The road noise problem in the frequency range of 35-40 Hz generated by a certain vehicle model on rough asphalt pavement at a speed of 30 km/h is analyzed by simulation and testing methods. Based on structural noise transfer path analysis （TPA） and node contribution analysis， it is determined that the problem frequency noise is mainly caused by low-frequency vibration of the vehicle body sheet metal parts. Therefore， based on the partial resonance principle， a resonance structure that can be industrialized according to the shape and local space of the sheet metal is designed， which is composed of a metal bracket and a metal block. By attaching this resonance structure to the sheet metal area with a significant contribution， simulation results show that the structure reduces the vibration of the sheet metal， thereby optimizing the interior noise. The peak noise levels in the front and rear seats are reduced by 8.6 and 6.4 dB， respectively. Meanwhile， the vehicle test results indicate that the interior noise has been improved by 7.0 and 5.8 dB. This provides a new research method and optimization scheme for controlling low-frequency noise caused by vibration of sheet metal.

In shared road space， there are path conflicts between different road users moving in various directions. Road users must negotiate right-of-way through driving interactions to avoid collision risks， thus resolving potential conflicts. The description and modeling of interactive behaviors is crucial for accurately understanding and predicting the dynamic environment. Therefore， a semantic-level representation and extraction method for multi-vehicle interactive behaviors is proposed in this paper， taking interactive trajectory primitives as analysis units. Firstly， a nonparametric Bayesian method is utilized to segment interactive behaviors， obtaining interaction segments with significant behavior patterns. Then， the sticky hierarchical Dirichlet-Hidden Markov Model is employed to extract interaction primitives from these interaction segments. Finally， unsupervised clustering is applied to the normalized interaction primitives to obtain semantic-level behavioral features of interaction scenarios. An empirical study based on 20 797 pairs of multi-vehicle interaction data from the NGSIM highway dataset shows that the method proposed in this paper can extract and analyze complex interactive scenarios involving multiple participants， breaking through the limitation of existing research that only constructs interaction primitives for two vehicle interaction scenarios， and supporting the analysis of interaction among multiple traffic participants. The experimental results show that the proposed method can segment continuous driving behaviors into discrete interaction primitives. The clustering results correspond to actual interaction scenarios and can be used to characterize the interaction behaviors among vehicles in different interactive trajectory primitives. Furthermore， the method can enhance performance of downstream driving tasks in complex scenarios. In multi-step vehicle trajectory prediction， by integrating with baseline prediction methods， the proposed method can reduce the average prediction error and final position error by 19.3% and 14.6%， respectively， compared to baseline methods.

For the problem that existing numerical simulation methods are unable to accurately reflect the probabilistic temperature variations caused by thermal runaway in lithium batteries， s a modeling method for the thermal runaway propagation of lithium battery modules based on probability function triggering is proposed. This method calculates the triggering probability of thermal runaway in each temperature range by statistical analysis of the temperature range and distribution of actual thermal runaway events in lithium batteries. Based on the proposed probabilistic trigger simulator， the simulation process is probabilistically triggered. The effectiveness of the method is verified by comparing simulation results with experimental data， showcasing a high degree of correlation. Then， the thermal spread paths and their probabilities under different triggering conditions of the probability function are analyzed， revealing multiple potential routes for thermal runaway， including the jump thermal runaway event. The sequential thermal runaway path is identified as the most probable， while the jump phenomenon is deemed least likely. The proposal of this method further improves the consistency between numerical simulation and actual process of thermal runaway in lithium batteries， providing an effective research tool and analysis method for studying the probability of thermal runaway propagation in lithium battery modules.

In order to solve the problems of low efficiency and insufficient system scalability of single-machine test platforms in the vehicle-road-map collaborative simulation environment， a distributed autonomous driving simulation platform architecture for vehicle-road-map cooperative simulation is proposed in this paper， named VIMS （Vehicle-Infrastructure-Map System）. The VIMS platform uses CARLA as the virtual simulation engine. By introducing in real high-definition maps and connecting the hardware-in-the-loop devices such as driving simulators and signal machines to VIMS， the virtual-real traffic scene is formed. Considering the interaction of functions， the VIMS platform is divided into four modules， namely， the main world， the intelligent vehicle， the intelligent roadside， and the high-definition map， adopting ROS distributed architecture to realize the relative independence of the modules and interconnection between the modules. Considering the computational reliability and availability of the platform， distributed computing is used to realize independent computation among the four modules. Through the lane-keeping and vehicle-road-map collaborative positioning algorithm as examples for application validation， data acquisition， transmission and algorithm validation tests and evaluation are realized through the platform. The results show that the platform proposed in this paper can realize the real-time simulation of vehicle， road， and map collaboration to ensure that the modules operate organically and that the system architecture is highly scalable.

In order to reduce fuel consumption and transportation cost of heavy-duty truck， this paper coordinates the human-vehicle-road interaction system， integrates multi-dimensional information of vehicles and intelligent network environment， and proposes an adaptive range-domain predictive cruise control strategy （ARPCC） based on iterative dynamic programming （IDP）. Firstly， by combining the vehicle status and multi-dimensional information of the front environment， an adaptive distance domain model is established based on the longitudinal dynamics of the vehicle to reconstruct the road network， simplify the number of grids， and obtain the global optimal speed sequence by IDP. Secondly， on the basis of the global optimal speed sequence， the segmented optimal speed sequence taken from the adaptive distance domain is obtained to realize the fast solution of vehicle control state. Finally， Matlab/Simulink is used to verify the results， and the results show that the algorithm can effectively improve the computational efficiency and vehicle fuel economy by reducing the grid several times.

