What a Drag! Streamlining the UAV Design Process with Design Grammars and Drag Surrogates

Unmanned Aerial Vehicles (UAVs) continue to pro-liferate, revolutionizing tasks such as cargo transport, surveillance, and search and rescue operations. With the discovery of novel use cases or specialized tasks for aerial vehicles, there is an increased need for improved design space exploration and performance estimation techniques for candidate UAV designs. Typical pipelines for this design process rely on time-consuming human efforts to identify productive design geometries or ex-pensive computational approaches for performance analysis to reconcile aerodynamic, electrical, and physical interactions. In this work-in-progress paper, we propose the use of a design process that uses a design grammar for UAV design generation and a Graph Neural Network (GNN)-based drag surrogate trained on simulation data for accelerated UAV design space exploration. We formulate a UAV design grammar and provide preliminary performance results from the GNN drag surrogate for randomly generated designs. We expect our approach to accelerate the exploration of UAV design geometries using a learned surrogate drag model to circumvent resource-hungry Computer-aided design (CAD) and simulation routines.