基于激光雷达特征提取和增量式数据关联的障碍分类判决

针对移动机器人导航中障碍检测与辨识问题,结合自行研制的移动机器人MORCS-1实验平台,在对二维激光雷达获得的障碍测距数据分析的基础之上,提出将无监督聚类学习应用于障碍的特征提取,基于有效性索引函数和自调整学习机制解决未知特征个数的类别划分,并将模糊逻辑引入到增量式测距数据关联对障碍的动静态属性进行分类判决.在研究所的办公室环境进行了特征提取和障碍分类实验,实验结果验证了该方法的有效性.

Obstacles classification decision based on feature extraction and incremental data association of laser scanners

Aimed at the detection and identification of obstacles in mobile robot navigation, combined with the experimental platform MORCS-1 designed by the authors, an unsupervised clustering algorithm was presented to realize the feature extraction of obstacles based on the analysis of ranging data obtained from 2D laser scanners. Considering the unknown clustering number in advance, the validation index function was introduced into the self-learning mechanism to determine the accurate clustering number automatically. At the same time, fuzzy logic was intergrated into incremental data association of obstalce features to make the static or dynamic obstalces classification decision to reduce the uncertain influence. The experiment of feature extraction and incremental obstalces classification in the office as the operating environment was implemented, and the results verified the effectiveness of the approach.