To rapidly and accurately predict the crashworthiness indicators of thin-walled tubes, a bionic gradient circular multi-cell thin-walled tube axial compression model for predicting the average collision force is established based on the Simplified Super Folding Element (SSFE) theory, and a Long Short-Term Memory (LSTM) network model is built to predict the crashworthiness indicators of the bionic gradient circular multi-cell thin-walled tube under different geometric parameters. The results show that the theoretical prediction error is less than 6% compared with the simulation results, indicating the reliability of the theoretical model. The LSTM network model exhibits an error of less than 2% for Energy Absorption (EA) and Initial Peak Force (IPF) on the validation set, and an error of less than 5% on the test set, demonstrating excellent prediction accuracy and generalization capability.

In the safety property development of side pole under crash condition, the relationship between vehicle structural characteristics (such as initial side space, speed waveform) and occupant injury is not uncertain, and there are few strategies for matching side airbags to vehicle structural characteristics. In order to address the above issues, this paper proposes a lateral Pole collision Occupant Load Criterion (POLC) to correlate the structural characteristics of the vehicle’s side structure and occupant injuries, and categorize vehicles into two groups based on the relative magnitude of their POLC values to the occupant’s average acceleration tolerance limit a_{ave_max}. For a given vehicle A with a POLC value less than a_{ave_max}, the paper analyzes deformation of the side airbag matched for the vehicle from the perspective of occupant kinematics, then analyzes whether the side airbag stress-deformation curve meet occupant protection demand from the perspective of stress limit of occupant torso, in an attempt to provide a theoretical basis for the design of vehicle side structure characteristics at the early stage of vehicle crash safety development.

In order to analyze pedestrian leg crash injury in active and passive safety fusion scenes with intervention of Automatic Emergency Braking (AEB) system, this paper firstly compares and analyzes THUMS (Total Human Model for Safety) human body finite element model and advanced Pedestrian Legform Impactor (aPLI), then selects THUMS dummy model with higher biological fidelity to study the influence rule of collision speed and vehicle body posture on leg injury index with the intervention of AEB system. The results show that the motion response of THUMS dummy is closer to practical collision state compared with aPLI. With the change of collision speed and vehicle body posture, different types of vehicles show different leg-type injury rules at different collision positions, and the influence of collision speed on leg injury index is greater than that of vehicle pitch attitude. The increase of vehicle pitch angle and the decrease of collision speed will cause the decrease of upper leg collision speed, but it has no obvious effect on the collision angle. The change of collision speed and vehicle body posture will affect the collision posture of pedestrian leg and vehicle, thus influencing various leg injury indicators. Therefore, the impact of vehicle body braking on pedestrian collision should be considered in the pedestrian lower limb injury evaluation system, so as to evaluate the pedestrian lower limb injury more comprehensively and objectively.

In order to reduce the risk of passenger lower leg injuries in crash test, this paper analyzes the influencing factors of dummy lower leg injuries based on a large number of test data, and then studies the influence of passenger foot posture on lower leg injuries through leg impact test and simulation. The results show that the direct cause of foot injury is the excess lower tibial moment M_y value. Lower leg injuries could be effectively reduced by adjusting the foot posture. When the foot is tilted forward by 10°, the lower tibial moment M_y is 94.21 N·m, and the tibial index T_i is 0.44, which could be reduced by 37% compared with the standard foot posture.

To reduce injuries of through-type headlamp to pedestrians’ pelvic limb in collisions, a series of pedestrian upper leg impact tests are designed to evaluate the crash response characteristics of a car equipped with through-type headlights at different impact points. The test results show that when the upper leg collides with the middle area of the headlamp, the impact force peak value reaches up to 6 424 N, the injury risk is very high. The paper proposes two optimizations to headlamp including low-temperature treatment and increasing initial injury, and a series of test are performed. The results show that the risk of injury to pedestrians’ legs is reduced by reducing the structural toughness of the through-lamp lampshade by low-temperature treatment which cause the lampshade crack at early stage, which effectively reducing impact force peak value and improving the energy absorption capacity in collisions. In contrast, the effect of only weakening toughness of interior plastic parts is not obvious.

