基于模糊逻辑的驾驶员制动意图辨识方法研究

发布时间：2018-06-26 11:31

本文选题：模糊逻辑 + 制动意图辨识　；参考：《吉林大学》2014年硕士论文

【摘要】：为了适应日益加快的社会节奏和满足人类对舒适、安全等性能越来越高的要求，驾驶员制动辅助系统的种类越来越多，人-机控制模式的平滑转换对于这些驾驶员辅助系统是非常重要的。同时线控制动系统取消了传统系统的机械连接，中央控制单元感知驾驶员制动意图再发送相应指令给对应的执行机构。可见，不管是驾驶员制动辅助系统还是线控制动系统，要实现线控制动系统功能以及人-机模式的平滑转换，都需要对驾驶员的制动意图进行辨识。本文通过总结、分析国内外关于驾驶意图辨识方法的现有成果，选择基于模糊逻辑方法建立驾驶员制动意图辨识模型。利用TruckSim和LabVIEW仿真软件搭建商用车驾驶模拟器，并进行典型制动工况的模拟试验获得用于训练模型的试验数据，应用离线训练的方法获得驾驶员制动意图模糊辨识模型。在此基础上，利用搭建的驾驶模拟器对本文提出的驾驶员制动意图辨识模型进行验证。论文主要进行了以下几方面的研究工作： (1).制动意图选取及试验方法设计建立制动意图辨识的模糊模型，首要任务就是确定需要辨识的制动意图，本文主要根据各驾驶员智能辅助系统及线控制动系统的功能，以提高驾驶舒适性和安全性为目标，尽量包含直线制动工况的所有情况选取了四种典型制动意图。四种典型制动意图分别为紧急制动、坡路起步辅助制动、调节制动及持续制动，分析四种制动意图特点制定了不同工况下的试验方法，试验采集的所有特征参数数据可以构成驾驶员制动意图辨识模型训练的数据集。 (2).商用车驾驶模拟器搭建 4种典型工况的模拟试验环境是通过搭建商用车驾驶模拟器实现的，，进而获得用于训练驾驶员制动意图辨识模型的试验数据。驾驶模拟器包括一个主机和一个实时目标机。主机负责车辆参数设置、工况设置、并同时具有显示仿真动画和数据监控的功能；实时目标机负责运行实时车辆模型以及驾驶员操作数据采集和车辆控制程序。 TruckSim和LabVIEW的联合仿真构成了驾驶模拟器的软件基础。LabVIEW与TruckSim程序接口设计包括实时仿真接口设置、求解器设置、车辆输入输出变量定义；LabVIEW主程序设计首先需要建立一个LabVIEW RT项目，并分别对主机程序以及下位机程序进行设计，最后通过共享变量实现主机与下位机的数据通讯。 (3).建立驾驶员制动意图辨识模型驾驶员驾驶行为是一个时序过程，但特定的驾驶环境下驾驶员的操作规律基本保持一致，并反映在某些参数的变化上。因此，本文选取了四个特征参数用于建立驾驶员制动意图辨识的模型。其中，制动踏板开度及其变化率用于辨识驾驶员制动的紧急程度，由期望减速度辨识模型的输出、轮速及其方差作为特征参数被用于多工况的模糊辨识模型输入参数。然后，结合基于模糊逻辑建立隶属度的方法建立用于驾驶员制动意图辨识的辨识模型。 (4).驾驶员制动意图辨识模型试验验证对搭建的驾驶员制动意图辨识模型需要进行离线试验验证和结合驾驶模拟器的在线试验验证。通过试验数据确定的驾驶员制动意图辨识模型参数并结合Matlab和LabVIEW的Fuzzy工具箱搭建了制动意图辨识模型。辨识出的驾驶员制动意图是量化的，本文基于最大相似性原则对辨识数据归类为对应的制动意图。
[Abstract]:In order to adapt to the increasing social rhythm and to meet the increasing requirements of human comfort and safety, the variety of driver's brake auxiliary system is more and more. The smooth conversion of human machine control mode is very important for these driver assistance systems. The central control unit perceiving the driver's braking intention and then sending the corresponding instruction to the corresponding actuator. It is visible, whether it is the driver's brake auxiliary system or the wire brake system, to realize the function of the line control system and the smooth transformation of the man machine mode, it is necessary to identify the driver's braking intention.
By summarizing and analyzing the existing achievements of driving intention identification methods at home and abroad, this paper selects the identification model of driver's braking intention based on fuzzy logic method. Using TruckSim and LabVIEW simulation software to build a commercial vehicle driving simulator, and carry out the simulation test of typical braking conditions to obtain the test data for the training model. The fuzzy identification model of driver's braking intention is obtained by off-line training. On this basis, the driver's braking intention identification model proposed in this paper is verified by using the built driving simulator.
The main research work of the paper is as follows:
(1). The design of the braking intention and the design of the test method
In order to establish a fuzzy model for identification of braking intention, the first task is to determine the braking intention that needs to be identified. This paper is mainly based on the functions of the driver's intelligent auxiliary system and the line control system, in order to improve the driving comfort and safety as the goal, and the four typical braking intentions are selected in all cases including the linear braking condition as far as possible.
The four typical braking intentions are emergency braking, auxiliary braking of slope road, adjusting brake and continuous braking. The analysis of four kinds of braking intent features the test methods under different working conditions. All the characteristic parameters collected by the test can form the data set of the driver's braking intention identification model.
(2). Construction of a commercial vehicle driving simulator
The simulation test environment of 4 typical operating conditions is realized by building a commercial vehicle driving simulator, and then the experimental data used to train the driver's braking intention identification model. The driving simulator includes a host and a real-time target. The host is responsible for the vehicle parameters setting, the working condition setting, and the display simulation animation and the simulation animation. Function of data monitoring; real-time target machine is responsible for running real-time vehicle model and driver operation data acquisition and vehicle control program.
The joint simulation of TruckSim and LabVIEW constitutes the software base of driving simulator.LabVIEW and TruckSim program interface design including real-time simulation interface setting, solver setting, and vehicle input and output variable definition. LabVIEW main program design first needs to establish a LabVIEW RT project, and the host program and the lower computer program respectively. Finally, the data communication between host computer and slave computer is realized through shared variables.
(3). Establish the identification model of driver's braking intention
Driver's driving behavior is a sequential process, but the driver's operation rules are basically consistent and reflect the changes of some parameters in a specific driving environment. Therefore, four characteristic parameters are selected to establish the model of driver's braking intention identification. The emergency degree of the member braking is used for the input parameters of the fuzzy identification model of multiple conditions by the output of the expected speed reduction identification model, the wheel speed and its variance as the characteristic parameters. Then, the method of establishing the membership degree based on the fuzzy logic is used to establish the identification model for the identification of the driver's braking intention.
(4). Test verification of driver's braking intention identification model test
The driver's braking intention identification model needs to be verified by off-line test and combined with the on-line test of driving simulator. The driver's braking intention identification model parameters are identified by the test data and the braking intention identification model is built with the Fuzzy toolbox of Matlab and LabVIEW. The driver's braking intention is identified. Quantified, this paper classifies the identification data into corresponding braking intentions based on the principle of maximum similarity.
【学位授予单位】：吉林大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：U491.25

【参考文献】