In the context of carbon peaking and carbon neutrality, integrated electric drive axles have emerged as a key pathway for commercial-vehicle electrification. Firstly, the paper introduces the typical configurations and layouts of integrated e-axles for commercial vehicles. Given the complexity and diversity of vehicle segments, it analyzes the suitability of electric drive axles for passenger cars, light trucks and pickups, as well as medium-and heavy-duty trucks. Next, the paper focuses on the motor, inverter, and transmission, which are the three core components, and summarizes recent advances in the key technologies supporting commercial-vehicle e-axles. Finally, the paper discusses the challenges these technologies still pose and describes their future prospects, providing a reference for the development and broader adoption of integrated electric drive axle systems in commercial vehicles.

To address the difficulty of balancing safety and ride comfort, this paper proposes a new safety model considering both factors, and introduces a corresponding early-warning braking strategy. First, the second-order time-to-collision (TTC) model is constructed by incorporating the accelerations of both vehicles into the classical TTC model. Then the safety distance model based on braking process analysis is modified according to road conditions. Combining the safety and comfort requirements, the paper employs the second-order TTC model during the warning phase, and the modified safety distance model during the braking phase to judge dangerous states. The hierarchical AEB control system is designed, and co-simulation and real-vehicle test platforms are built for verification. The results show that, under various test scenarios, the proposed car-following control strategy for AEB systems successfully avoids collisions, initiates braking without disrupting normal driving, provides warning times that give the driver ample reaction time, keeps the minimum inter-vehicle distance after braking within a reasonable range, and achieves effective hazard avoidance even on wet roads.

The rapid development of connected and intelligent vehicles is accelerating the exploration and commercialization of artificial intelligence (AI) technologies. Yet the broader and deeper application of AI in automated driving also brings increasingly prominent safety risks. Thus, developing safety testing and assessment methods for AI-applied automated driving systems is crucial for balancing technological innovation with safety concerns. From a system-safety perspective, this paper proposes a safety assessment method covering three stages: design and development, testing and evaluation, and deployment and operation. The method integrates the life cycle of AI system, safety requirements, verification and validation methods, and continuous risk assessment and safety analysis. Furthermore, the measures for development, design, testing, and optimization to ensure system safety are proposed, providing a reference for future testing and safety assessment of AI-based automated driving systems.

The European New Car Assessment Programme (Euro NCAP) is an important reference for consumers choosing vehicles, and a leading indicator for global advances in automotive safety technology. This paper provides an in-depth interpretation of the latest trends in Euro NCAP testing protocols and compares the latest assessment results. Focusing on the segmentation of the safety-protection assessment systems, the paper reviews research progress in safety assessment techniques throughout the entire process, from safe driving and collision avoidance, to crash protection and post-crash safety. It also summarizes the current status of mainstream assessment systems, discusses the performance and characteristics of leading models, and offers practical guidance for improving China's vehicle-safety evaluation system and supporting the overseas expansion strategies of domestic brands.

Existing driver-assistance systems often deliver late or inaccurate alerts when a vehicle cuts in suddenly from an adjacent lane. To address this issue, the paper develops a collision-warning model that detects the lane-change intention of the lead vehicle. The vehicle-vehicle communication is utilized to send that intention to the following vehicle, which then predicts the cut-in path and performs collision detection. The collision time TTC-S is proposed, the avoidance time TTA is re-examined, and a tiered warning strategy is designed. In order to verify the effectiveness of the collision warning system, a joint simulation platform is built based on Simulink and PreScan. The results show that the collision-warning model achieves an average true-positive rate of 90.32%, outperforming the Mazda model by 8.44% and the traditional TTC model by 11.66%. The system also provides earlier alerts, extending the average warning lead time by 1.42 s and 1.9 s, respectively, which provides a larger safety margin.

In this paper, based on the 2024 C-NCAP evaluation regulations, a finite element model of an SUV front end impacting a pedestrian's leg was established using ANSA software. Numerical simulations were carried out using LS-DYNA, and test data were employed to validate the model. Evaluation of the injury values obtained from the simulation and testing shows that the thigh bending moment and knee ligament elongation comply with the high performance limits of the 2024 C-NCAP, whereas the calf bending moment T1 does not. Further analysis shows that the lower front-end grille presses against the knee, occupies the X-direction energy-absorbing space, limits the deformation of the front bumper to absorb energy, and thus increases the calf bending moment. Two targeted structural improvements were made: the solid structure in the center of the energy-absorbing foam was replaced by an open-cell structure with a small groove at the upper end, and the license-plate mounting bracket was lowered by 10 mm along the X-direction. After these modifications, the calf bending moment T1 drops to 265.1 Nm, meeting the 2024 C-NCAP high-performance limit, and the thigh bending moment and the knee ligament elongation continue to satisfy the same criteria.

