In order to improve the safety of a commercial vehicle cab complying with the requirements of GB 26512—2011 real vehicle test and make it meet the requirements of GB 26512—2021, the Box-Behnken test design and Sequential Quadratic Planning (SQP) algorithm improvement were carried out for problems occurring in the cab simulation such as excessive peak force of longitudinal beam in frontal impact test, insufficient energy absorption in A-pillar impact test and serious crushing in side 20° impact test; and cab was optimized by setting bending beam and filling Expanded Polypropylene (EPP) foam material. The results show that after optimization, the longitudinal beam peak torque is reduced from 92.3 kN to 72.4 kN, the structural energy absorption is improved by 36.9%, the survival space is sufficient to meet the requirements of the new standards, and the cab safety is significantly improved.

In order to improve the occupant protection performance of a vehicle in the frontal 25% offset crash test, the research on the test methods and optimization schemes of the evaluation items of the injury level of the dummy, the restraint system and the motion level of the dummy was carried out. First, an accurate model was established by simulation and benchmarking test. Then, side air curtain was added, the seat belt and ignition time and parameters of airbag were optimized, and the structure of steering column bracket and instrument panel bracket were also optimized. Finally, it was verified by CAE. The results show that the grades of the thighs and hips, legs and feet of the dummy have been improved from Marginal (M) and Poor (P) to Good (G) and Acceptable (A), and the evaluation items of the restraint system and the motion grade of the dummy have been improved from Poor (P) to Good (G). The overall evaluation of small offset crash was improved from Poor (P) to Good (G). The occupant Weighted Injury Criteria (WIC) decreased by 22.97%.

In order to explore the kinematic responses and injury mechanisms of far-end elderly occupants under the current Euro-NCAP side impact condition, a sled model was established using the elderly human body model CHARM-70, then the Advanced European Mobile Deformable Barrier (AE-MDB) side impact and oblique pole side impact finite element model were established, to conduct the vehicle collision simulation, and extract the acceleration pulse at the bottom of the B-pillar on the non-collision side. Finally, the single-occupant and double-occupant sled simulation models were established based on the THUMS_AM50_V4.0 human body model and the elderly human body model CHARM-70, to conduct the AE-MDB side impact and oblique pole side impact simulation. The results show that in the side impact, the seat belt can’t effectively limit the lateral offset caused by the inertia of the far-end occupant; in the single-occupant sled simulation, the head and neck injury values of the far-end elderly occupants are far less than the threshold, but the chest suffers serious rib fracture; in the double-occupant sled simulation, the far-end elderly occupants suffer severe head and chest injuries.

To study the influence of different service cycles of BioRID II dummy on test evaluation results in the whiplash tests, 4 whiplash tests are carried out by taking 2 dummies with different service cycle seating the H-point x at the baseline position +5 mm and -5 mm respectively, meanwhile, the dummy’s neck structure and calibration descriptions are discussed. The results show that the difference in service cycle between BioRID II dummy has significant influence on the Neck Injuries Criteria (NIC) and the lower neck axial force, and the gap in the whiplash test evaluation may reach more than 0.5. The rationality of the current way of adjusting the neck My and the method and dimension of dummy calibration remain to be discussed.

In order to improve the safety of rear occupants with different sizes, this paper analyzed the damage of rear occupants in 100% frontal collision condition at 50 km/h by simulation. Firstly, a rear occupant restraint system model was established and verified. Then THUMS_AM50, THUNS_AF05 and CHARM-70 models were used to simulate different rear passenger sizes, Constant Loading Limit (Con-LL) curve, Progressive Loading Limit (Pro-LL) curve and Descending Loading Limit (Dig-LL) curve were respectively applied to explore the adaptability of seat belt loading curve to passengers of different sizes. The results show that the risk of neck ligament injury is great under this condition. The Dig-LL curve showed significant protection for the CHARM-70 but poor protection for the THUMS_AM50 during a frontal collision. Con-LL curve is not suitable for the protection of THUMS_AM50 and CHARM-70 under this condition. Pro-LL curves showed good protection for THUMS_AM50 and THUNS_AF05. In the future, the seat belt restraint system should be adapted to different loading curves according to the occupant type and collision condition to better protect the occupant.

