racing drone dataset
收藏资源简介:
该数据集由莫斯科斯科尔科沃科技学院智能空间机器人实验室创建,包含200个赛 drone 飞行任务,约20000张图像,涵盖穿越拱门形门、方形门和圆形赛道上的门等不同任务。数据采集过程中,使用Vicon运动捕捉系统记录了无人机的速度,T265相机以30Hz的频率捕捉图像。每个数据样本包括位置数据、速度分量、偏航角变化和同步的视觉帧。该数据集使用强化学习数据规范格式化,用于微调OpenVLA模型,以实现赛drone的高速度自主导航。
This dataset was created by the Intelligent Space Robotics Laboratory of the Skolkovo Institute of Science and Technology (Skoltech). It contains 200 racing drone flight missions and approximately 20,000 images, covering diverse tasks including passing through arched doors, square doors, and doors on circular race tracks. During data collection, the Vicon motion capture system was used to record the drone's velocity, while the Intel RealSense T265 camera captured images at 30Hz. Each data sample includes position data, velocity components, yaw angle changes, and synchronized visual frames. Formatted according to reinforcement learning data specifications, this dataset is utilized for fine-tuning the OpenVLA model to enable high-speed autonomous navigation of racing drones.




