2017/07/18 by Erwan Lecarpentier, Sebastian Rapp, Lecarpentier, Erwan +5
Engineering · Earth and Planetary Sciences · #Aerospace and Aviation Technology #Meteorological Phenomena and Simulations #Air Traffic Management and Optimization
paper · pdf · doi:10.48550/arxiv.1707.05668
Autonomous unpowered flight is a challenge for control and guidance systems:\nall the energy the aircraft might use during flight has to be harvested\ndirectly from the atmosphere. We investigate the design of an algorithm that\noptimizes the closed-loop control of a glider's bank and sideslip angles, while\nflying in the lower convective layer of the atmosphere in order to increase its\nmission endurance. Using a Reinforcement Learning approach, we demonstrate the\npossibility for real-time adaptation of the glider's behaviour to the\ntime-varying and noisy conditions associated with thermal soaring flight. Our\napproach is online, data-based and model-free, hence avoids the pitfalls of\naerological and aircraft modelling and allow us to deal with uncertainties and\nnon-stationarity. Additionally, we put a particular emphasis on keeping low\ncomputational requirements in order to make on-board execution feasible. This\narticle presents the stochastic, time-dependent aerological model used for\nsimulation, together with a standard aircraft model. Then we introduce an\nadaptation of a Q-learning algorithm and demonstrate its ability to control the\naircraft and improve its endurance by exploiting updrafts in non-stationary\nscenarios.\n