基于多传感器信息融合的轮履混合移动机器人路况识别方法

针对未知环境下轮履混合移动机器人的路况识别问题,提出一种将支持向量机(SVM)与Dempster-Shafer(D-S)证据理论相结合的多传感器信息融合路况识别方法.设计了一个基于超声波传感器和红外传感器的数据采集系统,以提取路况的信息特征;以Platt后验概率为基础,建立了多类SVM的后验概率模型,并构造D-S证据理论所需基本概率分配(BPA)函数;同时,将SVM与D-S相结合的信息融合识别方法应用于轮履混合移动机器人的3种典型路况识别实验.结果表明,所提出的方法能够满足轮履混合移动机器人识别平坦路面、斜坡和台阶等路况的要求.

Road Recognition Method of Wheel Tracked Robot Based on#br# Multi sensor Information Fusion

Abstract： In view of the wheel tracked robot’s road status recognition under the unknown circumstance, a new identification method was proposed based on multi information fusion, and the method compounded with support vector machine (SVM) theory and Dempster Shafer (D S) evidence theory. First, a data acquisition system composed of ultrasonic sensors and infrared sensors was designed in order to get the features of the road condition. Then on the basis of Platt posteriori probability, the multi class posteriori probability model and BPA function were constructed. Finally, experiments with SVM+D S fusion and identification method were carried out to identify the three typical road conditions. The results showed that flat road, slope and step conditions could be identified effectively, hence the method could satisfy the performance requirement of the wheel tracked robot.