摘要
自主移动机器人在未知环境中探索和估计路标的方法主要基于SLAM技术。提出一种以全局定位误差最小化为指导的基于SLAM的探索策略。以全局定位误差的估计为准则,采用Monte Carlo采样来贪心地优化每一步的行走路径。考虑到SLAM估计的惯性,文中对较大转弯角度进行惩罚,使机器人更倾向于平滑的行走轨迹,从而进一步提高路标位置的估计精度。文中还将全局定位信息引入SLAM的机器人自身位置估计中,由于全局定位信息历史运动轨迹,该方法能够有效地校正当机器人移动变化过大时SLAM估计的误差。实验显示了文中方法的有效性。
Exploration and estimation of landmarks in an unknown environment is an important skill for autonomous robots,most of which are based on the SLAM technique. This paper presents an SLAM based exploration strategy to minimize the global localization error,via greedily optimizing every step by Monte Carlo sampling. Due to the inertia of the SLAM method,we then penalize a large change of direction for a smoother trajectory,which may result in a more accurate estimation of landmarks. To further calibrate the estimation error for a large range of movement,the global localization information is introduced to the procedure of the estimation of the robot,since it depends less on the historical movement trajectory. Experiments verified the effectiveness of the proposed method.
出处
《智能系统学报》
CSCD
北大核心
2014年第3期313-318,共6页
CAAI Transactions on Intelligent Systems
基金
国家自然科学基金面上项目资助项目(61375061)
江苏省自然科学基金青年基金资助项目(BK2012303)