基于经典DSmT的Sonar栅格地图创建  

Sonar Grid Map Building under Static Environment Based on Classic DSmT

作　　者：李新德[1] 黄心汉[1] 王敏[1] 

机构地区：[1]华中科技大学控制系智能控制与机器人实验室,湖北武汉430074

出　　处：《计算机应用研究》2007年第3期209-212,共4页Application Research of Computers

基　　金：国家自然科学基金资助项目(69585003)

摘　　要：基于最近提出的一种在贝叶斯和DST扩展而来的信息融合算法DSmT(Dezert-Smarandache Theory),在实验的基础上,结合Sonar测量的基本特性,对静态结构化环境建模,并构造了广义基本信度赋值函数,利用经典DSm融合规则,融合每个栅格的声纳冗余信息,计算栅格占用的Bel。最后,以Pioneer II移动机器人作为试验平台,并在线对小型环境进行了3D栅格占用信度分布地图创建,其俯视图与实际2D地图中的物体外观轮廓及所在位置进行比较,其比较结果充分验证了算法的有效性,为进一步研究应用基于折扣理论的DSmT解决异类或同类非可靠多源信息融合,基于Hybrid DSmT的动态环境地图创建,以及多机器人联合创建地图和自定位奠定了坚实的基础。A new method of information fusion recently developed from Bayesian theory and Dempster-Shafer theory was applied and on the base of experiment, according to sonar＇ s metrical characteristic, the general basic belief assignment functions were constructed for modeling of static environment. After sonar＇ s redundant information for each grid was fused according to classic DSm rule, the Bel for grid＇ s occupancy was calculated. Finally, Pioneer Ⅱ mobile robot served as the experiment platform, and 3D Map was built based on DSmT on-line respectively. By comparing its planform from objects＇ profile and location in the real 2D environment, the validity of this method was fully verified. In short, would establish a firm foundation for further research on solving heterogeneous or homogeneous multi-source information fusion according to DSmT based on discount theory, on map building in dynamic environment and multi-robot＇ s map building together and self-localization based on Hybrid DSmT.

关 键 词：不确定性 信息融合算法 栅格地图 信息融合 移动机器人 

分 类 号：TP24[自动化与计算机技术—检测技术与自动化装置]