In order to give full pay to the advantages of rapid torque response and independent controllable torque of the distributed driven electric vehicle, this paper adopted the chassis stability coordination control method, through the coordinated control of the electronic stability control system and the torque vector control unit to deal with the stability requirements of the vehicle under various working conditions-improve the dynamic performance of the vehicle under stable conditions through TVCU independent control, ensures the power and stability of the vehicle by ESC when it tends to be unstable. The robustness of the stability coordination control method was verified based on a real vehicle.

In order to recognize driving fatigue, this paper proposed a fatigue detection method based on wavelet characteristics and Long Short-Term Memory (LSTM) neural network classifier. Two kinds of EEG signals (non-fatigue and driving fatigue) were collected in the real driving environment. The EEG signals were decomposed by wavelet, and the statistical values, energy values and relative energy values of four wavelet coefficients were calculated as the characteristic data, which were used for classification training and test of the LSTM neural network. The results of experiment show that the classification performance of LSTM neural network gradually improves with the increase of the number of channels involved in constructing characteristic data. Especially, in the scheme of 14 channels, the average classification accuracy is about 96.1%.

A test method and test bench for knee module impacting dashboard were proposed and designed, to reproduce the contact of the dummy leg with dashboard in real vehicle crash. The collision process between the knee and the dashboard in the car was studied pertinently, and the knee invasion curve on the dashboard was obtained by the high-speed camera dynamic displacement calculation method, which was interpolated and integrated to obtain the relationship between the injury index of the dummy leg and the amount of knee intrusion on the dashboard, to directly reflect and verify the collision protection performance of the dashboard to the leg.

This paper provides a comprehensive review of the application of computer vision technology in the automatic detection of road potholes and objective assessment of road conditions. Typically, cameras and various types of depth sensors are used to acquire two-dimensional and three-dimensional road data for road imaging. Pothole detection is carried out based on computer vision technology, with the main detection algorithms including classic two-dimensional image processing, three-dimensional point cloud modeling and segmentation, deep learning, and their hybrid methods. The hybrid methods, which exploit the advantages of various algorithms, can greatly improve the accuracy of detection. However, while existing algorithms have achieved good results in pothole detection, they still face many challenges, such as the need to improve the robustness of road geometry reconstruction, high algorithm complexity, and the model’s strong dependence on large-scale well-annotated datasets. Therefore, future research should focus more on unsupervised stereo matching algorithms and deep learning algorithms with few samples.

In order to effectively control vehicle NO_x emission, this paper studied vehicle NO_x emission from the perspective of intake temperature and road conditions with the single factor variable method and obtained the influence rule of different factors on vehicle NO_x emission. The results show that with the intake temperature increasing by 10 ℃, the original NO_x emission concentration increases by about 5% and the tail gas emission concentration increases by 19%~44%. The influence range of different road conditions on NO_x emission shows great difference.

In order to ensure the accuracy of trajectory tracking and maneuvering stability of intelligent vehicles at high speeds on roads with different coefficients of adhesion, the Adaptive Model Predictive Control (AMPC) method was used to estimate the tire cornering stiffness on-line and update the parameters of the vehicle dynamics model in real time under high-speed driving conditions, and the Non-Fixed Step (NFS) discrete method was used to prolong the prediction time domain while the slip stability constraints were added into the objective function, thus maintaining high stability and control in real time under high-speed driving conditions. The proposed method can improve the maneuvering stability and control accuracy when the vehicle is driving on the road with different adhesion coefficients under high-speed driving conditions, as shown in the joint simulation results of CarSim and MATLAB/Simulink.

For the problem of intelligent vehicles recognizing lane line deviation in different environments, this paper proposed a lane line detection and lane deviation warning algorithm based on sliding window search. Firstly, real-time road images were preprocessed, including camera calibration, color and gradient filtering, perspective transformation, etc., and then the sliding window search algorithm was used to detect the lane lines and fit the lane lines using quadratic polynomials. Then, the camera monocular ranging principle was used to calculate the relative position of the vehicle to the centerline of the lane and determine whether it deviated from the lane. The experimental results show that the algorithm’s lane line detection accuracy reaches 98.59%, the lane deviation warning rate reaches 99.58%, and the processed video frame rate is about 25 frames/s, which meets the accuracy and real-time requirements.

In the context of high-speed mixed traffic and intricate multi-vehicle interaction, existing driving intention recognition models for research vehicles mostly neglect driving styles and vehicle-vehicle information interaction, this paper introduced a novel driving intention recognition model based on an enhanced Bidirectional Long Short-Term Memory (Bi LSTM) network, with the driving trajectory sequence of the target vehicle, driving style, and interaction features of surrounding vehicles as inputs for effective training and learning, to facilitate the classification and recognition of the driving intention feature dataset, specifically considering diverse driving styles. Additionally, the Whale Optimization Algorithm (WOA) was employed to optimize hyperparameters, encompassing the number of hidden layer nodes and learning rate, effectively mitigating the adverse impacts of manual parameter adjustment. The model’s efficacy was validated using the NGSIM dataset, exhibiting an impressive recognition accuracy of 97.5% in precisely identifying vehicle driving intentions.

Hub motor air gap eccentricity leads to electromagnetic force imbalance and torque fluctuation increase affecting the vehicle driving stability. To solve this problem, firstly, the Fourier function was used to derive the relationship curves of inductance, magnetic chain, radial electromagnetic force and torque of the switched reluctance motor concerning the current and angle, and then, the spatial electromagnetic characteristics of the electromagnetic force and torque were analyzed based on Maxwell’s stress tensor method in the two operating conditions. The results show that the air gap eccentricity leads to the imbalance of electromagnetic force and torque at the eccentric position. Then, the contribution weights of structural parameters were determined by sensitivity analysis, and a multi-objective optimization scheme was formulated, finally, the multi-objective optimization search of structural parameters was carried out with the help of NSGA-Ⅱ, and two groups of optimization schemes were screened, of which the optimal scheme C1 improves the optimization results greatly, and the longitudinal and transverse stability of the wheel-hub motor-driven automobile is improved efficiently.

In order to realize ecological driving behavior decision-making of autonomous vehicle at intersections, the driving behavior of the vehicle without the influence of the vehicle in front was determined through empirical fuel-saving strategies. Using NGSIM dataset, combined with Gaussian Mixture Model (GMM) cluster analysis method and Support Vector Machine (SVM) theory, the driving style of the driver in front of the vehicle was identified, and a binary logistic regression-based intersection driving behavior decision model for self-driving cars was established. According to the different behaviors of decision making and considering vehicle fuel consumption, the multi-stage ecological driving optimal control issue was transformed into a full-stage optimal control issue. The paper proposed an eco-driving strategy for self-driving cars at intersections, and the effectiveness of the strategy was verified by simulation.