| 引用本文: | 代嘉惠,许鹏程,李小波.二阶中心差分粒子滤波FastSLAM算法[J].控制理论与应用,2018,35(9):1382~1390.[点击复制] |
| Dai Jia-hui,XU Peng-chen,LI Xiao-bo.Second order central difference particle filter FastSLAM algorithm[J].Control Theory and Technology,2018,35(9):1382~1390.[点击复制] |
|
| 二阶中心差分粒子滤波FastSLAM算法 |
| Second order central difference particle filter FastSLAM algorithm |
| 摘要点击 2891 全文点击 1133 投稿时间：2016-11-11 修订日期：2018-05-05 |
| 查看全文 查看/发表评论 下载PDF阅读器 |
| DOI编号 10.7641/CTA.2018.60849 |
| 2018,35(9):1382-1390 |
| 中文关键词 FastSLAM、即时定位与地图构建、协方差矩阵、粒子滤波、二阶中心差分粒子滤波、移动机器人 |
| 英文关键词 FastSLAM, simultaneous localization and mapping, covariance matrix, particle filter, second order central difference particle filter,mobile robot |
| 基金项目 国家自然科学基金项目(61164015, 61305132), 江西省自然科学基金项目(20151BAB207043)资助. |
|
| 中文摘要 |
| 为改善SLAM算法中非线性系统状态估计精度不高，计算繁杂的问题，本文创新性地提出了基于二阶中心差分滤波并融合最新观测数据来产生建议分布函数的新算法。新算法基于二阶sterling插值公式处理SLAM中的非线性系统问题，无须计算雅可比矩阵，容易实现。此外，该算法使用Cholesky分解技术，在SLAM概率估计中直接依据协方差平方根因子进行传播，保证协方差矩阵正定性的同时减小了局部线性化的截断误差。仿真试验表明，在粒子数相同的情况下，二阶中心差分FastSLAM(SOFastSLAM)在不同噪声条件下的估计精度均优于FastSLAM2.0、UFastSLAM算法，且用时最少，证实了SOFastSLAM算法的优越性。 |
| 英文摘要 |
| For improving the accuracy of nonlinear system state estimation and calculating complex problems in a SLAM algorithm, this paper proposes an innovative method based on a second-order central difference filter and the novel observational data to generate a proposal distribution function. The new algorithm uses the second order sterling interpolation formula to handle the nonlinear system problem in SLAM without calculating the Jacobian matrix and is easily implemented. Furthermore, by using Cholesky decomposition technology to directly propagate the covariance square root factor in SLAM probability estimation, the proposed algorithm not only guarantees the covariance matrix is positive and definite but also reduces the truncation error of local linearization. Notably, as verified by simulation testing with equivalent number of particles, the second-order center differential FastSLAM (SOFastSLAM) showed better estimation accuracy than that of FastSLAM2.0 and UFastSLAM algorithms in different noise conditions with the shortest computation time, confirming the superiority of the SOFastSLAM. |