摘要
针对移动机器人难以单纯依赖自身传感器定位的问题,提出了一种分布式感知协作的扩展Monte Carlo定位方法.在定位过程中,机器人根据感知更新前后采样分布信息熵、有效采样数目及采样分布均匀性的变化,适时地从环境传感器的检测模型进行重采样,从而有效减少其位姿估计的不确定性.在算法的具体实现过程中,采用彩色摄像头作为环境传感器,摄像头的参数由机器人进行在线标定;然后依据标定的参数获得摄像头的检测模型.实验验证了该算法在解决全局定位和机器人绑架问题时的有效性.
In order to overcome the difficulty of a mobile robot to perform localization only with its onboard sensors, an extended Morte Carlo localization algorithm based on collaborative distributed perception is proposed. In the process of localization, the robot timely executes resampling from detection models of environmental sensors according to the changes of sampling distribution information entropy, effective sample size and sampling distribution uniformity before and after the robot＇s perceptive update, and thus the pose estimation uncertainty is reduced effectively. When the algorithm is implemented, color cameras are adopted as environmental sensors and their parameters are calibrated by the robot online. And then detection models of the cameras can be obtained based on the calibrated parameters. Experiment results illustrate the validity of the approach in solving problems of global localization and ＂ kidnapped robot＂. ,
出处
《机器人》
EI
CSCD
北大核心
2008年第3期210-216,共7页
Robot
基金
国家863计划资助项目(2006AA040202
2007AA041703)
