Pinpin Qin

To address the spatially concentrated loss of relative speed and spacing measurements in expressway tunnel clusters, this route-specific case study evaluates a Projector-Transformer Deep Learning (PTDL) pipeline that adds a linear reconstruction head to a Transformer following-speed predictor. Sixty trajectory segments came from two instrumented-vehicle trials on a 42.086 km route with 15 tunnels. Errors were summarized over tunnel cluster cells (Tce). Coupled variants share an encoder under a weighted dual-task loss. Decoupled variants train separate encoders sequentially, freezing reconstruction during forecasting. Under the reported development records, decoupling reduced missing-feature MSE by 40.7% (from 0.0246 to 0.0146), while complete-feature MSE was nearly identical (0.0075 versus 0.0074). With complete inputs, D-PTDL-S gave a lower ASE than the reported IDM, LSTM, and D-PTDL values over one and two Tce. In the naturally missing subset, its ASE was 45.0–52.6% lower than D-PTDL across one to five Tce. The 60 trajectory segments were partitioned segment-wise into disjoint training, validation and test subsets (80/10/10), with validation used solely for early stopping and the test subset read once after training. The baselines were neither information- nor calibration-matched, no repeated-seed inference was performed, and no ground truth is available within the naturally missing intervals. The results therefore establish route-specific feasibility rather than cross-route generalization. Safety, energy, emissions, and life-cycle outcomes were not measured. Sustainability relevance is limited to operational continuity under incomplete sensing.

To address the significant degradation of steering performance in dual-motor steer-by-wire (DMSBW) systems caused by the coupling of communication delays and motor faults, a robust fault-tolerant control strategy is proposed under the dual-motor collaborative driving mode. First, a matrix polytopic model is employed to describe the nonlinearities introduced by delays, establishing a delay-dependent DMSBW system dynamics model. Second, for electrical faults such as internal motor short circuits that cause a sudden drop in rotational speed, an adaptive motor-switching fault-tolerant mechanism is designed based on a smooth monitoring function to achieve rapid fault detection and steering function reconstruction. Furthermore, considering the coupled impact of delays and faults, a robust linear quadratic regulator (LQR) controller with feedforward compensation is designed to enhance system fault tolerance and robustness. Simulation results demonstrate that under steering wheel angle step input with delays, the proposed strategy achieves a rapid response without significant overshoot, and the steady-state tracking error is significantly reduced. In variable-speed single lane change maneuvers with coupled delays and severe motor faults, the peak and root mean square (RMS) errors of the front wheel angle are reduced to 0.0112 rad and 0.0031 rad, respectively. Compared to the delay-compensated nonlinear model predictive control (NMPC) and sliding mode control (SMC) strategies that do not account for delays, the peak error is reduced by 15.79% and 45.37%, while the RMS error decreases by 27.91% and 35.42%, respectively. Additionally, the peak and RMS errors of the sideslip angle and yaw rate are substantially reduced, validating the strategy’s excellent steering fault tolerance, robustness, and vehicle handling stability.

Lightweight design and crashworthiness of protective structures are critical for battery safety in electric vehicles (EVs). This study addresses the limited research on cross-sectional shape design of high-strength steel double-cell roll-formed tubes (DCRFTs), widely used in EV bumper beams, battery boxes, and electric bus frames. A parametric design method is proposed based on three parameters: middle flange offset (o), upper deflection angle (α), and lower deflection angle (β). Under the constraints of constant cross-sectional height and enclosed area, this method systematically generates diverse shapes, including square, trapezoid, hexagon, re-entrant hexagon, and various hybrid shapes. Validated finite element models were employed to analyze the deformation modes and crashworthiness of DP980 steel DCRFTs under idealized lateral three-point bending with simple supports. The results indicated that the re-entrant hexagon section reduced maximum deformation (Disp) by 2.95%, peak crushing force (PCF) by 9.53%, and improved crushing force efficiency (CFE) by 13.88% compared to the baseline square section. The parametric study and sensitivity analysis confirmed that the offset (o) was the most critical parameter, contributing over 80% of the variance in Disp, PCF, and CFE. Multi-objective optimization using an RBF surrogate model and the NSGA-II algorithm yielded Pareto optimal solutions. Compared to the baseline, three representative solutions achieved Disp reductions of 11.83–25.10% and CFE improvements of 15.63–22.26%, each with distinct trade-offs among objectives. This work establishes a methodological framework for parametric cross-section design of roll-formed profiles; its extension to realistic boundary conditions will further facilitate practical EV protective structure design.

