摘要
针对视觉机器人定位问题,提出一种改进的蒙特卡罗定位方法.在特征粒子的观测更新上,用机器人当前位置确定局部地图的范围,用局部地图的搜索方式来匹配不惟一地标,减弱了粒子分布的多值性;并在粒子的重采样过程中应用历史观测信息和运动信息产生新粒子点,使新粒子有效的概率加大,加快了结果的收敛速度.该方法应用于TJArk队四腿机器人上,在2008机器人世界杯足球赛上取得了令人满意的结果.
An improved Monte Carlo localization approach was presented to solve the vision-based localization of legged robot.In respond to nature landmarks,dynamic map which moves synchronously with the position of robot instead of the global map was used.The unspecial landmarks captured by camera can be matched in the dynamic map,so the system is more robust against the multi-value caused by the global map′s symmetry.And the method used history information,including vision and odometry sensors′ infomation,to create ...
出处
《华中科技大学学报（自然科学版）》
EI
CAS
CSCD
北大核心
2008年第S1期109-112,133,共5页
Journal of Huazhong University of Science and Technology(Natural Science Edition)
基金
国家自然科学基金资助项目(60875057)
关键词
步行机器人
蒙特卡罗定位
粒子优化
动态地图
legged robot
Monte Carlo localization
particle filters optimize
dynamic map
