Overview

Unmanned aerial vehicle (UAV) design requires the optimal configuration of components and frame geometries to satisfy mission-specific performance requirements. However, conventional drone design processes often rely on repeated design modifications, numerical analysis, and flight tests, leading to substantial computational cost and time consumption.

Although data-driven approaches have recently gained attention as an alternative, the lack of multi-modal design datasets that jointly capture the complex mechanical characteristics and geometric diversity of multicopters has limited the development of AI-based drone design methods.

To address this limitation, this study proposes a large-scale drone design dataset consisting of 1,961 topology-optimized frame geometries generated under diverse component configurations and mechanical performance conditions, namely a TopoDroneX.

Proposed dataset is constructed based on five key design factors, including payloads and propulsion systems, and provides engineering attributes such as thrust, torque, and stress, together with multi-modal data including multi-view 2D projection images, 3D meshes, and point clouds.

We provide benchmark results for representative prediction and generation tasks across different data modalities. By incorporating both geometric information and physics-related design attributes, proposed dataset provides a physically grounded dataset for drone design optimization and broader data-driven engineering design.