To address the high computational and storage demands of object detection algorithms that limit real-time performance of edge devices, this study proposes an improved lightweight road object detection algorithm based on YOLOv7-tiny. First, prior anchor boxes optimized for road object detection are generated using the K-means++ clustering algorithm. Secondly, the backbone network is streamlined by modifying the ELAN structure, while a Lightweight Multi-scale Feature (LMS) module is introduced to optimize the neck network. Finally, the Sigmoid Linear Unit (SiLU) activation function is adopted to accelerate model convergence, and the MPDIoU loss function is employed to further improve detection accuracy. Experimental results demonstrate that the improved model achieves an 18.3% reduction in parameters, a 15.0% decrease in computational complexity, and a 1.1% increase in mean average precision across all categories. When deployed on Jetson TX2 with TensorRT acceleration, the detection speed reaches 48 frames per second, essentially meeting real-time requirements for road object detection applications.

Based on heat pump technology, an integrated thermal management system that fully utilizes the waste heat of the motor has been designed. The system uses heat exchangers to connect the independent circuits and achieve efficient energy utilization. In order to address the difficulty of controlling thermal management systems, this paper proposes two optimization fuzzy control approaches, anti-saturation integral fuzzy control and multi-level fuzzy control. An integrated thermal management system model is built based on AMESim, and Simulink control strategy models for working mode switching and key components are established to jointly simulate and analyze the thermal management control effect of the entire vehicle. The simulation results show that at 0 ℃, the integrated thermal management system reduces the cabin heating time by about 27.8% compared with the independent thermal management systems of each circuit, and the energy efficiency ratio has been increased by an average of about 31.3%, in addition the winter driving range has been increased by about 9.57%. The control effect of optimized fuzzy control is significantly improved with the heating time of the cab in winter shortened by about 18.4%, and the fluctuation and overshoot of the cabin temperature in summer reduced, and the battery cooling time shortened by about 3.6%.

This study aims to enhance the lateral stability of distributed drive electric vehicles when traveling on highways and low-adhesion roads by adopting a hierarchical control strategy. The upper level is the torque decision-making layer, which designs a hierarchical Sliding Mode Control (SMC) based on a two-degree-of-freedom dynamic model to optimize the additional yaw moment and introduces Active Rear-wheel Steering (ARS), while simultaneously using PID control to achieve vehicle speed tracking. The lower level is the torque distribution layer, which optimizes the distribution of driving torque among the four wheels based on wheel load and road adhesion coefficient. Through co-simulation using Carsim and Matlab/Simulink, the control effectiveness was verified under double lane change, different adhesion road surface, and slalom conditions. The results indicate that the SMC+ARS control maintains high stability under high-speed double lane change conditions and outperforms SMC under various adhesion conditions, while also reducing energy consumption during turning.

To address the challenges associated with the complex steps and inability to identify the optimal hardware parameters of the CLLLC converter when designing parameters by using the Fundamental Harmonic Analysis (FHA) method, this paper proposes a CLLLC converter parameter design and optimization approach based on the fundamental wave analysis method and the Dung Beetle Optimizer (DBO) algorithm. Firstly, the FHA is employed to derive the design boundary of the converter parameters as the design constraints. Secondly, the efficiency function of the converter is established according to the relationship between converter efficiency and hardware parameters. Then, the DBO algorithm is used to optimize the objective function within the design constraints to obtain the hardware parameter values at the best efficiency point. The experimental results indicate that the efficiency of the designed prototype can reach 97%, further proving feasibility of this scheme.

To address the issue of low accuracy in estimating the State of Charge (SOC) of lithium batteries using the Unscented Kalman Filter (UKF) algorithm, a combined ELM-UKF algorithm with a state detection mechanism is proposed, leveraging the complementary advantages of Extreme Learning Machine (ELM) and UKF for estimating the SOC of lithium batteries. Firstly, the algorithm uses the relevant filtering data estimated by UKF for battery SOC as a sample set to train the ELM model. The successfully trained ELM model is then used to online compensate for the SOC estimation error of UKF, thereby achieving real-time correction of estimation deviations. Secondly, the algorithm designs a state detection mechanism for the predictive output of the ELM model to reduce the impact of overfitting in the ELM model’s predictive output on the smoothness of the SOC estimation waveform. Experimental results show that, compared to single-type algorithms, the proposed combined algorithm exhibits good robustness and generalization, effectively enhancing the estimation performance of lithium battery SOC.

