In order to improve success rate and accuracy of automatic parking, firstly, the input image features were extracted based on Convolutional Neural Network (CNN) model, and then the encoding-decoding mechanism of Transfomer model was used to tile the image features extracted by CNN for calculation and inference. Finally, the target prediction results were obtained by feedforward neural network. In this paper, fisheye images were used to recognize the target. The center point of the parking angle and the center point of the empty parking entrance were expressed by two-dimensional coordinate points, which reduced the redundancy of the output information and optimized the model structure. The test results show that the algorithm can better adapt to different parking space line marking mode and different natural environment, with the recall rate of target perception reaches 98%, and the average error of parking space corner center location is less than 3 cm, which meets the requirements of real-time application for robustness, real-time and accuracy.

The 3D point cloud object detection algorithm based on deep learning is prone to issues such as inability to maintain network performance and poor transferability when changing scenes or devices. To address this issue, this article proposes an Accurate, Flexible, and highly transferable two-stage 3D point cloud object detection algorithm (AF3D). In the first stage of the AF3D detection algorithm, a segmented fitting algorithm is used to remove the road surface from the collected laser point cloud, then DBSCAN algorithm is used to cluster non-ground point clouds and obtain several clustering clusters. In the second stage of the AF3D detection algorithm, a point cloud fully connected network PFC-Net is established, and features are extracted and classified. Through experiments, it has been proven that this algorithm can achieve good detection performance on public KITTI datasets, and the detection accuracy for cars, pedestrians, and cyclists on real vehicle datasets is 69.74%, 41.25%, and 54.33%, respectively, indicating good transferability.

In response to the privacy leakage issues in the data sharing process among the fleet vehicles in Vehicular Social Network (VSN), this paper proposed a ciphertext policy attribute-based encryption scheme. Firstly, utilizing attribute-based encryption techniques, the scheme ensures that only authorized vehicles can access the data, thereby preventing privacy leakage during the fleet vehicle data sharing process. To address the issue of high time complexity in generating access policies during fleet vehicle data sharing, and low data sharing efficiency, an access tree was constructed to design the access policies. The access tree was then transformed into an access matrix using roadside units, enabling fast generation of access policies. Experimental analysis shows that this scheme achieves secure data sharing among fleet vehicles in VSNs, the proposed access policy generation approach effectively reduces the computational overhead for vehicles.

To achieve accurate prediction of EV battery energy information, this paper proposed a method for battery energy analysis and prediction based on big data of chargeable pure electric vehicles. Firstly, the big data of vehicles with the same battery model from different regions were obtained through a big data platform, and then the interval average method and Support Vector Regression (SVR) were used to fit the relationship between mileage and total energy for both the total data and typical regional data, to predict degradation of the battery total energy. Finally, the predicted results were compared with that obtained from Long Short-Term Memory (LSTM) neural network, and the accuracy of the proposed method was verified by vehicle test. The results show that: the SVR-based model can quantitatively fit the degraded battery capacity, which has high prediction accuracy.

In order to improve the energy management of Range Extended Electric Vehicle (REEV), firstly Long Short-Term Memory (LSTM) neural network was used to predicate vehicle speed, then calculates the demand power in the prediction time domain, and the demand power in the prediction time domain and the demand power at the current moment were jointly inputted to the Deep Deterministic Policy Gradient (DDPG) agent, which outputted the control quantity. Finally, the hardware-in-the-loop simulation was carried out to verify the real-time performance of the control strategy. The validation results show that using the proposed LSTM-DDPG energy management strategy reduces the equivalent fuel consumption by 0.613 kg, 0.350 kg, and 0.607 kg compared to the DDPG energy management strategy, the Deep Q-Network (DQN) energy management strategy, and the power-following control strategy, respectively, under the World Transient Vehicle Cycling (WTVC) conditions, which is only 0.128 kg different from that of the dynamic planning control strategy when the dynamic planning control strategy is used.

In order to improve the attitude angle solving accuracy of Micro-Electro-Mechanical System Inertial Measurement Unit (MEMS IMU) in unmanned vehicle system, this paper proposed a Particle Swarm Optimization (PSO) based algorithm and a Strong Tracking Adaptive Unscented Kalman Filter (STAUKF) data fusion method. Firstly, two kinds of IMU modules with different precision were filtered by STAUKF algorithm. Secondly, two kinds of error functions were constructed and PSO algorithm was introduced to fuse the two kinds of IMU posterior estimation. Finally, the test was carried out on the built unmanned vehicle platform. Experimental results show that, compared with the data solved by two single IMU sensors, the root mean square error of the transverse roller shaft and pitch shaft angle solved by the proposed algorithm is reduced by 56.67% and 58.94%, respectively, and the data solved is reduced by 36.55% and 52.15% respectively compared with direct weighted average of the redundant dual IMU system. Therefore, the algorithm proposed in this paper is more accurate and robust.

At present, low-end electric vehicle do not have direct slope signal information. To address this issue, this article proposed a road slope estimation algorithm based on the existing CAN bus signals. Firstly, vehicle speed signal and longitudinal acceleration signal values input from CAN were preprocessed, and a Kalman filtering equation was constructed based on the relationship between slope, vehicle speed and longitudinal acceleration, to estimate the road slope. Then, the impact of noise variance parameters on the slop estimation results was analyzed, and an adaptive tuning Kalman filter was designed to optimize the slope estimation results. Finally, the vehicle test was conducted, verifying the accuracy and real-time performance of the ramp estimation algorithm.

A method for optimizing the layout of carbon fiber composite floorings for BIW was proposed to enhance precision, efficiency, and structural lightweight. Initially, BIW finite element model was established and its efficiency was validated. Subsequently, material parameters for the carbon fiber composite were obtained through mechanical performance testing, followed by conceptual designing and modeling of the flooring layout. Subsequent utilization of continuous variable optimization determined the thickness, block shapes, and layers of the flooring, employing a discretization and rounding strategy to achieve discrete layer numbers for each layup angle. The optimization results show that the Particle Swarm Optimization-Bacteria Foraging Optimization (PSO-BFO) algorithm proposed herein improves flooring quality, static bending stiffness and BIW lightweight coefficient by 34.4%, 6.0% and 5.3%, respectively.