高速工况云控自动驾驶车辆队列数据集

为支撑高速公路场景下云控自动驾驶车辆队列控制与协同驾驶研究，构建并公开发布
高速工况云控自动驾驶车辆队列实车测试数据集。数据采集于支持自动驾驶试验的公开高速
公路（荣乌高速公路新线霸州北—定兴东约 46 km）开展的三车与十车队列试验，时间范围
覆盖 2025 年 9 月。数据包含两类：①三车队列车端本地记录的 CAN 总线原始报文（.asc），
提供帧级时间戳、 CAN ID、 DLC 及数据字节，可结合配套 DBC 进行信号级解码；②十车
队列车端解析关键信号并通过 MQTT 上传云端汇聚形成的结构化数据（.csv），包含云端接
收时间（毫秒级精度）、车辆编号/角色、车速及部分车辆 GNSS 经纬度等字段，并提供十
车队列长距离稳定行驶与云控/非云控舒适性对比数据。预处理遵循“尽量保留原始记录”
原则， 进行文件完整性检查、字段规范化、时间排序去重、缺失标注与轻量尖刺抑制。该数
据集是面向公开高速道路多车型十车队列的实车测试数据，具有较高的工程实用价值，可用
于云控队列控制算法验证、队列行为分析、通信链路与运行稳定性评估以及教学与基准测试。

To support research on cloud-controlled autonomous vehicle platoon control and cooperative driving in highway scenarios, this real-world test dataset for cloud-controlled autonomous vehicle platoons under high-speed operating conditions has been constructed and publicly released.
The data was collected from 3-vehicle and 10-vehicle platoon tests conducted on the publicly accessible highway designated for autonomous driving experiments (Xiong'an-Ulanhot Expressway New Line, approximately 46 km between Bazhou North and Dingxing East), covering the period of September 2025.
The dataset comprises two categories of data: ① Raw CAN bus messages (.asc format) recorded locally on the on-board terminals of the 3-vehicle platoon, which provide frame-level timestamps, CAN IDs, DLC (Data Length Code) and data bytes, and can be decoded at the signal level with supporting DBC files; ② Structured data (.csv format) aggregated in the cloud after being uploaded via MQTT protocol by the on-board terminals of the 10-vehicle platoon, which had parsed key signals in advance. This CSV dataset includes fields such as cloud reception time (millisecond-level precision), vehicle ID/role, vehicle speed, and GNSS latitude and longitude of some vehicles, as well as comparative data on driving comfort between long-distance stable operation of 10-vehicle platoons under cloud-controlled and non-cloud-controlled conditions.
Preprocessing follows the principle of "retaining original records as much as possible", including file integrity checks, field standardization, time-based sorting and deduplication, missing value annotation, and mild spike suppression.
This dataset consists of real-world test data for multi-type vehicle 10-vehicle platoons on public highways, with high engineering practical value. It can be applied to the verification of cloud-controlled platoon control algorithms, platoon behavior analysis, communication link and operational stability evaluation, as well as teaching and benchmark testing.

搜集汇总

背景与挑战

背景概述

该数据集是一个针对高速公路场景的云控自动驾驶车辆队列实车测试数据集，采集于2025年9月的荣乌高速公路新线，包含三车队列的CAN原始报文和十车队列的云端结构化数据，经过预处理后，可用于云控队列控制算法验证、队列行为分析和工程教学等研究。