基于传感器与BP神经网络的无人车路况识别与分类研究  

Research on Unmanned Vehicle Road Condition Recognition and Classification Based on Sensors and BP Neural Networks

作　　者：敖翔 王黎明[1] 赵永辉 栾朝琨 马志扬 AO Xiang;WANG Liming;ZHAO Yonghui;LUAN Chaokun;MA Zhiyang(School of Electrical Engineering,Naval University of Engineering,Wuhan 430034)

机构地区：[1]海军工程大学电气工程学院,武汉430034

出　　处：《舰船电子工程》2024年第9期42-47,共6页Ship Electronic Engineering

基　　金：国家优秀青年科学基金项目“海洋大地测量(Marine Geodesy)”(编号:42122025);国家自然科学基金项目“基于仿生偏振光的UCSV自主精确导航方法研究”(编号:2022CFB865)资助。

摘　　要：为保证无人车在复杂的非结构化道路环境下稳定地行驶,需要对无人车所行驶的实时路况进行准确识别。为满足这一要求,论文详细介绍了一种新的方法:首先使用车载陀螺仪传感器和电机编码器来实时捕获无人车的运动数据,这些数据包括但不限于车辆的速度、角度等关键指标。然后将原始数据进行特征化处理,从中提取出对路况识别有帮助的关键特征,最后利用BP神经网络对提取出的特征进行分析和学习,以识别出无人车当前所行驶的道路类型。实验阶段借助装备有ROS操作系统的无人车和Matlab进行验证。实验结果表明,该研究提出的这种道路识别方法有效且准确,为后续的稳定性控制提供了较好的支撑。To ensure the stable operation of unmanned vehicles in complex unstructured road environments,it is necessary to accurately identify the real-time road conditions in which the unmanned vehicles are driving.To meet this requirement,this article provides a detailed introduction to a new method.Firstly,the onboard gyroscope sensor and motor encoder are used to capture real-time motion data of the unmanned vehicle,including but not limited to key indicators such as vehicle speed and angle.Then,the raw data is subjected to feature processing to extract key features that are helpful for road condition recognition.Then,the raw data is subjected to feature processing to extract key features that are helpful for road condition recognition.During the experimental phase,validation is conducted using unmanned vehicles equipped with ROS operating systems and Matlab.The experimental results show that the road recognition method proposed in this study is effective and accurate,providing good support for subsequent stability control.

关 键 词：非结构化道路 BP神经网络 车载传感器 识别 

分 类 号：TP242[自动化与计算机技术—检测技术与自动化装置]