FlightDiffusion: Revolutionising Autonomous Drone Training with Diffusion Models Generating FPV Video

Valerii Serpiva, Artem Lykov, Faryal Batool, Vladislav Kozlovskiy, Miguel Altamirano Cabrera, Dzmitry Tsetserukou · 2025

Read FlightDiffusion: Revolutionising Autonomous Drone Training with Diffusion Models Generating FPV Video by Valerii Serpiva, Artem Lykov, Faryal Batool, Vladislav Kozlovskiy, Miguel Altamirano Cabrera, Dzmitry Tsetserukou (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2509-14082.txt.

Description

Valerii Serpiva and five co-authors train drones inside generated video in a 2025 robotics paper. FlightDiffusion is a diffusion model framework. It produces realistic first-person drone footage from a single starting frame. It also produces the action sequences that would generate the footage. A drone can learn navigation from synthetic experience instead of costly real flights.

The generated flight paths proved physically plausible and flyable. Position and orientation errors against reference paths were small. Policies trained on this synthetic data planned smoother routes. They stayed robust in simulation and adapted to unseen conditions. Tests found no significant gap between simulated and real flight success rates. That indicates strong sim-to-real transfer. The released datasets give other aerial robotics groups a reusable training resource.

Catalog record

CREATORValerii Serpiva, Artem Lykov, Faryal Batool, Vladislav Kozlovskiy, Miguel Altamirano Cabrera, Dzmitry Tsetserukou
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE32 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2509-14082

Browse the full library · About the Tower of Babel