基于姿态测量模块和闭环检测算法的仿生SLAM研究

发布时间：2018-10-12 11:42

【摘要】：基于传感器模型的SLAM导航策略精度较高,但由于摩擦等因素误差长时间累计会造成里程计的漂移现象。依靠视觉里程计进行导航的Rat SLAM系统,通过在局部场景细胞中引入闭环检测策略实现累计误差的校正,在静态环境下具有一定的鲁棒性,但在复杂场景里,如移动障碍物的出现,视觉里程计会提取到错误的速度等姿态信息,导致航迹出现较大偏移,有时无法通过场景重定位进行闭环校正。结合两种模型的优势提出一种带姿态测量模块和闭环检测算法的仿生SLAM模型。实验表明,相较于仅带带闭环检测的Rat SLAM系统或仅由姿态测量模块构成的导航系统,本文提出的新系统更能适应长期复杂场景下的导航,且鲁棒性更强。
[Abstract]:The precision of SLAM navigation strategy based on sensor model is high, but the drift of mileage will be caused by the accumulated error of friction and other factors for a long time. The Rat SLAM system which is guided by vision odometer can correct the cumulative error by introducing closed-loop detection strategy into the cells of local scene. It is robust in static environment, but in complex scene. For example the appearance of moving obstacle the visual mileage meter will extract the wrong attitude information such as velocity and so on which leads to a large deviation of the track and sometimes can not be corrected by the scene re-location closed loop. Combining the advantages of the two models, a bionic SLAM model with attitude measurement module and closed loop detection algorithm is proposed. The experimental results show that the proposed new system is more robust than the Rat SLAM system with closed loop detection or the navigation system composed of attitude measurement modules.
【作者单位】：安徽工程大学安徽省电气传动与控制重点实验室;
【基金】：2016年安徽高校自然科学研究项目(KJ2016A794) 2016安徽工程大学研究生实践与创新基金项目(Y040116004)
【分类号】：TP212;TP242

【参考文献】