Tham khảo tài liệu 'field and service robotics - corke p. and sukkarieh s.(eds) part 7', kỹ thuật - công nghệ, cơ khí - chế tạo máy phục vụ nhu cầu học tập, nghiên cứu và làm việc hiệu quả | Further Results with Localization and Mapping Using Range from Radio 235 Results In order to evaluate the performance of the filter we turn to the two commonly used error metrics the Cross-Track Error XTE and the Along-Track Error ATE . The XTE accounts for the position error that is orthogonal to the robot s path . orthogonal to the true robot s orientation while the ATE accounts for the tangential component of the position error. As part of our error analysis of the path estimates we observe the average of the absolute values of the XTE and ATE for each point in the path as well as the maximum and standard deviation of these errors. In the experiment illustrated here the true initial robot position from GPS was used as the initial estimate. Furthermore the location of each tag was known. Figure 1 shows the estimated path using the Kalman filter along with the GPS ground truth with 2 cm accuracy for comparison. Path Estimatefrom Kalman Filter Localization y position m Fig. 1. The path estimate from localization red ground truth blue and beacon locations are shown. The filter uses odometry and a gyro with range measurements from the RF beacons to localize itself. The path begins at 0 0 and ends at 33 0 travelling a total of km and completing 11 identical loops with the final loop km shown above. Note the axes are flipped . Numerical results are given in Table 1. Table 1. Cross-Track and Along-Track Errors for Kalman filter Localization estimate for the entire data set using the Kalman Filter with gyro bias compensation. XTE ATE Mean Abs. m m Max. m m Std. m m 236 J. Djugash S. Singh and P. Corke Failure Sensor Silence. An issue that requires attention while dealing with the Kalman filter is that of extensive sensor silence. When the system encounters a long period during which no range measurements are received from the beacons it becomes heavily dependant on the odometry and its estimate diverges. Upon .
