many real-world applications because of its simple and easy implementation. However, it is challenging to decelerate the AGV’s moving speed when turning on a large curve path. Moreover, this paper addresses the kidnapped-robot problem occurring in spare LiDAR environments. This paper proposes an improved Pure Pursuit algorithm so that the AGV can predict the trajectory and decelerate for turning, thus increasing the accuracy of the path tracking. To solve the kidnapped-robot problem, we use a learning-based classifier to detect the repetitive pattern scenario (e.g., long corridor) regarding 2D LiDAR features for switching the localization system between Simultaneous Localization And Mapping (SLAM) method and Odometer method. As experimental results in practice, the improved Pure Pursuit algorithm can reduce the tracking error while performing more efficiently. Moreover, the learning-based localization selection strategy helps the robot navigation task achieve stable performance, with 36.25% in completion rate more than only using SLAM. The results demonstrate that the proposed method is feasible and reliable in actual conditions.

This paper examines the environmental Kuznets curve (EKC) in Vietnam between 1977 and 2019. Using the autoregressive distributed lag (ARDL) approach, we find an inverted N-shaped relation between economic growth and carbon dioxide emissions in both the long- and short-run. The econometric results also reveal that energy consumption and urbanization statistically positively impact pollution. The long-run Granger causality test shows a unidirectional causality from energy consumption and economic growth to pollution while there is no causal relationship between energy consumption and economic growth. These suggest some crucial policies for curtailing emissions without harming economic development. In the second step, we also employed the back-propagation neural networks (BPN) to compare the work of econometrics in carbon dioxide emissions forecasting. A 5-4-1 multi-layer perceptron with BPN and learning rate was set at 0.1, which outperforms the ARDL’s outputs. Our findings suggest the potential application of machine learning to notably improve the econometric method’s forecasting results in the literature.

This study introduces a novel frequency-based attention network (FANet) designed to acquire frequency information in surface electromyographic signals (sEMG) while mitigating potential high-frequency disturbances. The key innovation is the utilization of the discrete cosine transform (DCT) in dedicated attention modules, enabling the conversion of sEMG features to the frequency domain and facilitating the extraction of spectral characteristics. This attention module replaces the self-attention module within the multilayer perceptron (MLP) based transformer, offering a streamlined means for internal frequency features analysis in deep learning models. To evaluate the proposed models, we focus on the dominance of the “rest” gesture in available NinaPro DB4 and NinaPro DB5 datasets. Consequently, we employ common evaluation metrics tailored to the class-imbalanced hand gesture classification task. Through comprehensive experiments and ablation studies, the results confirm the effectiveness of the proposed network in capturing frequency insights, leading to substantial improvement in hand gesture recognition. Notably, the DCT-based attention modules achieve accuracies of 78.70% and 89.64% in the NinaPro DB4 and DB5 datasets, respectively, while maintaining slight computational cost with log-linear complexity.

This study introduces a novel approach for real-time hand gesture classification through the integration of a Frequency-based Attention Neural Network (FANN) with Subject-Adaptive Transfer Learning, specifically tailored for surface electromyography (sEMG) data. By utilizing the Fourier transform, the proposed methodology leverages the inherent frequency characteristics of sEMG signals to enhance the discriminative features for accurate gesture recognition. Additionally, the subject-adaptive transfer learning strategy is employed to improve model generalization across different individuals. The combination of these techniques results in an effective and versatile system for sEMG-based hand gesture classification, demonstrating promising performance in adapting individual variability and improving classification accuracy. The proposed method's performance is evaluated and compared with established approaches using the publicity available NinaPro DB5 dataset. Notably, the proposed simple model, coupled with frequency-based attention modules, achieves accuracies of 89.56% with a quick prediction time of 5ms, showcasing its potential for dexterous control of robots and bionic hands. The findings of this research contribute to the advancement of gesture recognition systems, particularly in the domains of human-computer interaction and prosthetic control.

This article introduces a self-balancing bicycle utilizing a reaction wheel for stability control. To manage hardware complexity, a simplified test platform was first used for physical modeling and controller design with a linear-quadratic regulator (LQR), validating the concept. For the final implementation, the full-scale bicycle adopts full-state feedback. Although the LQR improved stability, its frequent motor speed adjustments led to high power consumption. To address this, a novel adaptive equilibrium point gradual algorithm (AEPGA) was developed to adapt to environmental changes, such as lateral imbalances. Performance tests on a real bicycle showed the AEPGA achieved maximum speed of 17 km/h on campus roads. In tests with a 2 kg lateral payload, AEPGA achieved a 67.456% reduction in power consumption compared to LQR over 22 s. For path tracking, a 3D LiDAR system was integrated. On a 77 m indoor corridor, the system reached an average speed of 1.39 km/h with a 0.153 m mean absolute error, while on a 190 m outdoor asphalt road, it reached 2.31 km/h with 0.283 m MAE. These results demonstrate the effectiveness of AEPGA in enhancing energy efficiency and adaptability, while maintaining reliable path tracking in varied environments.

Accurate segmentation of drivable areas and road obstacles is critical for autonomous mobile robots to navigate safely in indoor and outdoor environments. With the fast advancement of deep learning, mobile robots may now perform autonomous navigation based on what they learned in the learning phase. On the other hand, existing techniques often have low performance when confronted with complex situations since unfamiliar objects are not included in the training dataset. Additionally, the use of a large amount of labeled data is generally essential for training deep neural networks to achieve good performance, which is time-consuming and labor-intensive. Thus, this paper presents a solution to these issues by proposing a self-supervised learning method for the drivable areas and road anomaly segmentation. First, we propose the Automatic Generating Segmentation Label (AGSL) framework, which is an efficient system automatically generating segmentation labels for drivable areas and road anomalies by finding dissimilarities between the input and resynthesized image and localizing obstacles in the disparity map. Then, we train RGB-D datasets with a semantic segmentation network using self-generated ground truth labels derived from our method (AGSL labels) to get the pre-trained model. The results showed that our AGSL achieved high performance in labeling evaluation, and the pre-trained model also obtains certain confidence in real-time segmentation application on mobile robots.