Reproduce “KinoUAV: A Benchmark for Safe and Stable UAV Vision-Language Navigation with Kinodynamic Modeling” 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
Chenyao Yang, Peijin Wang, Yingchao Feng, Wenhui Diao
KinoUAV builds UAV navigation trajectories with simplified flight dynamics, then trains an agent to use visual estimates of directional free-space distance (DFD) when choosing accelerations. On the Test Seen simulation benchmark, its agent achieves 89.3% success and 6.0% collision rate, compared with AerialVLA at 52.3% success and 28.3% collisions. These results are benchmark-specific and do not establish real-world flight safety.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.