Improved Tag-based Indoor Localization of UAVs Using Extended Kalman Filter

Indoor Visual-Based Localization System for Multi-Rotor UAVs

“…Accounting for a more quantitative analysis perspective, Table 2 reports the mean and standard deviation of the error (introduced in the previous section) for the cumulative trials in each hovering phase. We observe that for QR01, the error mean is always included in the range m. For HR01, instead, the range is m. These results are almost consistent with the T1 case, confirming the observations in Section 4.3.1 and highlight the better performance of the proposed VIO approach with respect to the strategy outlined in [ 48 ] where the reported error range is m. Furthermore, focusing on the error along the z -axis of the world frame, i.e., on , for both the UAVs we note that the error mean and its standard deviation remain almost inside limited boundaries, meaning that the performance of the proposed localization method do not downgrade in relation to distance from the map, contrarily to the results given in [ 47 , 48 ]. In detail, for QR01 the mean of is in the range m and the maximum standard deviation is m, while for the HR01 the range of error mean results m and the maximum standard deviation turns out to be m.…”

“…In this case, focusing on planar movements, the experimental tests reveal that the position estimation error increases concurrently with the distance of the UAV from a marker. To mitigate this fact, in [ 48 ] the acquired position data is fused with the IMU measurements using an Extended Kalman Filter (EKF) approach: on the plane, the error range reduces from [−1.0, 0.6] m to [−0.2, 0.6] m, and localization accuracy increases as the UAV gets closer to the tags.…”

“…Inspired by [ 48 ] and similar solutions, we adopt an EKF approach to fusing the position data derived by the marker-based localization method and the inertial measurements acquired by some onboard IMU sensors. We access the performance of such an indoor localization solution through an experimental testing campaign wherein the output of a VICON motion capture system is assumed as ground truth.…”

“…We access the performance of such an indoor localization solution through an experimental testing campaign wherein the output of a VICON motion capture system is assumed as ground truth. In particular, overcoming the results of [ 48 ], we focus on both planar and vertical movements, thus discussing the performance of the outlined strategy as a function of the distance from the tags but also in the case of rapid longitudinal maneuvers between two different points. Moreover, we test the outlined strategy on both a small-medium size custom quadrotor and a medium-large size custom hexarotor, sharing the same flight controller and sensing system.…”

Analysis and Improvements in AprilTag Based State Estimation

“…However, for a large outdoor environment, it is not possible to lay multiple AprilTags on the ground beneath a flying robot all the time for pose correction, so this makes the proposed approach not suitable for a large outdoor environment. In 2019, Kayhani et al [42] have proposed that the raw AprilTag pose is not accurate enough for autonomous operations and has improved the accuracy of an indoor multi-copter by fusing pose data from multiple AprilTags with the help of an Extended Kalman Filter.…”