车辆主动安全系统关键预测算法研究

发布时间：2018-01-21 11:16

本文关键词： 车辆主动安全预测模型越线时间自适应巡航控制系统目标辨识状态预测 Petri网　出处：《长安大学》2014年博士论文　论文类型：学位论文

【摘要】：现代汽车安全技术的主流趋势已经由被动安全系统转为主动安全系统。车辆主动安全系统能够在交通冲突发生的早期对驾驶员进行提示或者介入车辆的操控，从而避免交通冲突的进一步恶化而引发事故。从保证安全的角度而言，当存在交通冲突风险时，车辆主动安全系统应该尽早的作出辨识。从时间序列而言，如果能够使用车辆的现有行驶状态表征参数对车辆下一步的运动状态进行预测，则可以对“即将到来”的交通冲突进行预先准备，从而进一步提高车辆主动安全系统的生效时间。车辆的行驶状态表征参数纵多，同时车辆行驶交通环境类型也不一样，因此，如何利用现有参数对车辆行驶状态和交通环境进行预测是车辆主动安全系统在算法设计时需要重点考虑的问题。针对车辆主动安全系统对于参数预测的关键技术需求，本文利用多类型传感器搭建了车辆行驶过程中的表征参数同步采集试验平台，实现了车载环境下多车辆行驶车速、交通环境参数的同步采集。利用上述试验平台对10名被试在不同道路环境下开展了真实驾驶试验，获取了大量的车辆行驶状态真实参数。考虑车辆主动安全系统在线运行的真实特点，在对国内外现有技术进行分类总结的情况下，主要完成了以下的研究内容： 1、提出了基于几何分析方法的车辆换道过程中越线时间预测模型。通过使用车辆与车道线距离数据，分析了车辆换道过程中的几何特性。并结合车-路几何模型，提出了车辆换道过程中的车辆偏航角估计理论。针对直道路段和弯道路段，并考虑车辆换道方向与道路弯道方向，分别提出了车辆在直道路段和弯道路段换道过程中的越线时间预测模型。采用真实数据对预测模型的精度进行验证，结果表明，模型整体预测误差较小，且绝大部分的误差分布于零点附件。所进行的检验结果中，直道路段预测误差绝对值小于等于0.1s的比例达到了78.3%，弯道路段相应的比例达到了80.8%，且两种模型的预测误差均符合正态分布规律。 2、通过建立车-路之间的几何关系模型，，并采用车速与横摆角速度对道路曲率进行估计，提出并建立了ACC系统对有效目标、潜在有效目标和无效目标的辨识理论与模型。对辨识模型分别进行了单目标追踪、多目标追踪以及多目标状态切换追踪的检验，结果表明，本文所建立的模型能够有效的区分三类目标。在此基础上，利用模糊加权评价方法，采用目标车的速度、目标车跟车时距、目标车横向运动状态等参数建立了前方车辆状态切换的预测模型。采用真实试验数据对预测模型进行检验，结果表明，该模型对目标车不同状态的切换预测准确率均超过了90%。 3、针对车辆运行过程中对于自车运动状态参数的预测需求，以线性二自由度车辆模型为研究对象，采用模糊Petri理论建立了车辆运行轨迹模型，将车身横向、纵向加速度、俯仰以及侧倾角速度作为输入变量建立了车辆运行状态预测模型，分别实现了对自车运行速度、横摆角速度、运行轨迹等参数的预测。针对单纯BP神经网络模型在对车辆运动状态预测过程中存在的不足，本文提出采用贝叶斯滤波器对BP神经网络模型的结果进行优化，检验结果表明，该方法将预测准确率由83.6%提高到了92.4%。本研究得到了国家自然科学基金项目（51178053和61374196）和教育部长江学者和创新团队发展计划项目（IRT1286）的资助。
[Abstract]:The main trend of modern automotive safety technology has been from the passive safety system to active safety system. The vehicle active safety system can prompt intervention or manipulation of the vehicle driver in the early stage of traffic conflict, so as to avoid further deterioration of the traffic conflict caused the accident. From the security point of view, when there is traffic conflict risk the vehicle active safety system, should as soon as possible to make identification. From the time sequence, if the motion state of the existing driving state parameters can be used for vehicle vehicle next prediction, can the traffic conflict coming "were prepared in advance, so as to further improve vehicle active safety system. The effect of time parameter the running state of the vehicle longitudinal, and vehicle traffic environment types are not the same, therefore, how to use the existing parameters of The vehicle driving state and traffic environment prediction are the key problems to be considered when the vehicle active safety system is designed in the algorithm design.
For the vehicle active safety system for the demand of key technical parameters prediction, this paper by multi-sensor synchronous acquisition test platform parameters built during the running of the vehicle, the vehicle speed vehicle under multi environment, synchronous acquisition traffic environment parameters. On 10 subjects of the real driving test carried out in different road environment the use of the test platform, to obtain a large number of real vehicle state parameters. Considering the real characteristics of online vehicle active safety system, summarized the situation of existing technology at home and abroad, the main research contents of the following:
1, put forward the geometric analysis method of lane changing process and time prediction model based on vehicle and lane. By using distance data, analysis of the geometric characteristics of road vehicles to change process. Combined with the vehicle road geometry model is proposed for vehicle road vehicle yaw angle estimation process in straight theory. And curved sections, lane changing direction and the curve of the road direction and consider, respectively put forward more time line vehicles in straight and curved sections lane change process model. To verify the accuracy of the real data of the prediction model. The results show that the model of overall prediction error is small, and most of the error distribution in the zero attachment. The test results, the straight section of the prediction error absolute value is less than or equal to 0.1s ratio reached 78.3%, the proportion of the corresponding curve sections reached 80.8%, and the two kinds of model pre The measurement error is in accordance with the normal distribution.
2, through the establishment of vehicle geometry model between the road, and the speed and yaw rate of the road curvature estimation, put forward and established a ACC system for effective target, identification theory and model of potential targets and effective target. The invalid identification model were investigated by single target tracking, multi-target test tracking and multi-target tracking state switching. The results show that the model can effectively distinguish the three kinds of targets. On this basis, using weighted fuzzy evaluation method, the target vehicle speed, the car with the car away from the target, the target vehicle lateral motion parameter prediction model is set up in front of the vehicle state switch. By testing, the prediction model was the real test data. The results show that the handoff prediction model on the target vehicle in different states accurate rate of over 90%.
3, according to the running process of the vehicle for demand forecasting vehicle motion state parameters, the linear two degrees of freedom vehicle model as the research object, using fuzzy Petri theory to establish the vehicle trajectory model, the transverse, longitudinal acceleration, pitch and roll rate as input variables to establish the prediction model of vehicle running state, respectively. The realization of vehicle speed, yaw rate, prediction trajectory and other parameters. For the simple BP neural network model in the vehicle motion state prediction process problems, this paper adopts Bei Juliu filter to the BP neural network model optimization results, test results show that this method will predict accuracy increased from 83.6% to 92.4%.
The study was funded by the National Natural Science Foundation (51178053 and 61374196) and the Ministry of education, the Yangtze River scholar and the innovative team development program (IRT1286).

【学位授予单位】：长安大学
【学位级别】：博士
【学位授予年份】：2014
【分类号】：U461.91

【参考文献】