Aiming at the problems of slow detection speed, large prediction error and weak environmental adaptability of current vehicle collision warning system, this paper proposes a recognition method of slippery road surface and collision warning system based on deep learning. Firstly, this paper uses the on-board camera to monitor the environment and road conditions in front of the vehicle in real time, and a residual network model FS-ResNet50 is proposed, which integrated SE attention mechanism and multi-level feature information based on the traditional ResNet50 model. The FS-ResNet50 model is used to identify the slippery states of the current road, such as wet and snowy. Secondly, the yolov5 algorithm is used to detect the position of the vehicle in front, and a driving safety distance model with adaptive traffic environment characteristics is established based on different road conditions and driving conditions, and an early warning area that dynamically changed with the speed and the road slippery states is generated. Finally, according to the relationship between the warning area and the position of the vehicle, the possible collision is predicted and timely warned. Experimental results show that the method proposed in this paper improves the overall warning accuracy by 6.72% and reduces the warning false alarm rate for oncoming traffic on both sides by 16.67% compared with the traditional collision warning system. It can ensure safe driving, especially in bad weather conditions and has a high application value.

Based on small scale sample of accident data from specific scenarios, fully exploring the potential influencing factors of the severity of traffic accidents has become a key and effective research method. In order to analyze the factors mentioned above in the scenario of urban ring roads, this paper collected data records of 1250 traffic accidents involving different severity on urban ring road of a central city in northwest China in the past 3 years. Firstly, the Support Vector Machine (SVM) model of non-parametric method is utilized to analyze the data above, and three kernel functions of linear, inhomogeneous polynomial and Gaussian radial basis are constructed respectively. Considering comprehensively 16 potential influencing factors covering the driver-vehicle-road-environment integrated system, the SVM models of above three kernel functions are verified, accuracy reaches 0.771 and F1 reaches 0.841. Then, Bayesian Optimization (BO), Grids Search (GS) and Rough Set (RS) are utilized as optimizer to adjust the parameters of Gaussian radial basis function SVM model, the performance of BO-SVM is further improved and reaches the optimum, with an average accuracy of 0.875 on the test set and a F1 of 0.886, completely outperforming the benchmark models of GS-SVM, RS-SVM, Bilayer-LSTM and BP. Finally, the sensitivity analysis method is utilized to quantify the sensitivity of the potential influencing factors to the severity of road accidents, and the backward selection method is utilized to screen the core influencing factors that influence the severity of accident, concluded that core influencing factors are age, driving mileage and vehicle type. This paper will provide reference for the analysis of the significant influencing factors for road accidents severity, and to provide theoretical support for the precise formulation of accident improvement strategies.

In the evolving landscape of mixed human-machine driving environments, autonomous vehicles (AVs) confront the challenge of anticipating the lane-changing intentions and subsequent driving trajectories of neighboring vehicles. This capability is essential for optimizing safety, efficiency, and comfort in decision-making processes. This paper introduces a novel hybrid prediction model, the LSTM-GAT-Bilayer-GRU, which leverages deep learning to enhance predictive accuracy and real-time responsiveness in dynamic traffic scenarios. The proposed model consists of two main components: a lane change prediction model (LSTM-GAT) and a trajectory prediction model (G-BiLayer-GRU), to process and predict complex vehicular interactions and environmental dynamics effectively. The efficacy of this integrated model was tested using the HighD dataset for training, validation, and testing purposes. The results of a benchmark analysis indicate that the proposed model demonstrated superior prediction performance and reliability over the Support Vector Machine (SVM), Random Forest (RF), AlexNet and Back-Propagation Through Time (BPTT) in the context of lane change intention recognition. Combining LSTM for temporal data processing with GAT for spatial interaction analysis, along with the GRU’s precise trajectory prediction, achieved the best error evaluation metric and balanced prediction time consuming metric under the six prediction time-interval, marks a substantial advancement in AVs technology. This integration guarantees smooth operation of AVs in intricate driving scenarios, fine-tuning their reactions to bolster road safety and passenger comfort.

