ImagineUAV: Aerial Vision-Language Navigation via World-Action Modeling and Kinodynamic Planning
This work addresses the problem of grounding free-form instructions into 6-DoF flight for UAVs under partial observability, offering a practical imagination-driven framework.
ImagineUAV tackles aerial vision-language navigation by using a latent video diffusion model to generate future observations and infer 6-DoF motions, achieving superior performance over prior baselines on benchmarks and real-world flights with only 1.3B parameters.
Vision-language navigation (VLN) for UAVs demands grounding free-form instructions into 6-DoF flight under partial observability. While Vision-Language-Action (VLA) models excel at semantic reasoning, they suffer from brittleness due to geometric inconsistency and dynamics mismatch. To address this, we propose ImagineUAV, an imagination-driven framework leveraging cascaded world-action modeling. Instead of direct regression, ImagineUAV employs a latent video diffusion model to generate instruction-conditioned future observations, explicitly imagining environmental evolution, from which 6-DoF motions are inferred via an action extractor. A kinodynamic planner then refines these estimates into collision-free trajectories. Additionally, a step-distilled inference pipeline ensures real-time execution. With only 1.3B parameters, ImagineUAV outperforms prior VLN and VLA baselines on benchmarks and real-world flights, validating the practicality of imagination-driven aerial navigation.