Adaptively Adjusted EKF-Based Magnet Tracking Method for Fast-Moving Object
Adaptively Adjusted EKF-Based Magnet Tracking Method for Fast-Moving Object
复制标题
DOI：
10.1109/tim.2023.3253892
复制
发表时间：
2023
影响因子：
5.6
通讯作者：
Jiesi Luo;Qiaoyuan Huang;Houde Dai;Shushu Wang;Y. Chen
中科院分区：
文献类型：
--
作者：
Jiesi Luo;Qiaoyuan Huang;Houde Dai;Shushu Wang;Y. Chen
永磁跟踪（PMT）具有无线、无光遮挡、位置和方向同步跟踪等优点，可用于机器人和精密控制。然而，现有的基于优化算法的PMT研究仅限于静止或低速运动，随着磁体速度的增加，跟踪精度会迅速下降。提出了一种基于扩展卡尔曼滤波（EKF）的PMT算法。采用磁偶极子模型逼近圆柱永磁体周围的磁场分布，并采用随机游走模型作为系统的运动模型。根据不同的磁体速度自适应调整EKF估计的状态更新模型，同时通过磁力计阵列输出的统计分析近似初始磁体位姿。因此，与以往的研究相比，该方法可以在保证有效姿态精度的同时，以更高的速度跟踪目标。该方法在技术验证实验和实际应用中得到了验证。此外，还研究了磁体转速与定位精度之间的关系。当磁体速度达到40 mm/s时，实验结果表明，平均位姿误差和算法延迟分别为（1.83±0.25 mm）、（1.296 $\pm ~0.091 $\ circ}）和（0.485±0.064 ms）。在机器人停车充电实验中，当机器人速度达到40 mm/s时，位姿误差为（2.85±0.41 mm, 1.302 $\pm ~0.023^{\circ}$）。该方法可以极大地提高快速运动目标的跟踪性能，从而扩展了PMT的应用场景。
Permanent magnet tracking (PMT) can be employed in robots and precision control due to its advantage of being wireless, without optical occlusion, and simultaneous position and orientation tracking. However, existing PMT studies based on optimization algorithms are limited to stationary or low-speed motions, where the tracking accuracy rapidly decreases with the increasing magnet speed. This study presents an extended Kalman filter (EKF)-based PMT. The magnetic field distribution around the cylindrical permanent magnet is approximated with a magnetic dipole model, where a random walk model is utilized as the system motion model. The state update model of the EKF estimation is adaptively adjusted according to different magnet speeds, while the initial magnet pose is approximated by the statistical analysis of magnetometer array outputs. The proposed PMT method hence can track the object at a higher speed while ensuring an effective pose accuracy than the previous studies. The proposed method was tested in technical verification experiments and a practical application. In addition, the relationship between the magnet speed and the position accuracy is investigated. When the magnet speed reaches 40 mm/s, experimental results show that the average pose error and algorithm latency are (1.83 ± 0.25 mm, 1.296 $\pm ~0.091^{\circ }$ ) and 0.485 ± 0.064 ms, respectively. In the robot parking experiment for charging, the pose errors are (2.85 ± 0.41 mm, 1.302 $\pm ~0.023^{\circ }$ ) when the robot speed reaches 40 mm/s. The proposed method can greatly improve tracking performance for a fast-moving object is greatly improved by performing the proposed method, thereby expanding the application scenarios of PMT.