In the free-space optical communication (FSO) domain, orbital angular momentum (OAM) beams are vulnerable to disturbances from composite channels, including atmospheric turbulence and diffuse media, creating significant challenges for accurate mode identification. The present study investigates the transmission characteristics and mode recognition issues of a novel Bessel-Bessel-Gaussian (BBG) vortex beam after traversing a composite channel comprising atmospheric turbulence and diffuse media. Integrating the Fresnel diffraction model with the von Kalman turbulence model enables the numerical simulation of the optical field evolution of BBG beams under varying turbulence intensities and transmission distances. Simulation results in constructing a dataset comprising 9,600 distorted spot images. The evaluation of the recognition performance of multiple deep learning models, including traditional convolutional neural networks (VGG16, ResNet50) and advanced Transformer architectures (Swin-Transformer), is facilitated by this dataset. The simulation results demonstrate that BBG beams exhibit superior shape retention capabilities in complex channels due to their unique double-Bessel function structure. Concerning model performance, the Swin-Transformer model, benefiting from its efficient sliding-window self-attention mechanism, achieved a recognition accuracy of 98.2% under severe conditions of strong turbulence and 1 km transmission distance, significantly outperforming ResNet50 (94.8% ) and VGG16 (82% ). This study corroborates the practical application of BBG beams as highly robust information carriers. It validates the Swin-Transformer architecture’s superiority as an efficient demodulation tool under complex channel conditions, providing a novel solution for constructing high-capacity, highly reliable OAM-multiplexed FSO systems.

To address the conflicts among objectives and the decision-making challenges in the multi-condition adaptive design of battery boxes for new energy vehicles, this study proposes a multi-objective collaborative optimization method based on an improved relaxation factor, aiming to achieve a comprehensive enhancement in both structural lightweighting and mechanical performance. A finite element model of the CTB high-strength steel roll-formed battery box was established and validated through modal testing. According to the Chinese National Standard GB 38031-2025, the mechanical responses of the battery box under random vibration, extreme operating conditions, and impact loads were analyzed to identify performance weaknesses. Sensitivity analysis was conducted to screen the design variables, and an improved relaxation factor strategy based on weight distribution difference information was introduced to construct a multi-objective collaborative optimization model. Furthermore, the entropy-weighted TOPSIS method was employed to enable intelligent decision-making on the Pareto solution set. The results demonstrate that the proposed method outperforms conventional approaches in both convergence speed and solution distribution uniformity. After optimization, the mass of the battery box was reduced by 12.38%, while multiple mechanical performance indicators were simultaneously improved, providing valuable theoretical and engineering guidance for the structural design of power battery systems.

In this study, finite element numerical simulations were used to investigate the deformation of critical vehicle components and the damage characteristics of the retaining wall at the end of the emergency escape ramp after the impact of a Ford 800 truck on the retaining wall of the refuge lane. A finite element model of the reinforced concrete retaining wall of the truck was created using the LS-DYNA (R11.0) program and the correctness of the constructed finite element model was confirmed by tests. The parameters of the reinforced concrete retaining wall were determined using orthogonal tests. Finite element numerical simulations of vehicle impact on the retaining wall were carried out, and the results showed that two stages of deformation occurred at the front and rear sides of the cockpit during the impact process, and the damage of the retaining wall increased with the increase in the vehicle speed, the impact angle, and the bumper stiffness. Punching shear damage occurred in the impact region of the wall and shear damage occurred at the corners of the wall.

