【摘要】 针对机器人定位过程中传感器感知信息存在野值,加剧粒子退化,导致机器人状态参数滤波值失真,甚至出现定位失败的问题,提出一种机器人自适应抗差无迹粒子滤波定位算法。在重要性采样阶段利用无迹卡尔曼滤波产生优选的建议分布函数,降低系统估计误差,同时有效提升系统的抗噪声能力。同时利用抗差估计原理构造抗差方差分量统计量,并由该统计量引入的自适应因子调节增益矩阵,减弱野值对滤波的影响。实验结果表明,当观测数据中存在野值时,该算法能够有效地控制观测异常误差的影响,定位精度得到了很大提高,并在不同系统噪声和观测噪声方差下,具有较强的鲁棒性和实时性。更多还原

【Abstract】 Aiming at the problem that the sensor perception information contains outliers in the process of robot localization,which aggravates the particle degeneracy and causes the filtering value distortion of the robot state parameters,even leads to localization failure,a robot particle filtering localization algorithm is proposed based on adaptive robust unscented Kalman filtering. In the importance sampling step,the proposed algorithm utilizes unscented Kalman filter to produce a preferred proposal distribution function,reduce the system estimation error and effectively improve the anti-noise ability of the system. The robust estimation principle is used to construct the robust variance component statistics,and the adaptive factor based on the statistics is used to adjust the gain matrix,which decreases the effects of the outliers on filtering. The experiment results show that the proposed algorithm can effectively control the influence of abnormal observation error and greatly improve the localization accuracy when the measurement data contain outliers,and has good robustness and real-time under different system noises and observation noise variances.更多还原