Autonomous driving has now made great strides thanks to artificial intelligence, and numerous advanced methods have been proposed for vehicle end target detection, including single sensor or multi sensor detection methods. However, the complexity and diversity of real traffic situations necessitate an examination of how to use these methods in real road conditions. In this paper, we propose RMMDet, a road-side multitype and multigroup sensor detection system for autonomous driving. We use a ROS-based virtual environment to simulate real-world conditions, in particular the physical and functional construction of the sensors. Then we implement muti-type sensor detection and multi-group sensors fusion in this environment, including camera-radar and camera-lidar detection based on result-level fusion. We produce local datasets and real sand table field, and conduct various experiments. Furthermore, we link a multi-agent collaborative scheduling system to the fusion detection system. Hence, the whole roadside detection system is formed by roadside perception, fusion detection, and scheduling planning. Through the experiments, it can be seen that RMMDet system we built plays an important role in vehicle-road collaboration and its optimization. The code and supplementary materials can be found at: https://github.com/OrangeSodahub/RMMDet

翻译：自动驾驶现已借助人工智能取得巨大进展，并提出了多种用于车辆端目标检测的先进方法，包括单传感器或多传感器检测方法。然而，真实交通场景的复杂性和多样性要求我们研究如何将这些方法应用于实际道路条件。本文提出RMMDet，一种面向自动驾驶的路侧多类型多组传感器检测系统。我们采用基于ROS的虚拟环境模拟真实世界条件，特别是传感器的物理与功能构建。随后，在该环境中实现多类型传感器检测与多组传感器融合，包括基于结果级融合的摄像头-雷达与摄像头-激光雷达检测。我们制作了局部数据集与真实沙盘场地，并开展了多项实验。此外，我们将多智能体协同调度系统与融合检测系统相连接。由此，完整的路侧检测系统由路侧感知、融合检测与调度规划三部分构成。实验表明，我们构建的RMMDet系统在车路协同及其优化中发挥了重要作用。代码及补充材料可见：https://github.com/OrangeSodahub/RMMDet