Despite the significant impact of network hyperparameters on deep learning car-following models, there has been relatively little research on network hyperparameters of deep learning car-following models. Therefore, this study proposes a car-following model that combines particle swarm optimization (PSO) and gated recurrent unit (GRU) networks. The PSO-GRU car-following model is trained and tested using data from the natural driving database. The results demonstrate that compared to the intelligent driver model (IDM) and the GRU car-following model, the PSO-GRU car-following model reduces the mean squared error (MSE) for the speed simulation of following vehicles by 88.36% and 72.92%, respectively, and reduces the mean absolute percentage error (MAPE) by 64.81% and 50.14%, respectively, indicating a higher prediction accuracy. Dataset 3 from the drone video trajectory database of Southeast University and NGSIM’s I-80 dataset are used to study the car-following model’s cross-dataset adaptability, that is, to verify its transferability. Compared to the GRU car-following model, the PSO-GRU car-following model reduces the standard deviation of the test results by 60.64% and 32.89%, highlighting its more robust prediction stability and better transferability. Verifying the ability of the car-following model to produce the stop-and-go phenomenon can evaluate its transferability more comprehensively. The PSO-GRU car-following model outperforms the GRU car-following model in creating stop-and-go sensations through platoon simulation tests, demonstrating its superior transferability. Therefore, the proposed PSO-GRU car-following model has higher prediction accuracy and cross-dataset adaptability compared to other car-following models.

To address traffic congestion in urban expressways during the transition from wide to narrow sections, this study proposed a car-following strategy based on deep reinforcement learning. Firstly, a car-following strategy was developed based on a twin-delayed deep deterministic policy gradient (TD3) algorithm, and a multi-objective constrained reward function was designed by comprehensively considering safety, traffic efficiency, and ride comfort. Secondly, 214 car-following periods and 13 platoon-following periods were selected from the natural driving database for the strategies training and testing. Finally, the effectiveness of the proposed strategy was verified through simulation experiments of car-following and platoon-following. The results showed that compared to human-driven vehicles (HDV), the TD3 and deep deterministic policy gradient (DDPG)-based strategies enhanced traffic efficiency by over 29% and ride comfort by more than 60%. Furthermore, compared to DDPG, the relative errors between the following distance and desired safety distance using TD3 could be reduced by 1.28% and 1.37% in simulation experiments of car-following and platoon-following, respectively. This study provides a new approach to alleviate traffic congestion for wide-to-narrow road sections in urban expressways.

To explore the potential relationship between the leading vehicle and the following vehicle during car-following, we proposed a novel car-following model combining a convolutional neural network (CNN) with a long short-term memory (LSTM) network. Firstly, 400 car-following periods were extracted from the natural driving database and the OpenACC car-following experiment database. Then, we developed a CNN-LSTM car-following model, and the CNN is employed to analyze the potential relationship between the vehicle’s dynamic parameters and to extract the features of car-following behavior to generate the feature vector. The LSTM network is adopted to save the feature vector and predict the speed of the following vehicle. Finally, the CNN-LSTM model is trained and tested with the extracted car-following trajectories data and compared with the classical car-following models (LSTM model, intelligent driver model). The results show that the accuracy and the ability to learn the heterogeneity of the proposed model are better than the other two. Furthermore, the CNN-LSTM model can accurately reproduce the hysteresis phenomenon of congested traffic flow and apply to heterogeneous traffic flow mixed with adaptive cruise control vehicles on the freeway, which indicates that it has strong generalization ability.

There is limited research on the loss and reconstruction of car-following features. To delve into car-following's characteristics, we propose a car-following model based on LSTM-Transformer. By fully leveraging the advantages of long short-term memory (LSTM) and transformer models, this study focuses on reconstructing the input car-following features. Training and testing were conducted using 700 car-following segments extracted from a natural driving dataset and the Next Generation Simulation (NGSIM) dataset, and the proposed model was compared with an LSTM model and an intelligent driver model. The results demonstrate that the model performs exceptionally well in feature reconstruction. Moreover, compared to the other two models, it effectively captures the car-following features and accurately predicts the position and speed of the following car when features are lost. Additionally, the LSTM-Transformer model accurately reproduces traffic phenomena, such as asymmetric driving behavior, traffic oscillations and lag, by reconstructing the lost features. Therefore, the LSTM-Transformer car-following model proposed in this study exhibits advantages in feature reconstruction and reproducing traffic phenomena compared to other models.