China’s characteristic side barrier was developed with the SUVs as the research object. Through extraction of characteristic parameters of typical SUV models, the development objectives of honeycomb aluminum barrier and trolley were defined, and the Tresca theoretical model was used to guide the selection of barrier materials and the determination of trolley component specifications. Based on the performance test of barrier trolley dynamic load-cell wall, the performance of honeycomb aluminum barrier was improved and verified. The results show that the improved barrier is stable in deformation mode, the mechanical properties meet the design specifications, and can characterize the appearance and stiffness of SUV.

In order to study the feasibility of active-passive integration test based on acceleration sled, simulation and sled test were carried out on the 50th THOR dummy under 8.0 m/s² braking, and acceleration and deceleration simulation were carried out under frontal high-speed collision condition after AEB system braking. The results show that the occupant out-of-position fluctuates under AEB system braking, the maximum displacement occurs before the 300th ms, and the fluctuation range is small after the 400th ms. Under the simulation of acceleration sled, after the 400th ms, the protective effect of airbag on head and the stiffness performance of dummy chest are closer to the deceleration test. It therefore can be concluded that the active-passive integration test can be carried out on the acceleration sled by reducing the deceleration time of AEB under the circumstance of the reduced simulation time of AEB should not be less than 300 ms.

In order to solve the problem that existing LiDAR attenuation models rely on statistics to generate point cloud and lack of noise interpretation, this paper presents a LiDAR attenuation model for raining environment. Firstly, the LiDAR emission-reception model is established, and the spatial distribution of raindrops is simulated according to the raindrop size distribution model. Secondly, the light intensity change of the laser propagation process is obtained by coupling scattering model and noise model, and simulation imaging of point clouds is obtained. Finally, the image of point clouds in normal weather and raining weather is collected, to simulate and generate the attenuated point clouds for different rainfalls. The point clouds obtained from the attenuation model is compared with the point clouds image corresponding to actual raining weather. The results show that the presented model is superior to the existing models in each evaluation index, and effectively explains the attenuation effect of the raining environment on LiDAR.

For the problem that current target detection methods generally require a high-power consumption GPU computing platform and are easily affected by lighting conditions, this paper proposed 2 infrared pedestrian detection methods in front of vehicles based on embedded platform: the trained YOLOv4-tiny model was optimized using NVIDIA’s open source inference acceleration library TensorRT and deployed on the embedded platform; the YOLOv4-tiny model was used as the basic architecture of the algorithm, which was combined with the visual attention mechanism and the spatial pyramid pooling idea, and a YOLO layer was added at the same time, a YOLOv4-tiny+3L+SPP+CBAM network model was proposed. The 2 methods were trained and tested on the FLIR dataset, and tested on the Jetson TX2 embedded platform. The test results show that: compared with the original network YOLOv4-tiny, the average accuracy of the first method is reduced by 0.54%, and the inference speed is increased by 86.43% (frame rate up to 26.1 frame/s), the average accuracy of the second method is improved by 16.21%, and the inference speed is reduced by 22.86% (frame rate up to 10.8 frame/s). Both methods can take into account the accuracy and real-time performance, and meet the needs of infrared pedestrian detection in front of vehicle.

To solve the problem of low accuracy when classifying traffic signs using Convolutional Neural Networks(CNN), this article proposes a cascade super-resolution network structure by connecting an image super-resolution network to a classification network. A modified dual-attention mechanism super-resolution network is first used as a sub-network of the cascade network, and then the image classification network is trained for classifying the super-resolution processed images, and finally the classification accuracy is used to measure the effectiveness of super-resolution reconstruction for the image classification task. The validation results of both simulation and real traffic sign datasets show that the super-resolution processed images achieve higher classification accuracy in the classification model, which proves that the super-resolution technology has a facilitating effect on the improvement of the classification accuracy of traffic sign images.