Uncertain characteristics along with unknown exterior disturbances are fundamental issues in the path‐following control of autonomous ground vehicles (AGVs). Here, a novel robust non‐singular terminal sliding mode (NTSM) control method based on a recurrent neural network (RNN) structure is proposed to enhance the path‐following performance of AGVs. First, based on the dynamic model and path‐following model of AGVs, the robust NTSM steering controller is proposed to ensure that the lateral offset converges to zero within a limited time, and concurrently suppresses the chattering phenomenon that occurs in the regular sliding mode control (SMC) strategy. Then, the RNN is applied to approximate the unknown dynamics section of the system online. With an implied feedback loop, its study ability precedes the regular neural network to catch up with the dynamic response. In addition, the adaptive learning algorithms of the parameters in the RNN are derived using the Lyapunov stability theorem and Taylor linearisation technique to ensure the stability of the closed‐loop system. Finally, the robustness and effectiveness of the proposed control approach against parametric uncertainties and unknown external disturbances are verified in the Simulink‐CarSim simulation platform by comparison with selected reported control methods.

The large-scale preparation of stable graphene aqueous dispersion has been a challenge in the theoretical research and industrial applications of graphene. This study determined the suitable exfoliation agent for overcoming the van der Waals force between the layers of expanded graphite sheets using the liquid-phase exfoliation method on the basis of surface energy theory to prepare a single layer of graphene. To evenly and stably disperse graphene in pure water, the dispersants were selected based on Hansen solubility parameters, namely, hydrophilicity, heterocyclic structure and easy combinative features. The graphene exfoliation grade and the dispersion stability, number of layers and defect density in the dispersion were analysed under Tyndall phenomenon using volume sedimentation method, zeta potential analysis, scanning electron microscopy, Raman spectroscopy and atomic force microscopy characterization. Subsequently, the long-chain quaternary ammonium salt cationic surfactant octadecyltrimethylammonium chloride (0.3 wt.%) was electrolyzed in pure water to form ammonium ions, which promoted hydrogen bonding in the remaining oxygen-containing groups on the surface of the stripped graphene. Forming the electrostatic steric hindrance effect to achieve the stable dispersion of graphene in water can exfoliate a minimum of eight layers of graphene nanosheets; the average number of layers was less than 14. The 0.1 wt.% (sodium dodecylbenzene sulfonate: melamine = 1:1) mixed system forms π–π interaction and hydrogen bonding with graphene in pure water, which allow the stable dispersion of graphene for 22 days without sedimentation. The findings can be beneficial for the large-scale preparation of waterborne graphene in industrial applications.

Computer vision techniques have significantly advanced autonomous driving but still face specific challenges in complex environments, such as accurately detecting small, occluded, or low-contrast objects amidst dynamic backgrounds and varying lighting conditions. This paper introduces FOS-YOLO, a novel model designed to enhance detection performance in these scenarios. FOS-YOLO integrates the Focal Nets mechanism to improve feature representation, effectively addressing target scale variations and low contrast through multi-scale context aggregation and multi-level modulation. Additionally, the Zoomed Spatial Convolutional Block Attention Module (ZS_CBAM), is introduced to improve the detection of small targets, occlusions, and illumination changes by effectively fusing channel and spatial attention with a focus on scaled spatial features. The model also includes two lightweight modules, the Adaptive Depthwise Enhanced Module (ADEM) and the Lite Enhanced Reduction Module (LERM), which reduce parameters and computational load, accelerating convergence and improving accuracy. Experimental results, compared with state-of-the-art methods, show that FOS-YOLO achieves a mAP of 90.4% on the KITTI dataset and 64.3% on the RTTS dataset, with a 20.07% reduction in parameters, significantly enhancing real-time detection accuracy and efficiency.

The method chosen to conduct vehicle dynamic modeling has a significant impact on the evaluation and optimization of ride comfort. This paper summarizes the current modeling methods of ride comfort and their limitations. Then, models based on nonlinear damping and equivalent damping and the multibody dynamic model are developed and simulated in Matlab/Simulink and Adams/Car. The driver seat responses from these models are compared, showing that the accuracy of the ride comfort model based on nonlinear damping is higher than the one based on equivalent damping. To improve the reliability of ride comfort optimization and analysis, a ride comfort optimization method based on nonlinear damping and intelligent algorithms is proposed. The sum of the frequency‐weighted RMS of the driver seat acceleration, the RMS of dynamic tyre load, and suspension working space is taken as the objective function in this article, using nonlinear damping coefficients and stiffness of suspension as design variables. By applying the particle swarm optimization (PSO), cuckoo search (CS), dividing rectangles (DIRECT), and genetic algorithm (GA), a set of optimal solutions are obtained. The method efficiency is verified through a comparison between frequency‐weighted RMS before and after optimization. Results show that the frequency‐weighted RMS of driver seat acceleration, RMS values of the suspension working space of the front and rear axles, and RMS values of the dynamic tyre load of front and rear wheels are decreased by an average of 27.4%, 21.6%, 25.0%, 19.3%, and 22.3%, respectively. The developed model is studied in a pilot commercial vehicle, and the results show that the optimization method proposed in this paper is more practical and features improvement over previous models.

Forward collision warning (FCW) is a critical technology to improve road safety and reduce traffic accidents. However, the existing multi-sensor fusion methods for FCW suffer from a high false alarm rate and missed alarm rate in complex weather and road environments. For these issues, this paper proposes a decision-level fusion collision warning strategy. The vision algorithm and radar tracking algorithm are improved in order to reduce the false alarm rate and omission rate of forward collision warning. Firstly, this paper proposes an information entropy-based memory index for an adaptive Kalman filter for radar target tracking that can adaptively adjust the noise model in a variety of complex environments. Then, for visual detection, the YOLOv5s model is enhanced in conjunction with the SKBAM (Selective Kernel and Bottleneck Attention Mechanism) designed in this paper to improve the accuracy of vehicle target detection. Finally, a decision-level fusion warning fusion strategy for millimeter-wave radar and vision fusion is proposed. The strategy effectively fuses the detection results of radar and vision and employs a minimum safe distance model to determine the potential danger ahead. Experiments are conducted under various weather and road conditions, and the experimental results show that the proposed algorithm reduces the false alarm rate by 11.619% and the missed alarm rate by 15.672% compared with the traditional algorithm.