面向自动驾驶的5G边缘计算业务连续性技术研究

为了提升在5G边缘计算场景下的自动驾驶业务连续性，保障车辆在高速移动下的鲁棒性和实时性，解决由于车载设备与5G边缘计算之间的通讯网络路径发生突变和变长，导致网络连接不可靠和拥塞丢包的问题，提出了增强型随机早期检测算法(stochastic fair-random early detection, SF-RED)和端到端双连接双链路技术保障业务鲁棒性，并提出硬件加速技术保障业务实时性。增强型SF-RED算法通过统一资源定位器(uniform resource locator, URL)识别应用，并基于不同应用来设计数据报文转发路径策略，实现基于多路径的转发和数据缓存以及报文拥塞下的丢弃控制，进一步改善丢包队列的管理。通过自动驾驶应用的实际案例证明，采用增强型SF-RED和端到端双连接双链路技术能够有效提升高速移动的自动驾驶车辆业务可用性，采用的硬件加速技术显著降低其业务时延。

Research on 5G edge computing service continuity technology for automatic driving

To improve the auto driving service continuity in the 5G edge computing scenario and ensure the robustness and timeliness of vehicles in the high-speed movement, the enhanced SF-RED(stochastic fair-random early detection) and end-to-end dual-link technology are proposed to guarantee the service robustness for the communication between the on-board equipment and 5G edge computing due to the sudden change and change of network path, leading to the problem of unreliable network connection and packet loss due to congestion. The hardware acceleration technology is proposed to guarantee real-time service performance. The enhanced SF-RED identifies the applications through URL (uniform resource locator) and designs the packet forwarding path policy based on different applications to implement multi-path-based forwarding, data buffering and packet discard control in case of packet congestion. In this way, the management of packet loss queue is improved. The application of the enhanced SF-RED and the end-to-end dual-link technology are shown to be able to effectively improve the auto driving service availability at a high speed, and the hardware acceleration technology is used to significantly reduce the service delay.