To address the inefficiency and weak robustness of existing rollback methods when a remote update of automotive ECU fails, this paper designs an efficient and stable upgrade rollback strategy. Firstly, an optimized rollback backup method and a comprehensive security-verification strategy are proposed to enhance rollback stability. Then, to avoid the long duration of traditional full rollback, a differential rollback technique based on an improved Bsdiff algorithm is proposed. The improved algorithm can significantly increase rollback efficiency by using the higher-ratio LZMA2 (Lempel-Ziv-Markov chain-Algorithm 2) compression and optimizing the patch format. Finally, the proposed strategy is tested on real-vehicles. The results show that the rollback strategy using the improved differential algorithm reduces rollback time by 84.69%, and achieves a 100% test-case pass rate. The proposed strategy preserves vehicle functionality when an ECU upgrade fails while improving rollback efficiency.

To address the lack of objective and quantitative data support in the user-experience evaluation of in-vehicle software usability, the paper proposed an evaluation method that blends subjective and objective data. The subjective psychological responses and objective physiological signals captured during the user-experience evaluation were used as evaluation indices of vehicle software. Twenty target users were recruited for the study. A head-mounted eye tracker, a finger-trajectory tracking system, and a satisfaction questionnaire were used to collect the eye-movement data, finger-movement data and subjective ratings while participants evaluated three kinds of in-vehicle software. Eye-movement and hand-operation data were processed with D-lab and EthoVision XT software, and a multi-dimensional comprehensive evaluation model combining psychological and physiological indices was established. The evaluation model was then validated. The results show that the navigation software provides the best user experience, the air-conditioning software ranks second, and the media-player software performs the worst.

When an autonomous vehicle (AV) is in motion, the driving risks caused by the external environment can lead to occupants' distrust, reducing their acceptance of AVs. Therefore, quantifying occupants' perceived risk is crucial for designing and evaluating AV behavior, as it provides theoretical support for mitigating that risk. The paper quantifies the relationship between objective scenario-level risk factors and subjective perception in overtaking scenarios using a logistic regression model. Firstly, based on 92 overtaking segments of data collected in real-world driving experiments, 7 candidate risk factors are identified. Then, a logistic regression model is established in which the 5 risk factors that passed the hypothesis test are used as the independent variables and the binary classification of occupants' perceived risk serves as the dependent variable. The model analysis indicates that three factors, i.e. risk in adjacent areas (S), time to collision (t_{\mathrm{c}\mathrm{o}\mathrm{l}}) and time headway (t_{\mathrm{h}\mathrm{e}\mathrm{a}\mathrm{d}}), are significantly related to occupants' perceived risk, with t_{\mathrm{h}\mathrm{e}\mathrm{a}\mathrm{d}} being the most influential factor. To classify whether occupants perceive risk, the cut-off value of the prediction model is set at 0.462, which is calculated from the Receiver Operating Characteristic (ROC) curve. By using the HighD dataset, the cut-off value is verified and the accuracy of the prediction model is found to be 89.1%. On this basis, three optimized driving strategies are formulated to mitigate high perceived risk in overtaking scenarios. These three strategies are compared in driving-simulator tests in terms of traffic efficiency and perceived risk, confirming the validity of the model's analysis conclusions.

Virtual simulation testing has become the industry-wide norm for intelligent connected vehicles. Such testing demands a rich set of simulation scenarios. Based on data collected from naturalistic highway-driving scenarios, the paper develops an automated strategy for extracting lane-change cut-in events. A perception risk coefficient is introduced, and one-way ANOVA is used to analyze discrete factors while Pearson correlation analysis is employed for continuous factors, revealing the relationships between scenario elements and risk and identifying the key influencing factors. Typical scenarios are then derived with K-Means clustering and the elbow method, key parameters are retained, and each scenario is constructed on the Prescan simulation platform. The results show that this method can effectively extract critical lane-change cut-in scenarios from real-world data, cluster them into typical scenarios, and reconstruct these scenarios on the simulation platform, providing scientific support for autonomous-driving system testing.