The continuous increase in the penetration rate of autonomous vehicles in highway traffic flow has become an irreversible development trend; in this paper, a novel hybrid prediction model of deep sequence learning and an integrated decision tree is proposed for human–machine mixed driving heterogeneous traffic flow scenarios, so as to realize the accurate prediction of the driving intention of the target vehicle in the traffic environment by autonomous vehicles (AVs). Firstly, the hybrid model uses the attention mechanism-based double-layer gated network model (Bilayer-GRU-Att) to effectively capture the time sequence dependence of the target vehicle’s driving state, and then accurately calculate its trajectory data in different prediction time-domains (t_pred). Furthermore, the hybrid model introduces the eXtreme Gradient Boosting decision tree optimized by the Grey Wolf Optimization model (GWO-XGBoost) to identify the lane-changing intention of the target vehicle, because the prediction information of the future trajectory data of the target vehicle by the aforementioned Bilayer-GRU-Att model is properly integrated. The GWO-XGBoost model can accurately predict the lane-changing intention of the target vehicle in different prediction time-domains. Finally, the efficacy of this hybrid model was tested using the HighD dataset for training, validation, and testing purposes. The results of a benchmark analysis indicate that the hybrid model proposed in this paper has the best error evaluation index and balanced prediction time consuming index under the six prediction time-domains. Meanwhile, the hybrid model demonstrates the best classifying performance in predicting the lane-changing intentions of “turning left”, “going straight”, and “turning right” driving behaviors.

As a core application of V2X technology, vehicle–road collaboration enables dynamic coordination among road users (pedestrians, vehicles), infrastructure, and networks through real-time, omnidirectional information exchange. This system represents a pivotal solution for addressing critical transportation challenges, including traffic congestion, safety risks, and environmental sustainability. Its experimental teaching, as the core linkage of theoretical innovation and technical verification, is of vital importance to the cultivation of intelligent transportation talents. Compared with traditional experimental teaching, the V2X semi-physical simulation platform effectively reduces capital investment, completely eliminates the safety risks of actual road tests, and emulates the real traffic environment. To verify the teaching effectiveness of this platform, based on the OBE concept and the BOPPPS teaching method, this study constructed an experimental curriculum framework driven by learning goals and conducted an empirical analysis taking global path planning as an example. Teaching evaluation adopts a combination of subjective and objective methods: Subjective evaluation is conducted through questionnaire surveys, and the proportion of those satisfied with the teaching effect reached more than 80%. The objective evaluation consists of eight performance indicators before class, during class and after class. Through reliability analysis, the performance of students in the observation group was shown to increase by 17.39% compared with that in the control group. The results show that the experimental teaching mode based on the V2X semi-physical simulation platform significantly improves the teaching effectiveness of the vehicle–road collaboration course compared with traditional methods.

To address the challenge of lane-changing prediction for autonomous vehicles (AVs) within highway weaving sections characterized by dynamic interactions in human-machine mixed traffic flows, this paper proposes a multi-expert collaborative prediction model based on a dynamic conflict field (DCF) and Generative Adversarial Networks, termed G-MoE-WGAN. This model quantifies the dynamic game-theoretic intensity among interacting vehicles via the DCF. The proposed model constructs a Mixture of Experts (MoE) system to facilitate intention decision-making and trajectory generation, employing adversarial training to optimize the physical plausibility and interaction adaptability of the prediction results. Furthermore, a Physics-Informed Neural Network (PINN) is introduced for the reconstruction and smoothing of raw naturalistic driving trajectories to validate the model. Experimental results demonstrate that the G-MoE-WGAN achieves a lane-changing intention classification accuracy of 94.35%, representing an improvement of 3.89% to 16.55% compared to baseline models. Within a 3-second prediction horizon, the Final Displacement Error (FDE) and Average Displacement Error (ADE) of the proposed model are significantly reduced by 8.61%–40.55% and 3.21%–40.27%, respectively. In the 5-second long-term prediction, benefiting from the dynamic expert fusion mechanism, the ADE and FDE metrics exhibit a reduction in error of 3.19%–39.29% relative to comparative methods. The study indicates that the proposed conflict potential representation and the multi-expert adversarial training mechanism effectively capture the spatiotemporal heterogeneity of high-dynamic interactions in weaving sections significantly enhancing the robustness and interpretability of lane-changing predictions in complex scenarios.