Overview
Applied ML-driven optimization techniques to autonomous vehicle component design — specifically propellers for unmanned underwater vehicles (UUVs). This work contributes to the DARPA Assured Autonomy program’s underwater vehicle challenge problem by enabling efficient, automated propeller design.
Approach
The system integrates computational fluid dynamics (CFD) simulation with machine learning models to predict propeller performance metrics, replacing expensive full CFD evaluations with learned surrogates that maintain accuracy while dramatically reducing computation time.