Vehicular applications such as cooperative driving, teleoperation, and real-time perception increasingly rely on low-latency wireless communication. In this context, ITS-G5, based on IEEE 802.11p, represents a key technology for enabling direct vehicle-to-vehicle and vehicle-to-infrastructure communication. Despite its relevance, experimental studies focusing on the performance of UDP-based traffic over IEEE 802.11p under realistic conditions remain limited. This paper presents an experimental evaluation of UDP transmission over an IEEE 802.11p ITS-G5 testbed composed of Raspberry Pi-based onboard units and commercial roadside units. The analysis investigates the impact of different modulation and coding schemes (MCS). It also evaluates two network-layer configurations (IPv4 unicast and IPv6 multicast) and the use of CAKE for active queue management. In addition to synthetic traffic generated with iPerf, the evaluation includes real-time video streaming using MPEG-TS over UDP to emulate latency-sensitive vehicular applications. Results show that the modulation scheme is the dominant factor influencing latency at low traffic loads, while the choice of transmission mode and IP version becomes increasingly significant under congested conditions. Higher-order modulations significantly reduce latency and variability, whereas IPv6 multicast exhibits greater delay dispersion than IPv4 unicast. Furthermore, active queue management does not seem to improve delay predictability. These findings provide practical insights for configuring ITS-G5 networks supporting latency-sensitive vehicular services.

翻译：诸如协同驾驶、远程操控和实时感知等车载应用，日益依赖于低延迟无线通信。在此背景下，基于IEEE 802.11p的ITS-G5技术成为实现车与车、车与基础设施直接通信的关键技术。尽管其重要性显著，但针对IEEE 802.11p在真实条件下承载UDP流量性能的实验研究仍较为有限。本文通过在由树莓派车载单元与商用路侧单元组成的IEEE 802.11p ITS-G5测试平台上开展UDP传输实验评估，分析了不同调制与编码方案（MCS）的影响。实验同时评估了两种网络层配置（IPv4单播与IPv6组播）以及采用CAKE进行主动队列管理的效果。除使用iPerf生成合成流量外，评估还包含基于UDP传输MPEG-TS流实现的实时视频传输，以模拟对延迟敏感的车载应用场景。结果表明：在低流量负载下，调制方案是影响延迟的主导因素；而在拥塞条件下，传输模式与IP版本的选择对延迟的影响愈发显著。高阶调制可显著降低延迟与抖动，而IPv6组播相较于IPv4单播表现出更大的延迟离散度。此外，主动队列管理似乎未能改善延迟的可预测性。这些发现为配置支持延迟敏感车载服务的ITS-G5网络提供了实践指导。