Flightmare Simulator

该数据集提供了一个模拟环境，用于训练基于视觉的无人机导航神经网络的策略。此外，该数据集支持利用完整状态信息来训练一种感知意识教师策略。其任务是在杂乱环境中训练基于视觉的控制策略，以实现四旋翼无人机的有效导航。This dataset provides a simulated environment for training vision-based neural network policies for drone navigation. Additionally, it supports training a perception-aware teacher policy by leveraging full state information. The core task of this dataset is to train vision-based control policies to enable effective navigation of quadrotor unmanned aerial vehicles (UAVs) in cluttered environments.

搜集汇总

背景与挑战

背景概述

Flightmare是一个模块化的四旋翼无人机模拟器，具有可配置的渲染和物理引擎，支持多模态传感器和并行计算，适用于强化学习、路径规划等多种应用。