城市工况云控自动驾驶车辆数据集

本数据集面向城市道路环境下自动驾驶车辆纵向控制性能研究，在典型城市纵
向跟驰及交叉口通行工况下，设置云控策略与智能驾驶员模型（IDM)两种控制条件采集数
据。数据以高时间分辨率时序形式记录，时间步长约为 10-2 s，覆盖车辆从起步、加速、稳
定跟驰到减速停车的完整过程，时间范围包含不同信号配时和车流波动情形。空间范围聚
焦于具有信号灯控制、车流扰动和限速约束的城市道路纵向区段。数据采用统一的 JSON
与表格文件格式存储，字段结构包括时间步、车辆实际速度与目标速度、加减速控制量、
制动/驱动请求及电池荷电状态等关键变量，并配套速度曲线图用于直观展示不同控制策略
下的速度跟踪轨迹。数据生成过程中对字段取值和单位进行一致性检查，对多源数据时间
轴进行对齐，并剔除明显违背车辆动力学特性的异常点，以保证时序连续性和物理合理
性。本数据集可为云控车辆与 IDM 纵向性能评估、城市交通流建模与仿真、强化学习与模
型预测控制训练样本构建以及车路云数字孪生验证等研究提供基础支撑，具有较高的重用
价值与应用潜力。

This dataset is designed for research on the longitudinal control performance of autonomous vehicles in urban road environments. Data was collected under two control conditions: cloud-based control strategy and Intelligent Driver Model (IDM), covering typical urban longitudinal car-following and intersection traffic scenarios. The data is recorded as high temporal resolution time series with a time step of approximately 10^-2 s, covering the complete process of vehicles from starting, accelerating, stable car-following to decelerating and stopping, with time ranges covering different signal timings and traffic flow fluctuation scenarios. The spatial scope focuses on longitudinal sections of urban roads with signal light control, traffic flow disturbances and speed limit constraints. The data is stored in unified JSON and tabular file formats. The field structure includes key variables such as time step, actual vehicle speed and target speed, acceleration/deceleration control commands, brake/drive requests, and battery state of charge (SOC). Supporting speed curve plots are provided to intuitively display speed tracking trajectories under different control strategies. During the data generation process, consistency checks are performed on field values and units, time axes of multi-source data are aligned, and outliers that obviously violate vehicle dynamic characteristics are removed to ensure temporal continuity and physical plausibility. This dataset can provide basic support for research such as longitudinal performance evaluation of cloud-controlled vehicles and IDM, urban traffic flow modeling and simulation, reinforcement learning and model predictive control training sample construction, and vehicle-road-cloud digital twin validation, and has high reuse value and application potential.

搜集汇总

背景与挑战

背景概述

该数据集专注于城市道路环境下自动驾驶车辆的纵向控制性能研究，通过云控策略和智能驾驶员模型（IDM）两种控制条件，采集了高时间分辨率（时间步长约0.01秒）的时序数据，覆盖车辆从起步到停车的完整过程。数据以JSON和表格格式存储，包含速度、控制量等关键变量，适用于车辆性能评估、交通流建模和数字孪生验证等领域，具有较高的重用价值。