Accurate prediction of collision risk is crucial for ensuring the driving safety of intelligent vehicles. However， the risk differentiation among heterogeneity vehicle types and its coupled effect in longitudinal and lateral directions has rarely been considered in existing driving risk assessment methods. Therefore， firstly， the behavior patterns of drivers of heterogeneous vehicle types are explored to analyze the influence of vehicle types on drivers' sensitivity to risk in this paper. Secondly， the heterogeneous risk thresholds for different combinations of vehicle types are identified， and the risk differentiation in such traffic surroundings is further quantified based on two-dimensional indicators. Finally， the coupled two-dimensional collision risk prediction model considering vehicle types is proposed， and the effectiveness of the model is validated through comparative analysis. This research helps to enhance the driving safety of intelligent vehicles， which also can provide a theoretical foundation for the development of collision warning systems for human-driven vehicles.

For the data imbalance problem of the current automotive CAN network intrusion detection algorithm due to the lack of attack samples， a CAN intrusion detection data enhancement method based on BEGAN is proposed， which introduces in one-hot coding to image the CAN message features and combines with the constructed Generative Adversarial Network to generate valid samples with the same format as the real attack and with different content. The practicality of the generated enhanced dataset is verified from the perspectives of feature maps， t-SNE visualization， statistical analysis and classifier validation by collecting real vehicle data as real samples for training， which can improve the intrusion detection classifier accuracy. With higher accuracy compared with the traditional oversampling algorithms including Random Oversampling （ROS）， Synthetic Minority Oversampling Technique （SMOTE）， SMOTE combined with Edited Nearest Neighbors （SMOTE-ENN） and Adaptive Synthetic Oversampling （ADASYN）.

The load spectrum of commercial vehicle cab assembly is the key factor affecting the accuracy and computational efficiency of virtual fatigue prediction. In this article， key links such as road load spectrum collection and editing， high fidelity dynamic modeling， and virtual iteration are explored， in order to obtain accurate and efficient external point time-domain loads from the engineering application perspective. Firstly， the full path road load spectrum of the driver's cab assembly is collected from the actual vehicle in the test field， and the original data is normalized， split， and reassembled considering random errors to obtain a statistically strong total damage target in the test field. Then， using the principle of equal damage， 9 operating conditions and their number of cycles are optimized， which not only controls the error within 10%， but also increases the efficiency by 75%. Subsequently， based on the performance parameters of the measured damping components， a high fidelity rigid flexible coupling dynamic model of the cab is established， and the accuracy of the model is verified through a 7-channel road simulation bench in the cab. Finally， the load decomposition of the optimal operating conditions is completed through virtual iteration， with an iteration error of less than 10%. Based on the above optimization and decomposition of the external connection point load， the virtual fatigue calculation of the cab body is efficiently completed， and the failure of the cab welding points is accurately predicted， which has a high degree of consistency with the durability test results of the road simulation bench， providing strong technical support for the design and optimization of commercial vehicle cabins.

The road adhesion coefficient has an important impact on the vehicle dynamics control performance. In order to accurately obtain the road adhesion coefficient in real time and improve the estimation accuracy and convergence speed of the algorithm under different road surfaces and driving conditions， an interactive multiple model adaptive unscented Kalman filter （IMM-AUKF） based on the seven-degree-of-freedom vehicle dynamics model and Dugoff tire model is proposed in this paper for the distributed four-wheel-drive vehicles. The algorithm first introduces the improved Sage-Husa noise estimator into the UKF algorithm to construct the AUKF observer， which updates the measurement noise in real time and ensures the positive characterization of its covariance matrix， improves the weight of the new observation data， and enhances the real-time tracking accuracy and stability of the algorithm. Afterwards， the algorithm selects different observation variables to construct the longitudinal driving condition AUKF observer and the lateral-longitudinal coupling driving condition AUKF observer. And the IMM algorithm is also used to switch the observer model， so as to realize the algorithm's accurate estimation of the road adhesion coefficient under different driving conditions. The results of simulation tests on high/low attachment， joint and u-split roads and real vehicle road tests show that the proposed IMM-AUKF algorithm has higher estimation accuracy and faster convergence speed than the traditional UKF algorithm， and it can adapt to the real-time and accurate estimation of the road adhesion coefficient under different driving conditions.

The pose control method of the vehicle with dual motor active lateral stabilizer bar is studied in this paper. Firstly， a dynamic model of a dual motor active lateral stabilizer bar and an eight-degree-of-freedom vehicle model including roll， lateral， yaw pitch， and suspension vertical displacement are established. Secondly， for the problem that the control algorithm parameters are difficult to adjust under complex working conditions， the actual pitch angle estimation method， the ideal pitch angle calibration method， and the pitch condition identification method based on vehicle state information such as suspension height signal and road slope signal are proposed. The pitch sub-controller and the roll sub-controller based on PID control algorithm are designed with the dual motor active lateral stabilizer bar as the actuator. Genetic algorithm is used to tune the parameters of each sub-controller. Finally， combined with the pose control matrix， the vehicle pose joint control algorithm is designed and verified through experiments. The Hardware in Loop results of MATLAB/Simulink CarsimRT and Rapid ECU show that the improvement of the roll angle， roll angle speed， and pitch angle of the vehicle equipped with dual motor active lateral stabilizer bar is more than 10% under different complex working conditions， which proves the feasibility and universality of the control algorithm.