D4RL
收藏资源简介:
D4RL是由加州大学伯克利分校和谷歌大脑联合创建的数据集,旨在为深度数据驱动的强化学习提供基准。该数据集包含20个任务,涉及从简单的导航到复杂的机器人操作等多个领域。数据集的创建过程中,考虑了实际应用中的关键属性,如使用人类演示者、多任务数据集和混合策略收集的数据。D4RL的应用领域广泛,包括机器人学、自动驾驶和医疗保健,旨在解决这些领域中的复杂决策问题。
D4RL is a dataset jointly created by the University of California, Berkeley and Google Brain, aiming to provide benchmarks for deep data-driven reinforcement learning. This dataset includes 20 tasks spanning multiple domains, ranging from simple navigation to complex robotic manipulation. Key attributes in real-world applications were taken into account during the dataset's creation, such as data collected via human demonstrators, multi-task datasets, and data gathered using mixed policies. D4RL has wide-ranging application areas including robotics, autonomous driving and healthcare, and it is designed to solve complex decision-making problems in these fields.

- 1D4RL: Datasets for Deep Data-Driven Reinforcement Learning加州大学伯克利分校, 谷歌大脑 · 2021年



