In order to accelerate the large-scale commercialization of autonomous driving system and provide the remote control capability for autonomous vehicles, this paper proposes a remote driving system by integrating 5G communication network, cloud control platform, video monitoring and other technologies. The vehicle can connect with remote users through the cloud platform, and send the vehicle speed, gear, position, tire pressure and other information to the system in real time. Drivers can monitor the vehicle status information in real time through the system so as to make comprehensive decisions, and realize remote driving. It can provide safety redundancy and strong guarantee for the commercial landing of advanced automatic driving. The framework of the remote driving system based on 5G technology and the architectural design based on each part of the system are analyzed in this paper. Through the real vehicle test, the performance of the cloud-control remote driving system and its feasibility and superiority in application are verified.

In-vehicle software under the new automotive electronic and electrical architecture needs to be reusable, easy to expand, loosely coupled, compatible and interoperable. In order to abstract the application program on automobile ECU into a service, a distributed service framework based on VSOME/IP is proposed based on the open source distributed communication middleware VSOME/IP and the remote procedure calling framework CommonAPI C++. The framework utilizes francadil service interface description language to improve the efficiency of developers in building services; Through the routing manager, the service registration center component in the service framework is realized, which provides services publishing, service subscription, status synchronization, message notification and other functions for the distributed system. SOME/IP serves as the underlying communication protocol to provide the system with publish-subscribe and request-response service invocation methods.

With the advent of the software-defined automobile era, the necessity of the application of automobile software update technology becomes more and more prominent. To avoid potential security risks, vehicle manufacturer needs to ensure OTA security of the product. In view of the current challenges faced by OTA, the development status of key technologies for OTA update is summarized from the aspects of data integrity and service integrity. The requirements of existing software update management laws and regulations are analyzed, and the corresponding OTA capacity building suggestions are put forward to vehicle manufacturers.

Based on the current status of lane recognition based on road feature model under the background of intelligent driving, the gray processing algorithm, filtering processing algorithm, and region of interest extraction technology in image preprocessing for intelligent vehicles are compared and analyzed respectively, and the applicability of different image preprocessing methods in lane recognition algorithm is studied. The principle of edge detection technology and the conversion algorithm of road feature conditions in the real-time lane extraction algorithm are comprehensively used and analyzed. The lane recognition algorithm model based on road features is built. After verification on the Visual Studio, the algorithm model meets the lane recognition requirements of intelligent driving vehicles.

On the basis of the existing detection network structure, a new decoupling prediction branch is added to realize the multi-task recognition of pedestrian position, head position, pedestrian age and occlusion attributes under the vehicle-mounted camera in this paper. The influence of classification branch on detection performance at different network locations is demonstrated by experiments. Aiming at the problem that some nonlinear activation function operators in engineering applications are not supported by the deployment environment, a method using piecewise linear function to fit the nonlinear activation function is proposed, which not only improves the performance of network identification but also provides convenience for engineering deployment.

In high-speed digital circuits, electrical signal crosstalk is one of the main factors affecting signal integrity. The theoretical model of the near end electrical signal crosstalk noise and the far end electrical signal crosstalk noise is established in this paper, and some parameters affecting electrical signal crosstalk are simulated by using the signal integrity simulation tool HyperLynx, such as the distance between aggressor line and victim line, the coupling length, the thickness of dielectric layer from the signal line to reference plane and the rates of rising as well as falling edges of signals. The simulation results show that the reasonable design can effectively suppress the electrical signal crosstalk and improve the signal integrity of the circuit.

In order to intuitively evaluate the driving control performance of the vehicle based on the actual travel scenarios of Chinese consumers, this paper proposes a driving control performance testing and evaluation method. According to Chinese road conditions, combined with engineering experience and national standards, driving control performance test conditions and the evaluation indexes and weights in the test conditions are selected. Three vehicles are selected for the road test under the test conditions, and obtain the index limits through data analysis. The results show that the driving control performance scores of the three vehicles are related to the level and energy type of the vehicle，are consistent with the subjective driving experience. It indicates this driving performance testing and evaluation method is feasible and accurate, and can be used for strategy calibration and validation evaluation of the entire vehicle driving performance for engineering samples in the development or validation stage.