In order to improve the accuracy of predicting the Remaining Useful Life (RUL) of batteries, an intelligent digital-analogue fusion method of Particle Swarm Optimization (PSO) optimized Extreme Learning Machine (ELM) combined with Random Perturbation Untraceable Particle Filtering (RP-UPF) is used to predict the RUL of batteries B0005, B0006 and B0018 based on fusion of the health indexes and the constructed battery capacity decline model. The research results show that the proposed intelligent digital-analogue fusion method not only significantly improves the accuracy of battery RUL prediction, but also maintains high prediction accuracy throughout the life cycle of the battery.

In order to enhance the vehicle braking energy recovery efficiency and maintain braking stability, this paper proposes a comprehensive energy recovery method that takes the drivers’ styles and road adhesion characteristics into account. Firstly, the driving style feature parameters are extracted from different drivers’ data, and the driving style recognition model is trained based on the Support Vector Machine (SVM). Then, the road images are preprocessed using the U-Net network, and the lightweight network MobileNet V3 is trained to recognize the road surface efficiently. Finally, combined with the recognition results of driving styles and road surfaces, the variable ratio of braking force of the front and rear axles of the vehicle is allocated, and a method is proposed to determine the regenerative braking force considering the weight of driving styles and road adhesion conditions, the braking energy recovery strategy is formulated on this basis. The simulation results show that the braking efficiency and stability are significantly improved for different road surfaces; the SOC of the battery is improved by 2.13 percentage points and 1.02 percentage points in the WLTC and NEDC cycle conditions respectively, further improving the overall braking stability and economy of vehicles.

To Improve the comprehensive economy of Hybrid Electric Velicle (HEV) in driring, aiming at the problem of insufficient adaptability of existing HEV energy management strategies to drivers with different driving styles an adaptive equivalent consumption minimization energy management strategy based on real-time recognition of driving styles is proposed. Firstly, the changes in driving demand caused by the real-time pedal signal variation are analyzed, and the real-time driving styles in the rolling time window by data characteristic extraction is identified. Then, in the equivalent consumption minimization strategy considering the cost of battery degradation, and an adaptive function of equivalent factor is constructed by quantifying the driving style coefficient, and the optimal equivalent factor and battery degradation cost weighting coefficient corresponding to different driving styles are solved. Based on this, an adaptive law for equivalent factor based on artificial potential field method is designed. The simulation results showed that, compared with the equivalent energy consumption minimization without consideration of driving style, the proposed method improves fuel economy by 2.82% while reducing the battery loss by 25% under the same driving condition.

To address the issues of a low recovery rate of motor braking energy and significant fluctuations in the DC bus voltage, a braking energy tracking strategy is put forward. This strategy can adjust the reference value of the supercapacitor current in real time by considering multiple parameters, including motor power, the efficiency of the bidirectional DC/DC converter, and the supercapacitor voltage. The discrete state space model of the energy storage unit is used to predict the current and voltage of the supercapacitor, and the current reference value of the supercapacitor is calculated by combining the other parameters in the system. The current loop based on Proportional-Integral (PI) is used to control the current of the supercapacitor and track the reference value of the current in real time. Simulation and experiments compared the tracking of regenerative energy under two different control strategies. The results indicate that the proposed strategy enhances the efficiency of regenerative energy recovery from 55.93% to 86.76%, and it restricts the voltage fluctuation of the DC bus within 0.9%.

With the increasing applications of new technologies such as smart driving, autonomous driving and Internet connectivity, traditional in-vehicle networks are difficult to meet the Quality of Service (QoS) demands of diverse applications. In order to improve the data transmission rate and guarantee the QoS demand of services in the in-vehicle network, a deep reinforcement learning QoS routing algorithm based on SDVN is designed in combination with Software-Defined Vehicular Network (SDVN) technology. The algorithm can realize intelligent control and optimized management of data transmission in the in-vehicle network to ensure the control, distribution and monitoring of in-vehicle network traffic and improve the quality and efficiency of in-vehicle data transmission. The experimental results show that the routing algorithm can better reduce the delay of the in-vehicle network and has better optimization performance compared to the traditional routing algorithm.