The resrearch aims of the paper is to improve the optimization efficiency of the automobile sill beam, and address the challenges of optimization such as the limited energy absorption of the sill extrusion aluminum under the condition of the side column collision, long iterative cycle of explicit solution in the simulation and the high requirements of the manufacturing process. A collision model is established based on an SUV model, and the Design Of Experiment (DOE) analysis is made with sill beam thickness as variables, resulting in 144 groups of valid design data are obtained. A Reduced-Order Model (ROM) is formed by deep learning methods(Rapidminer, romAI) and used as the alternative model of optimization and simulation. The results of CAE simulation verifies that accuracy of romAI reaches more than 95% in optimization and the solution speed is increased by more than 40 times under the limited data, which greatly shortens the R&D cycle.

In order to ensure the repeatability of simulation tests and reliability of test results, a simulation testing platform is built using Unity and Unreal Engine based on a systematic analysis of the definition and potential causes of performance consistency issues in autonomous driving simulation software, and static and dynamic scenarios are built at the same time. The study also involves the design of test cases, data processing scripts, automated testing scripts and the overall experimental process. Test results based on Unity and Unreal Engine simulation test platform are quantified through the inconsistency deviation. Influenced by various factors related to the simulation system’s software and hardware, experimental results reveal significant inconsistencies in multiple tests, with the maximum deviation reaching 15%. To address the identified deviation, this paper proposes several optimization strategies, including process optimization and power mode optimization, which are validated to reduce the deviation by more than 40%, thereby enhancing the reliability and consistency of simulation tests.

In order to facilitate a comprehensive understanding of the Autonomous Valet Parking (AVP) system including its behavioral limitations, and to ensure its Safety Of The Intended Function (SOTIF), the paper proposes the SOTIF analysis process for AVP systems. This procedure synthesizes the ISO 21448 and the System Theory Process Analysis (STPA) method, and elaborates on system safety and triggering conditions, layers triggering conditions and combines with fuzzy comprehensive evaluation to build SOTIF scenarios. The evaluation employs collision distance risk, collision time risk and braking deceleration risk as key performance indicators. The Criteria Importance Through Intercriteria Correlation (CRITIC) and Technique for Order Preference by Similarity to Ideal Solution (TOPSIS) methods are utilized for index weighting and evaluation. Finally, SOTIF scores of AVP systems under different scenarios are obtained through real vehicle testing, clarifying unsafe scenarios and providing targeted guidance for product optimization.

To address the issue of limited path tracking ability of Ultra-thin Carrier Robot (UCR) in complex test scenarios in intelligent connected vehicle testing, this paper proposes an improved Stanley lateral motion control algorithm. Firstly, the kinematic model of the UCR is established based on the Ackermann steering principle, and the bicycle model is used to simplify the steering system of the UCR. Then, a fuzzy PID controller is introduced on the basis of the traditional Stanley algorithm in order to enhance the adaptive feedback capability of the algorithm. The results of the joint simulation of CarSim and MATLAB/Simulink show that the modified Stanley algorithm reduces the average lateral error by 50.67% and the maximum lateral error by 41.76% at different speeds compared with the traditional algorithm. The real-vehicle test further confirms that the improved algorithm performs well in medium and high-speed straight line and lane changing scenarios, with the average lateral error less than 0.05 m and the maximum lateral error less than 0.17 m. It meets the testing requirements of intelligent connected vehicles in different scenarios and realizes high-speed and high-precision lateral motion control of the UCR.

In order to enhance the testing and evaluation of systematicness and comprehensiveness of Moving Off Information System (MOIS) for commercial vehicles, the paper analyzes the evaluation requirements of commercial vehicle MOIS, clarifies the working principle of MOIS, and proposes specific testing and evaluation plans and improvement measures. Commercial vehicles with the same MOIS perception and recognition scheme but different sizes are used as test samples, the alarm triggering and ending timing, alarm duration, etc. of MOIS are evaluated and analyzed for different target objects, blind spot positions, and bias rates. The results show that the same scheme of MOIS installed on commercial vehicles of different sizes can meet regulatory requirements, and the monitoring effect of the system on the forward blind spot shows a trend of right side>left side>middle. Moreover, when the vehicle moves together with the longitudinal target object in front, the larger vehicle has better blind spot monitoring effect.