The present work investigates the application of Artificial Intelligence-based optimization techniques
to the aerodynamic design of small-scale propellers operating at low Reynolds numbers. A
Single-Step Deep Reinforcement Learning (SDRL) algorithm is implemented to optimize propeller
geometry in terms of chord and twist distributions and airfoil shape. The performance of the
generated geometries is assessed using Blade Element Momentum Theory (BEMT), coupled with
NeuralFoil for fast and reliable aerodynamic coefficients evaluation. In the two optimization runs,
the rewards identifying the quality metric are set equal to thrust and efficiency, respectively. The
results prove that the intelligent agent is capable of autonomously identifying valid design solutions
in a complex design space, without prior aerodynamic knowledge.