Given multiple widespread stationary data sources such as ground-based sensors, an unmanned aircraft can fly over the sensors and gather the data via a wireless link. To minimize delays and system resources, the unmanned aircraft should collect the data at each node via the shortest trajectory. The trajectory planning is hampered by the complex vehicle and communication dynamics. We present a method that allows the ferry to optimize a multi-node data collection trajectory through an unknown radio field using reinforcement learning. The approach learns improved trajectories in situ obviating the need for detailed system identification. The ferry is able to quickly learn significantly improved trajectories compared to alternative heuristics.
展开▼
