摘要
针对FastSLAM算法对传感器精度要求较高,不适用于方向性差的超声传感器问题,提出了一种基于超声概率栅格地图环境特征点提取匹配的移动机器人粒子滤波同时定位与地图创建(SLAM)算法.该算法可分解为机器人位姿估计和环境路标估计2个部分.基于蒙特卡罗定位原理利用粒子滤波算法对机器人运动轨迹进行估计;在建立全局超声概率栅格地图的基础上,利用概率栅格地图环境特征提取算法对环境路标坐标进行估计.实验证明,该算法较好地解决了超声测距传感器由于散射角大带来的特征点估计不准的问题,对环境路标和机器人轨迹的估计都比较准确.并对移动机器人累计误差进行了有效的补偿,减少了由于累积误差造成的移动机器人轨迹扭曲失真.
Focusing on the limited application of FastSLAM algorithm due to its high accuracy requir- ment for sensor and unfitness for ultrasonic sensor with poor directivity, a simultaneous localization and mapping （SLAM） a/gorithm utilizing particle filter is provided, which is based on ultrasonic probability grid map feature points extraction and matching. This algorithm consists of robot pose esti- mation and environment landmark estimation. Particle filters are applied to estimate the robot trajectory according to Monte Carlo methods. Based on global ultrasonic probabilistic grid map, the environment landmark position is observed by environment feature extraction algorithm. The effectiveness of the proposed algorithm is validated by experimental result. The estimation accuracy of feature point posi- tion as well as environment landmark are improved. The cumulative error of the robot is compensated effectively and the mobile robot trajectory distortion is reduced.
出处
《东南大学学报（自然科学版）》
EI
CAS
CSCD
北大核心
2010年第1期117-122,共6页
Journal of Southeast University：Natural Science Edition
基金
教育部高等学校科技创新工程重大项目培育资金项目(708045)
江苏省"六大高峰人才"资助项目(06-D-031)
东南大学优秀青年教师基金资助项目(4022001004)
