典型场景交通流预测对比数据集

随着智能网联汽车渗透率不断提升，其与人工驾驶车辆的长期共存使混行交通流成为城市交通系统的常态形态。不同自动化水平车辆在跟驰、换道与协同机制上的差异，使交通运行呈现更强的非线性、不确定性与时空异质性，对交通状态刻画与云控策略效果评估提出了更高的数据支撑需求。为此，本文基于自主构建的混行交通流云控仿真测试平台，生成并整理了一套面向多场景、多渗透率的混行交通流仿真数据集。该数据集以SUMO微观交通仿真引擎为核心，融合真实城市路网结构、统一的24小时潮汐交通需求模型及精细化车辆行为建模，在可控环境下高保真复现混行交通运行状态。数据覆盖长直路、匝道合流区和信号化交叉口等典型场景，在不同CAV渗透率及有无云控与协同管控策略条件下生成对比数据。数据内容既包括车辆级轨迹、速度与延误等微观信息，也涵盖路段与区域级流量、密度等宏观指标，能够支撑交通流预测、效率评估及云控性能演化研究。通过统一路网、车流输入与实验配置，该数据集具备良好的一致性、可比性与可复现性，为混行交通流建模、优化控制与车路云协同研究提供了系统化的数据基础。

With the increasing penetration rate of Connected and Automated Vehicles (CAVs), the long-term coexistence of CAVs and human-driven vehicles has made mixed traffic flow a normal state of urban traffic systems. The differences in car-following, lane-changing and cooperative mechanisms among vehicles with different automation levels make traffic operations exhibit stronger nonlinearity, uncertainty and spatiotemporal heterogeneity, which puts higher data support requirements for traffic state characterization and the effect evaluation of cloud-based control strategies. To address this demand, this paper generates and organizes a multi-scenario, multi-penetration mixed traffic flow simulation dataset based on an independently developed cloud-based control simulation test platform for mixed traffic flow. Taking the SUMO microscopic traffic simulation engine as the core, this dataset integrates real urban road network structures, a unified 24-hour tidal traffic demand model and fine-grained vehicle behavior modeling to achieve high-fidelity reproduction of mixed traffic operation states in a controllable environment. The dataset covers typical scenarios including straight road sections, ramp merging zones and signalized intersections, and generates comparative data under different CAV penetration rates and conditions with or without cloud-based control and cooperative management strategies. The data content includes both microscopic information such as vehicle-level trajectories, velocity and delay, as well as macroscopic indicators such as road segment and regional-level flow and density, which can support traffic flow prediction, efficiency evaluation and research on the performance evolution of cloud-based control. With unified road network, traffic flow input and experimental setup, this dataset has good consistency, comparability and reproducibility, providing a systematic data foundation for mixed traffic flow modeling, optimal control and vehicle-road-cloud coordination research.

搜集汇总

背景与挑战

背景概述

该数据集是一个面向混行交通流的仿真对比数据集，基于SUMO微观交通仿真引擎构建，覆盖长直路、匝道合流区和信号化交叉口等典型场景，包含不同智能网联汽车渗透率及云控策略条件下的车辆轨迹和宏观交通指标数据。它旨在支持交通流预测、效率评估和车路云协同研究，具有良好的一致性和可复现性，数据量为10.02GB，包含811个文件。