基于激光扫描和迭代贝叶斯策略的定位

Localization with laser scan and iterative Bayesian strategy

传统贝叶斯数据融合算法中存在线性化缺陷,其在融合测量信息与里程计信息时,不能充分利用测量信息,导致机器人的定位出现较大误差。针对该问题,基于原始激光扫描数据进行特征提取,在得到有效测量信息后,引入迭代贝叶斯数据融合策略,利用非线性最优方式,通过一系列线性化点逐步接近最佳收敛值,达到降低定位误差的目的。实验中,将该特征提取方法应用于Victoria Park数据集,对迭代贝叶斯数据融合算法和传统算法进行性能比较,比较结果表明,该算法的定位轨迹与真实路径的吻合程度更高,相同条件下,对噪声的容忍能力更强。

Due to the linearization problem, the traditional Bayesian data association algorithm can not make full use of the mea- surement information when it fuses the measurements and odometry information, causing a larger error generated in the robot lo- calization. For the above problem, a feature detection method was used to extract effective observations from raw laser scan da- ta, and an iterative Bayesian data fusion strategy was put forward. A nonlinear optimal way was used to produce a series of li- nearization values which gradually approximated to the optimal convergent point. The localization error was reduced. In the experiments, the proposed feature detection method was applied to Victoria Park data set. According to the comparison between the proposed method and the traditional one, the estimated path from the former algorithm fits the ground truth better and its noise tolerance is greater than the latter one in the same conditions.