Kite-based airborne wind energy on Mars: A hybrid physics-analytical and machine learning feasibility framework with uncertainty quantification

Autonomous, dust-resilient power generation is a critical enabling requirement for sustained robotic and crewed surface operations on Mars. Solar photovoltaic systems suffer progressive dust-induced degradation, while radioisotope thermoelectric generators face severe scalability and fuel-availability constraints. This study presents the first integrated feasibility assessment of kite-based airborne wind energy for the Martian surface, using a two-level hybrid modelling framework. The analytical tier, parameterised with Mars Climate Database wind data for the Arsia Mons North site, establishes seasonal power bounds using the Luchsinger pumping-cycle formulation. The data-driven tier employs an Extreme Gradient Boosting surrogate trained on 7021 samples from the open-access Meridional flight-test dataset and validated via eight-fold leave-one-cycle-out cross-validation. A corrected physics-informed scaling procedure - incorporating kite area normalisation, Reynolds-consistent apparent wind iteration, and removal of a previously identified Reynolds number double-counting error - transfers the Earth-validated surrogate to Martian conditions without retraining. For a 60 m 2 kite at Arsia Mons North operating at 100 m altitude, the corrected machine learning estimate yields a mean net cycle power of 663 W and an annual energy production of approximately 1742 kWh/year under a 30% duty cycle. Sensitivity analysis over eight parameter perturbations shows the annual yield ranges from 686 to 3770 kWh/year, with the Reynolds power-law exponent and free-stream wind speed identified as the dominant uncertainty drivers. The results establish both theoretical bounds and realistic performance expectations, and identify altitude selection and seasonal timing as the primary mission design levers. • First hybrid physics - machine learning feasibility framework for kite wind energy on Mars. • Eight-fold cross-validation confirms Earth surrogate stability across all pumping cycles. • Corrected scaling removes Reynolds double-count; mean net cycle power revised to 663 W. • Sensitivity analysis spans 686-3,770 kWh/yr; wind speed is the dominant driver. • Annual 1742 kWh enables hybrid photovoltaic - kite - battery surface microgrids.