摘 要

非机动车在国内尤其是城市中由于其方便快捷性，是居民常使用的一种交通工具。随着非机动车尤其是电动自行车的数量增多，由非机动车辆而引发的交通事故逐年上升。自动驾驶作为当今汽车行业的发展热门方向，要实现上路安全驾驶，则必然涉及到与非机动车相关的识别测试。

本文从非机动车危险工况的发生机理分析切入，对非机动车常见的几种类型交通事故进行特性、成因及机理分析，为自动驾驶非机动车识别测试场景设计提供理论依据。然后提出七种自动驾驶非机动车识别测试场景及测试场景要求，最终借助3D建模软件3Dmax和游戏引擎软件Unity3D，对测试场景进行了模拟，以提供实际测试的参考实例，所得结果对于自动驾驶技术关于非机动车的识别测试场景设计具有指导意义。

论文主要研究了非机动车与机动车在自然驾驶和交通事故两种危险工况的特性、成因，并根据特性及成因分析了四种常见非机动车-机动车交通事故的发生机理。在发生机理和危险工况特性、成因的基础之上，首先确定了自动驾驶非机动车识别测试场景的测试参数，然后提出七种包括自然驾驶和交通事故两种类型危险工况下的测试场景。

研究结果表明：自动驾驶非机动车识别测试场景的设计不能只从交通事故方向上进行设计，同时自动驾驶汽车也应该能对自然行驶状态下的非机动车正确识别，如并列行驶、跟车和会车等场景。最终研究设计出的七种测试场景，包含有自然驾驶工况测试场景三种，交通事故类测试场景四种。各场景设置有特定的测试参数、测试情景以及统一的测试过程。最后还通过软件模拟场景，为测试场景提供可视化示例。

本文的特色：本文在对非机动车交通事故的分析之上，对四种常见非机动车交通事故）的发生机理进行分析；根据非机动车事故发生机理及自然驾驶状体非机动车特性，并结合相关交通相关法律法规，确定测试场景参数和测试场景。本文运用了unity3D软件对场景模拟进行了类游戏化处理，场景模拟仿真的过程类似于控制车辆的简易赛车游戏。

关键词：自动驾驶；机动车；非机动车；测试场景；计算机模拟

Abstract

Non-motor vehicles in China, especially in the city because of its convenience and speed, is often used by residents of a means of transport. With the increasing number of non-motor vehicles, especially electric bicycles, the traffic accidents caused by non-motor vehicles increase year by year. As a hot development direction of today's automobile industry, autonomous driving must involve the identification test related to non-motor vehicles in order to realize safe driving on the road.

Starting from the analysis of the occurrence mechanism of non-motor vehicle dangerous working conditions, this paper analyzes the characteristics, causes and mechanism of several common types of non-motor vehicle traffic accidents, so as to provide theoretical basis for the design of automatic driving non-motor vehicle identification test scenario. And then puts forward seven automated driving a non-motor vehicle identification test scenarios and the requirement of test scenarios, finally with the help of 3 d modeling software production software 3dmax and Unity3D game engine software, has carried on the simulation to test scenarios, in order to provide the actual test reference for instance, the results for automated driving technology of non-motor vehicles recognition test scenario design has a guiding significance.

This paper mainly studies the characteristics and causes of non-motor vehicles and motor vehicles in two dangerous working conditions of natural driving and traffic accidents, and analyzes the occurrence mechanism of four common non-motor vehicles and motor vehicle traffic accidents according to the characteristics and causes. On the basis of occurrence mechanism, characteristics and causes of dangerous working conditions, the test parameters of the automatic driving non-motor vehicle identification test scenario were firstly determined, and then seven test scenarios including natural driving and traffic accident under two types of dangerous working conditions were proposed.

The research results show that the design of the self-driving non-motor vehicle identification test scene can not only be designed from the direction of traffic accidents, and the self-driving car should also be able to correctly identify the non-motor vehicles under natural driving conditions, such as parallel driving, following and Meetings such as cars. The final seven research scenarios designed include three types of natural driving conditions test scenarios and four types of traffic accident test scenarios. Each scenario is set with specific test parameters, test scenarios, and a unified test process. Finally, a simulation example is provided through the software to provide a visual example of the test scenario.

Features of this paper: Based on the analysis of non-motor vehicle traffic accidents, this paper analyzes the occurrence mechanism of four common non-motor vehicle traffic accidents; According to the occurrence mechanism of non-motor vehicle accidents and the characteristics of non-motor vehicle in the natural driving state, and in combination with relevant traffic laws and regulations, determine the parameters of the test scene and the test scene. In this paper, unity3D software was used to gamify scene simulation, and the process of scene simulation was similar to the simple racing game of controlling vehicles.

Key Words：autonomous driving; motor vehicle; non-motor vehicle; test scenario; computer simulation

目 录

摘 要 I

Abstract II

第1章 绪论 3

1.1研究目的 3

1.2研究意义 3

1.3非机动车交通研究现状 4

1.4自动驾驶测试研究现状 4

1.5国内外汽车-非机动车常见碰撞场景类型研究现状 5

1.5.1国内研究现状 5

1.5.2国外研究现状 5

1.6国内外道路交通事故发生机理研究现状 6

第2章 非机动车危险工况特性分析 7

2.1非机动车交通事故特性 7

2.1.1事故数量及事故严重度特性 7

2.1.2事故时间及空间分布特性 8

2.1.3非机动车交通事故形态特性及其危险工况 10

2.2非机动车交通事故重要成因归纳 10

2.2.1人的因素 11

2.2.2 车辆因素 11

2.2.3道路因素 11

2.2.4环境因素 11

2.3 非机动车-机动车自然驾驶危险工况特性分析 12

第3章 非机动车危险工况发生机理 13

3.1非机动车交通事故发生机理理论框架构建 13

3.1.1非机动车交通事故发生机理含义 13

3.1.2非机动车交通事故成因和发生机理的区别及联系 14

3.1.3非机动车交通事故发生机理的分类 14

3.2非机动车交通事故发生机理分析方法 14

3.3非机动车典型交通事故发生机理 15

3.3.1非机动车入侵车道交通事故发生机理分析 15

3.3.2非机动车逆行 16

3.3.3非机动车横穿道路 17

3.3.4 十字交叉路口事故 18

第4章 自动驾驶非机动车识别测试场景设计 20

4.1测试参数确定 20

4.1.1道路特征 20

4.1.2测试角色 20

4.1.3测试机动车与非机动车速度 21

4.1.4环境条件 21

4.1.5道路设施 21

4.2测试场景提出 22

4.2.1自然驾驶工况场景 22

4.2.2交通事故类工况场景 24

4.3数据收集及测试有效性 27

4.3.1 数据收集 27

4.3.2 测试有效性 27

第5章 测试场景计算机软件模拟 29

5.1场景模拟的基本原理与过程 29

5.2 3Dmax与Unity3D建模和场景搭建过程 29

5.2.1建模软件的选择 29

5.2.2 3Dmax建模 30

5.2.3 Unity3D场景搭建 31

5.2.4机动车与非机动车运动程序设计 32