Artificial intelligence (AI) models, with their strong generalization and multi-task learning capabilities, have demonstrated extensive application potential in intelligent connected vehicles. This paper summarizes the challenges of the application of AI models in driving automation, analyzes the technical route of driving automation models, and the supporting platform technology of driving automation model development and validation, summarizes the application of intelligent cockpit, and explores the method for constructing scenario generation models based on large language models. From the perspectives of AI security and data governance, this paper summarizes the security governance practices associated with the application directions for AI models, providing a reference for the safety assessment and management of AI-related applications.

The application potential of on-device intelligent technology in distributed computing, with its inherent advantages in timeliness and privacy preservation, is progressively expanding within intelligent connected vehicles. Based on on-device intelligent principles, this paper proposes a device-cloud collaborative framework for expedited recommendation functionalities in intelligent connected vehicles, thereby establishing a novel technical solution for intelligent networking through on-device computing. The architecture incorporates the design of on-device model training and inference mechanisms that enable real-time local data analysis and decision-making at vehicular terminals. This implementation facilitates personalized shortcut recommendations tailored to occupants’ preferences. Experimental results demonstrate that the proposed scheme significantly enhances users’ experience while providing innovative insights into practical implementations of on-device intelligent technology for intelligent connected vehicles. This research contributes new perspectives for Intelligent connected vehicles applications through on-device.

The event-tracking data of intelligent cockpit contains rich information about driver and passenger actions. Analyzing and identifying specific action intents can benefit deeper insights into user needs. Considering the high cost, strong subjectivity, and repetitiveness of current methods that rely on manual tagging for action intent recognition, a new method based on Artificial Intelligence model for automated tagging and classification is proposed. By fine-tuning the Qwen2-14B model, this approach could rapidly identify action intents across multiple dimensions and granularities, enhance the efficiency of cloud data analysis and lay a theoretical foundation for real-time response to user needs on the vehicle side.

The Noise Vibration Harshness (NVH) performance of vehicles is one of the key indicators of overall vehicle qualities. To enhance ride comfort and meet increasingly stringent NVH requirements, this paper reviews the application of Artificial Intelligence (AI) algorithms in NVH optimization, both domestically and internationally. It analyzes feasible approaches for improving NVH performance using AI-based methods and discusses future trends and challenges in AI-driven NVH optimization. The study aims to provide valuable insights for leveraging intelligent algorithms to address automotive performance enhancement.

To enhance lateral active safety in intelligent connected vehicles, the identification and definition of unintended functional safety scenarios for the Electric Power Steering (EPS) system are integrated into system design and development process. By combining scenario domain analysis, equivalence partitioning, and boundary value analysis with the deconstruction of vehicle model variants and lateral motion scenarios related to EPS, more potential hazardous operational scenarios are expanded from known operational scenarios. Meanwhile, the Driver-In-the-Loop (DIL) system based on hardware-in-the-loop simulation is utilized. Relying on its advantages of “driver-vehicle-environment” closed-loop, more unknown risk scenarios are transformed into valid known scenarios. Subsequently, the existing safety technical requirements are optimized and the overall performance of the EPS system is further improved.

To optimize the opening and closing forces of the automotive Power Lift Gate (PLG) system, the literature on algorithms related to lift gate forces is reviewed, and a mechanical model of the lift gate system is established. The expressions for the moment exerted by the support rod on the lift gate and the moment exerted by gravity on the lift gate are derived, and the relationships between the length of the support rod, the force arm of the support rod, and the opening angle are analyzed. A simulation model is created using simulation software to perform calculations, and the rationality of the expressions is verified. This study provides a theoretical foundation and optimization direction for the design and improvement of PLG systems.

Europe has become one of China’s important automobile export markets. In order to enhance the product competitiveness of Chinese automobiles and assist Chinese automobile companies in conducting overseas adaptability tests. The market application of new energy vehicle technology and Internet Of Vehicle (IOV) in the European automotive market is studied. Based on the characteristics of automotive operating conditions in the European region, the European road test projects and necessary preparatory work are proposed. The results indicate that the European automotive market has broad application prospects in the fields of new energy vehicles and connected vehicles, and Chinese automotive products should pay attention to product adaptability development in the overseas application process.

Aiming at the reliability evaluation problem of small samples of long-life high-voltage relays, an analysis method of mean rank combined with least squares is proposed. Under the guidance of the basic principle of average rank method, the method of determining the fault rank of high voltage relay under random truncated multiple tests is studied. The least square method is used to estimate the reliability parameters, and the linear correlation coefficient and D-test method is used to test the fit. Under the condition that the fault data are subject to 2 distribution models at the same time, the total root-mean-square error method is introduced to optimize the distribution model, identify the optimal reliability model and conduct reliability evaluation.

The newenergy vehicle industry is faced with comprehensive upgrading and rapid competition, but the product development cycle is continuously shortened, and synchronization brings more severe test on efficiency and cost of R&D equipment. In order to further improve the equipment R&D efficiency and quickly meet the new energy high-end product development needs. A kind of predictive maintenance system for R&D equipment is researched and designed in depth. Fault diagnosis and life prediction algorithm model are developed through key technologies such as Internet of Things, wavelet transform, deep learning, multiple Gaussian distribution and long and short time memory neural network, so as to realize the prediction of key faults and remaining life of equipment. The results show that the system can significantly reduce the downtime and maintenance time, and achieve more efficient use of R&D and maintenance resources.