摘要： 异常运动状态识别对移动机器人航迹推算、定位、导航以及机器人安全等至关重要。该文以自行研制的移动机器人为研究对象，分析了4个驱动轮的状态特点(正常、打滑、受阻、卡死)，提取了8个反映4个驱动轮状态的特征，识别了驱动轮运动状态，针对受阻和被卡死两种异常设计了避让策略。实验结果表明，该方法可以有效地识别不同的运动状态，并可以进行有效的避让。

关键词: 移动机器人, 异常监测, 概率神经网络

Abstract: Abnormality detection plays an important role in dead-reckoning, locating, navigating and safety for wheeled mobile robots(WMRs). Based on the mobile robot developed independently by ourselves, this paper analyzes 4 kinds of states (i.e. normal, slipping, blocked, deadly stuck ) for driving wheels, extracts 8 kinds of features, recognizes state of motion, designs two strategies to avoid blocked or deadly stuck situations. Experimental results show that the method can identify different movement states and recover from the abnormality correctly and timely.

Key words: mobile robot, abnormality detection, probabilistic neural network

中图分类号: 

• TP18