德清自动驾驶标准法规仿真测试场景库数据

本自动驾驶仿真场景库适用于自动驾驶技术研发、车辆测试及安全评估等阶段，需高性能计算机与专业人员操作。涵盖城市道路各个场景，助力算法优化、功能测试与安全评估，提升自动驾驶性能与安全性，适用于研发企业、汽车制造商、科研机构等，解决算法性能提升、功能验证及安全风险评估等问题。基于德清智能网联汽车测试场路网数据，经解析转化成OpenDrive通用格式数据。再依据仿真测试相关政策规范和行业标准，使用自研场景库编辑工具，严格依据相关标准，在工具可视化界面编排和制作场景，再由编辑工具转化为OpenScenario格式的场景库结构化数据。最终将政策法规中的抽象场景解析成为具体的以姿态、动作、事件、操作、行为、故事、实体等组成的结构化数据表述。数据内容主要包括实体姿态（航向角、俯仰角、翻滚角）、实体位置（x、y、z）、场景中的相关动作（动作名称字段和动作类型字段。动作类型包括针对特定实体如车和人执行的专属动作PrivateAction，对整个场景或多个实体同时生效的全局动作GlobalAction，用户定义动作UserDefinedAction）、事件（事件名称字段和事件优先级字段。定义动作的触发条件和执行逻辑，包括基于条件触发的事件、基于时间触发的事件、基于状态触发的事件、基于用户定义的事件、全局事件、专属事件等）、操作和操作组（用于定义交通参与者行为的高级结构，通常包含一系列事件和动作）、行为（定义交通参与者执行的具体动作和操作）、故事版（整个场景的完整描述，包含了动作、操作、事件等）、实体（交通参与者，包括车辆、行人和其他目标）等。测试场景覆盖智能网联汽车和低速功能型无人车。确保测试的全面性、真实性和可重复性，满足自动驾驶系统在不同场景下的测试验证需求。原始数据采集及场景库构建未涉及公共数据和个人数据，无需匿名化、去标识化处理。数据格式兼容51Sim、ModelBase、RoadRunner Scenario、esmini等市面上所有支持OpenDrive和OpenScenario的仿真工具。

This autonomous driving simulation scenario library is suitable for stages such as autonomous driving technology R&D, vehicle testing and safety assessment, and requires high-performance computers and professional personnel for operation. It covers various urban road scenarios, assists in algorithm optimization, functional testing and safety assessment, improves the performance and safety of autonomous driving, and is applicable to R&D enterprises, automobile manufacturers, scientific research institutions and other parties, addressing issues such as algorithm performance improvement, functional verification and safety risk assessment. Based on the road network data of Deqing Intelligent Connected Vehicle Test Site, it is parsed and converted into OpenDrive general format data. Then, in accordance with relevant policy regulations and industry standards for simulation testing, the self-developed scenario library editing tool is used to strictly comply with relevant standards to arrange and create scenarios on the tool's visual interface, and the editing tool then converts them into structured scenario library data in OpenScenario format. Finally, abstract scenarios in policies and regulations are parsed into specific structured data representations composed of poses, actions, events, operations, behaviors, storyboards, entities and other elements. The data content mainly includes entity poses (heading angle, pitch angle, roll angle), entity positions (x, y, z), relevant actions in the scenario (action name field and action type field. The action types include PrivateAction (exclusive actions performed for specific entities such as vehicles and pedestrians), GlobalAction (global actions that take effect on the entire scenario or multiple entities simultaneously), and UserDefinedAction (user-defined actions)), events (event name field and event priority field, which define the trigger conditions and execution logic of actions, including condition-triggered events, time-triggered events, state-triggered events, user-defined events, global events, exclusive events, etc.), operations and operation groups (advanced structures used to define the behaviors of traffic participants, usually containing a series of events and actions), behaviors (defining specific actions and operations performed by traffic participants), storyboards (complete descriptions of the entire scenario, including actions, operations, events, etc.), entities (traffic participants, including vehicles, pedestrians and other targets) and other contents. The test scenarios cover intelligent connected vehicles and low-speed functional unmanned vehicles. It ensures the comprehensiveness, authenticity and repeatability of testing, and meets the test and verification requirements of autonomous driving systems in different scenarios. The original data collection and scenario library construction do not involve public data and personal data, so no anonymization or de-identification processing is required. The data format is compatible with all commercially available simulation tools that support OpenDrive and OpenScenario, such as 51Sim, ModelBase, RoadRunner Scenario, esmini and others.

搜集汇总

背景与挑战

背景概述

德清自动驾驶标准法规仿真测试场景库数据是一个包含539条记录的企业数据集，主要用于自动驾驶技术研发和测试。数据涵盖城市道路场景，包括实体姿态、位置、动作和事件等结构化信息，适用于研发企业、汽车制造商和科研机构。