ResearchResearch paperRobotics & Embodied AI · Reinforcement Learning · Large Language Models1 source · Oct 18, 2026

AeroDASH: Asynchronous Vision-Language-Action Control for Continuous UAV Flight

We present AeroDASH, an asynchronous VLA framework for continuous language-conditioned flight.

Onboard inference latency creates a mismatch between the trajectories predicted by vision-language-action (VLA) policies and the ongoing motion of unmanned aerial vehicles (UAVs).
A B-spline policy is initialized through supervised fine-tuning on expert demonstrations, providing continuous trajectory references in position and yaw.
Execution experiments show that AeroDASH sustains motion during inference and reduces acceleration, jerk, and speed fluctuations relative to asynchronous tracking of waypoint sequences.
We further demonstrate language-conditioned flight with onboard inference and asynchronous trajectory execution.

[1]AeroDASH: Asynchronous Vision-Language-Action Control for Continuous UAV Flight
arXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 18, 05:05 PM
We present AeroDASH, an asynchronous VLA framework for continuous language-conditioned flight.
Onboard inference latency creates a mismatch between the trajectories predicted by vision-language-action (VLA) policies and the ongoing motion of unmanned aerial vehicles (UAVs).

Extractive summary: sentences quoted from the sources.

Before this