BADGR Navigation Dataset

该数据集由BADGR系统在实际的城市和越野环境中收集的非策略数据组成，包含了各种导航经验。该数据集能够评估基于学习的导航策略与几何方法的效果，并提供了关于地形偏好和避障的洞察。数据集规模达到720,000个非策略数据点，其任务涉及移动机器人导航，在避免碰撞的同时，优先选择平滑地形以到达设定的GPS目标位置。This dataset consists of off-policy data collected by the BADGR system in real-world urban and off-road environments, encompassing a wide range of navigation experiences. It enables the evaluation of learning-based navigation strategies and geometric methods, while providing insights into terrain preference and obstacle avoidance. With a total of 720,000 off-policy data points, this dataset focuses on mobile robot navigation tasks, where agents prioritize traversing smooth terrain to reach predefined GPS target locations while avoiding collisions.