More efficient algorithms beyond Monte Carlo for uncertain UAV co-design trade-offs

Develop more efficient algorithms than Monte Carlo sampling to compute, compare, and visualize distributions of feasible trade-offs in the task-driven unmanned aerial vehicle co-design problem under distributional uncertainties in battery and actuation parameters when certain design choices must be fixed before parameter realizations.

Background

In the case paper on uncertainty-aware UAV co-design, the authors model uncertainties in battery energy density and actuation power parameters, and assume that actuator selection can occur after observing samples, while battery technology must be chosen beforehand, inducing a stochastic optimization problem.

To approximate the resulting distributions of lifetime cost versus payload trade-offs, the authors use Monte Carlo sampling. They explicitly state that designing more efficient algorithms to handle these computations remains open, highlighting the computational burden of comparing outcome distributions in this uncertain co-design setting.