Automatic landing is a feature that allows aerial robotic platforms to safely and accurately land without human intervention. This paper presents a plug-and-play tiny machine learning vision-based system for automatic landing compatible with the Pixhawk flight controller series. The proposed system is implemented on a low-power microcontroller, specifically OpenMV Cam H7 Plus, demonstrating that a constrained resources board can be used as a companion computer to enable autonomous functions for UAVs. The experiments confirm the proposed system's effectiveness, capable of correctly identifying a landing pad and consequently controlling the UAV to align it over the pad center before landing. The system overhead is only 2% of the UAV's total energy budget, with an accuracy of 93.5% and precision of 94.0%.
Automatic landing is a feature that allows aerial robotic platforms to safely and accurately land without human intervention. This paper presents a plug-and-play tiny machine learning vision-based system for automatic landing compatible with the Pixhawk flight controller series. The proposed system is implemented on a low-power microcontroller, specifically OpenMV Cam H7 Plus, demonstrating that a constrained resources board can be used as a companion computer to enable autonomous functions for UAVs. The experiments confirm the proposed system's effectiveness, capable of correctly identifying a landing pad and consequently controlling the UAV to align it over the pad center before landing. The system overhead is only $2\%$ of the UAV's total energy budget, with an accuracy of $93.5\%$ and precision of $94.0\%$.
Santoro, L., Albanese, A., Canova, M., Rossa, M., Fontanelli, D., Brunelli, D. (2023). A Plug-and-Play TinyML-based Vision System for Drone Automatic Landing. Piscataway, New Jersey : IEEE [10.1109/metroind4.0iot57462.2023.10180179].
A Plug-and-Play TinyML-based Vision System for Drone Automatic Landing
Brunelli, Davide
Supervision
2023
Abstract
Automatic landing is a feature that allows aerial robotic platforms to safely and accurately land without human intervention. This paper presents a plug-and-play tiny machine learning vision-based system for automatic landing compatible with the Pixhawk flight controller series. The proposed system is implemented on a low-power microcontroller, specifically OpenMV Cam H7 Plus, demonstrating that a constrained resources board can be used as a companion computer to enable autonomous functions for UAVs. The experiments confirm the proposed system's effectiveness, capable of correctly identifying a landing pad and consequently controlling the UAV to align it over the pad center before landing. The system overhead is only 2% of the UAV's total energy budget, with an accuracy of 93.5% and precision of 94.0%.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
