| Full text | |
| Author(s): |
Caldas, Kenny A. Q.
;
Inoue, Roberto S.
;
Terra, Marco H.
;
IEEE
Total Authors: 4
|
| Document type: | Journal article |
| Source: | 2022 IEEE SENSORS; v. N/A, p. 4-pg., 2022-01-01. |
| Abstract | |
We propose in this work a non-parametric Gaussian Process Regression (GPR) model to estimate reduced drift velocity of a commercial Micro Air Vehicle (MAV) by using a single additional Inertial Measurement Unit and odometry data, without relying on expensive or processing demanding sensors. The main advantage of this approach is that it does not require sensor calibration and parameter tuning for accurate velocity estimation. We considered three different kernels in the GPR training and evaluated the proposed method on two datasets. The results are compared with the MAV odometry and a Visual Simultaneous Localization and Mapping (SLAM) algorithm against ground truth data obtained from a high precision motion tracking system. Our proposed method demonstrates superior performance in both datasets and provides a viable solution to reduce position drift on commercial MAVs. (AU) | |
| FAPESP's process: | 14/50851-0 - INCT 2014: National Institute of Science and Technology for Cooperative Autonomous Systems Applied in Security and Environment |
| Grantee: | Marco Henrique Terra |
| Support Opportunities: | Research Projects - Thematic Grants |
| FAPESP's process: | 18/13848-1 - Trajectory planning of autonomous heterogeneous robots for cooperative 3D mapping of an unknown environment |
| Grantee: | Kenny Anderson Queiroz Caldas |
| Support Opportunities: | Scholarships in Brazil - Doctorate |