In order to improve the ability of the collision warning system to perceive the surrounding environment, this paper proposed a collision warning system based on YOLOv5 and hazardous area judgment. Firstly, the discriminative ability and accuracy of the model were improved by the channel attention module, then, the extraction ability of the model for multi-size features was improved by using path aggregation network and spatial pyramid pooling, and finally, the warning accuracy of the warning system was improved by filtering relatively safe targets through the introduction of warning activation regions. The results show that the introduction of warning activation regions improves the accuracy, precision and recall of the warning system by 20%, 50% and 26.7%, respectively, the running speed is increased by 49.1%, which further proves the effectiveness of the method.

To assess the safety risks of autonomous vehicles during cut-in scenarios on congested Chinese highways, 64 cut-in samples were extracted from a natural driving dataset. Employing a six-level model and correlation analysis, the static and dynamic factors of the scenarios were defined. Subsequently, 1 000 test cases were randomly generated through sampling, and a safety assessment index system was established to analyze the safety of vehicle operations. Lastly, the random forest algorithm was applied to identify the key factors triggering risks. Results indicate that risk scenarios account for 5.3% of the total, with longitudinal relative velocity identified as the crucial factor. Under congested conditions, a high-risk cut-in scenario is formed when the speed of surrounding vehicles is 23% lower than that of autonomous vehicles, this indicator serves as a crucial predictive measure for identifying collision risks in congested cut-in scenarios for autonomous vehicles and may be applied in determining liability of accident in such scenarios.

Based on THUMS digital human model, this paper studied the occupant injury characteristics of rear attitude frontal collision in a large tilt seat and discussed the mechanism of occupant injury. As the result shows, compared with regular sitting posture, on the one hand, the traditional restraint system cannot realize the effective restraint on the occupant under the reclining posture, which easily causes injury of occupant’s abdomen, thorax and neck due to descending of occupant. On the other hand, the occupant’s spine is subjected to a large axial acceleration load, and there is a risk of injury due to the seat cushion. Finally, based on the characteristics and mechanism of occupant injury, the paper also discussed the strategies on occupant protection in a large seat.

In order to improve the safety score of the driver’s side airbag under the side pole collision condition, the model of occupant’s restraint system in the vehicle side pole collision was built according to the 2021 version of C-NCAP, and the vehicle collision and simulation results show that the safety scores of the head, abdomen and pelvis all reach the standard, and the chest injury is more serious and the score is lower. By optimizing the pouch shape and stomatal diameter of the side airbags, the chest rib safety score is improved. The optimization results show that the compression of the upper rib, middle rib and lower rib of the chest decreases by 30.97%, 8.26% and 13.71%, respectively, which verifies the effectiveness of the optimization scheme.

In order to investigate the effects of diverse occupant out-of-position postures on occupant crash injuries, a variety of typical occupant out-of-position postures were considered, and 1 800 frontal crash simulations were conducted with the help of a mature occupant compartment model and dummy model. The results show that the occupant will suffer more injury in typical out-of-position posture than the standard posture; the increase of seatback angle leads to the increase of occupant head and neck injury level, while the decrease of thorax injury level; the increase of braking intensity of Automatic Emergency Braking (AEB) system increases the occupant thorax injury level, while decreasing the head and neck injury level; the occupants’ head and thorax injury level is greater in the rightward offset condition than in the leftward offset condition.

In order to explore the working principle of legform impactor calibration test, this paper compared in detail the structural differences between advanced Pedestrian Legform Impactor (aPLI) and Flexible Pedestrian Legform Impactor (FlexPLI), summarized the methods for determining of dynamic calibration limits and static calibration limits of legform impactor as the mixed group mean estimation method and the three-point bending deflection test method respectively. The paper also summarized and compared the significant differences between dynamic calibration limits and static calibration limits of the two types of legform impactor, and analyzed the immanent cause for these differences. The main cause for the difference in dynamic calibration limits was the addition of mass module to simulate human upper limbs in aPLI, while the main cause for the difference in static calibration limit was the structural changes in the leg structure and knee structure of aPLI.

To improve the cooling effect, this paper proposed a highly symmetrical bionic network channel cold plate. It firstly analyzed the influence of the cold plate’s structure parameters on its performance through single-factor analysis, then, optimized the structure parameters of the cold plate using the Multi-ObjectiveParticle Swarm Optimization (MOPSO) algorithm, with the average temperature, temperature standard deviation, and coolant pressure loss of the cold plate serving as performance indexes. The optimal channel width, channel depth, and cold plate wall thickness were found to be 9.0 mm, 1.5 mm, and 1.4 mm respectively. The corresponding average temperature, temperature standard deviation, and pressure loss were measured as 33.20 ℃, 1.33 ℃, and 65.63 Pa respectively. When compared with the initial structural parameters, the optimized mean temperature and temperature standard deviation decreased by 1.92 ℃ and 0.02 ℃ respectively, while the pressure loss increased by 27.10 Pa. Finally, the optimization results were verified using the battery module.

To address the issue of understeer, difficulty in control effect calibration, complex controller design and difficulty in engineering application, this article proposed a novel approach to vehicle handling control by enabling adaptable rear wheel steering and braking intensity. Employing nonlinear model design and a feedforward/feedback structure, this method can adjust control intensity according to design parameters, allows for improving the vehicle’s center of mass sideslip angle according to expectation and enhancing yaw response speed. In addition, the approach also accommodates understeer mitigation and fuel economy optimization. In addition, the proposed method also has other advantages such as high accuracy, small computational complexity, no need for vehicle sideslip angle estimation, and ease of calibration. A method for vehicle handling tuning based on adjustable parameters is also provided.