Shufeng Wang

Vehicle interactions at unsignalized intersections are complex, and crossing conflicts may differ considerably in their potential severity. This study develops a trajectory-based framework for identifying and analyzing vehicle crossing conflict severity using data from the Bendplatz intersection in the inD dataset. Post-encroachment time (PET) and time to the second conflict point (T2) are applied as complementary criteria to identify 370 crossing conflict pairs, which are classified into Minor, Moderate and Severe levels based on relative speed. Thirteen explanatory variables are constructed from the pre-conflict motion states of both vehicles, the road-priority relationship, and short-term traffic exposure. Five classification models are compared using nested stratified cross-validation. LightGBM achieves the best overall performance, with an Accuracy of 0.8162, a Macro-F1 of 0.7850, and a macro-averaged area under the receiver operating characteristic curve (MAUC) of 0.9326. Out-of-fold SHapley Additive exPlanations (SHAP) analysis shows that the average speeds of the first-passing and later-passing vehicles together account for 48.09% of the summed global variable importance, while the lateral-motion state of the later-passing vehicle also makes an important predictive contribution. SHAP interaction analysis further reveals nonlinear interaction patterns involving the speed states of both vehicles and the lateral adjustment of the later-passing vehicle. Sensitivity analyses using alternative severity thresholds and traffic-exposure windows show that the principal variable-importance pattern remains broadly stable. The proposed framework can support event-level conflict screening and provide quantitative evidence for subsequent field investigation and safety monitoring at unsignalized intersections.

Due to the absence of traffic signal control and the difficulty in accurately estimating the future movements of surrounding vehicles, autonomous vehicle decision-making faces challenges at unsignalized intersections. This study proposes a trajectory prediction-aided deep reinforcement learning framework. First, a composite prioritized replay mechanism is introduced into the Twin Delayed Deep Deterministic Policy Gradient (TD3) algorithm, jointly considering temporal-difference error and reward-based event severity to enhance critical-experience reuse. Second, a convolutional multi-layer long short-term memory (CM-LSTM) model predicts surrounding-vehicle trajectories through convolutional local-motion encoding and stacked LSTM temporal modeling, and the predicted trajectories are incorporated into the deep reinforcement learning state representation. A multi-objective reward function is designed to balance collision avoidance, passing efficiency, lane keeping, and task completion. In CARLA go-straight and left-turn tests, CLS-TD3 achieves success rates of 93.8% and 90.2%, collision rates of 2.5% and 4.2%, and average passing times of 5.18 s and 5.58 s. Compared with TD3, the success rates increase by 6.3 and 8.6 percentage points, while average passing times decrease by 18.8% and 20.5%. These results demonstrate that the proposed framework improves the safety and crossing efficiency of autonomous vehicle decision-making at unsignalized intersections.

To address the problems of underutilization of samples and unstable training for intelligent vehicle training in the Twin Delayed Deep Deterministic Policy Gradient (TD3) algorithm, a TD3 algorithm based on the Composite Prioritized Experience Replay (CPR-TD3) mechanism is proposed. It considers experience immediate reward value and Temporal Difference error (TD-error) based and respectively to construct priorities to rank the samples. Subsequently composite average ranking of the samples to recalculate the priorities for sampling, uses the collected samples to train the target network. Then introduces the minimum lane change distance and the variable headway time distance to improve the reward function. Finally, the improved algorithm is proved to be effective by comparing it with the traditional TD3 on the highway scenario, and the CPR-TD3 algorithm improves the training efficiency of intelligent vehicles.

Taillight recognition is the key to predicting the intention of the vehicle ahead, but the current detection algorithms still have problems in small target and taillight matching. In order to solve the above problems, this paper analyzes the characteristics of automobile taillight signal, combines the image processing algorithm with the object detection technology, and puts forward a taillight signal recognition model based on image processing. Firstly, the characteristics of automobile taillight signal are analyzed, and the taillight is extracted by using HSV color space and corrosion expansion algorithm, and then the taillight is matched by artificial preset experience value to solve the problem of matching headlight. Secondly, P2 small target detection model is added to the YOLOv8s model to improve the recognition ability of small targets, and CA(coordinate attention) is inserted to reduce the interference of other light sources. Finally, EIOU Loss was added to solve the problem of sample imbalance caused by a small number of motorcycles. In this paper, the actual scene video is used for ablation experiments to verify the effectiveness of the improved algorithm. The experimental results show that the mAP value of the model is 9.3% higher than that of the YOLOv8s model.

In automatic driving, the recognition of Front-Vehicle taillights plays a key role in predicting the intentions of the vehicle ahead. In order to accurately identify the Front-Vehicle taillights, we first analyze the different characteristics of the vehicle taillight signal, and then propose an improved taillight recognition model based on YOLOv5s. First, CA(coordinate attention) is inserted into the backbone network of YOLOv5s model to improve small target recognition and reduce interference from other light sources; Then, the EIOU Loss is used to solve the class imbalance problem; Finally, EIOU-NMS is used to solve the problem of anchor box error suppression in the recognition process. We use the actual scene video and vehicle taillights dataset to conduct ablation experiments to verify the effectiveness of the improved algorithm. The experimental results show that the mAP value of the model is 9.2% higher than YOLOv5s.