Reproduce “AeroDASH: Asynchronous Vision-Language-Action Control for Continuous UAV Flight” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Zhe Liu, Yuxuan Zheng, Jinzhe Qin, Ziqi Gao, Boyu Zhou
AeroDASH predicts continuous B-spline trajectories for UAV flight and adapts delayed predictions by discarding expired segments and refitting the remainder to match the active path. It trains the policy with execution-aware GRPO against trajectories produced by asynchronous flight. On the UAV-Flow simulation test set, it reaches 76.56% success, versus 38.46% for OpenVLA-UAV; onboard flights are demonstrations, not quantitative real-world success-rate evaluations.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.