We developed an algorithm that can keep a fleet of drones in formation around a moving rover using UWB measurements. A prototype was built and tested by mounting UWB on 4 custom-made drones and a rover. A system was also developed that allows a drone to land on a platform (with 4 UWB devices on the vertices) attached to the moving rover, autonomously and with precision <15 cm.
The other activities concerned the use of UWB for robot navigation. In one of our contributions, we exploit the onboard odometer of a robot together with UWB; it accumulates an error that grows over time but whose uncertainty is modeled and used to acquire a new (accurate) UWB measurement reducing the use and therefore the energy consumed, without detriment to accuracy.
In many scenarios, the deployment of a mains-powered infrastructure is not feasible. We proposed and evaluated a system that allows the robot(s) to move in an environment aided by battery-powered anchors; an efficient protocol keeps them as much as possible in standby but can activate them quickly as soon as the rover appears in range.
A local planner trained with deep reinforcement learning was designed to provide a robust and noise-immune guidance system.
We designed and built deployable/retractable anchors, together with algorithms that allow their autonomous repositioning in hostile environments by a rover equipped with a manipulator arm.
Robot Localization via Odometry-assisted Ultra-wideband Ranging with Stochastic Guarantees
V. Magnago, P. Corbalán, G.P. Picco, L. Palopoli, D. Fontanelli. 32nd International Conference on Intelligent Robots and Systems (IROS), Macau (China), November 4-8, 2019. https://doi.org/10.1109/IROS40897.2019.896801
UNITN
Indoor Point-to-Point Navigation with Deep Reinforcement Learning and Ultra-Wideband
E. Sutera, V. Mazzia, F. Salvetti, G. Fantin, M. Chiaberge. International Conference on Agents and Artificial Intelligence, Volume 1, pages 38-47. https://doi.org/10.5220/0010202600380047
POLITO