A Multi-Scenario Multimodal Simulation Dataset for Vehicle–Infrastructure Cooperative Perception(车路协同感知仿真数据集)

This dataset was generated using CARLA 0.9.12 and SUMO 1.12.0 co-simulation within the OpenCDA 0.1.2 framework to support the testing and evaluation of vehicle–infrastructure cooperative perception algorithms. It comprises 200 traffic scenarios across five road environments: urban road segments, urban intersections, highway segments, highway curves, and highway ramps. Each scenario includes at least two roadside sensing nodes and two vehicle-mounted sensing nodes, with 100 frames collected per node. Each vehicle-side frame contains images from four RGB cameras and one bird's-eye-view (BEV) camera, a LiDAR point cloud, and a YAML metadata file recording sensor positions and intrinsic parameters, ego-vehicle speed, planned trajectory, and predicted and ground-truth ego-vehicle positions. Each infrastructure-side frame contains images from four RGB cameras, a LiDAR point cloud, and a YAML metadata file recording sensor positions and intrinsic parameters, along with the predicted and ground-truth positions of the ego vehicle. By providing multimodal vehicle-side and infrastructure-side observations across diverse road environments, the dataset supports research on cooperative perception, multisensor fusion, and related algorithm testing.(本数据集基于OpenCDA 0.1.2框架,采用CARLA 0.9.12与SUMO 1.12.0联合仿真采集生成,面向车路协同感知算法的测试与评估。数据集涵盖城市基本路段、城市交叉口、高速公路基本路段、高速公路弯道和高速公路匝道5类道路环境,共包含200个交通场景。每个场景配置至少2个路端感知节点和2个车端感知节点,并为各节点采集100帧数据。每帧车端数据包含4个RGB相机图像、1个鸟瞰视角(BEV)相机图像、1个激光雷达点云及1个YAML元数据文件,记录传感器位置与内参、自车速度、规划轨迹以及自车的预测位置与真实位置。每帧路端数据包含4个RGB相机图像、1个激光雷达点云及1个YAML元数据文件,记录传感器位置与内参,以及主车的预测位置与真实位置。数据集提供多类道路场景下的车端与路端多模态观测数据,可用于车路协同感知、多传感器融合及相关算法的测试研究。) Full Dataset Download URL: https://www.openits.cn/articles/169