NatSGLD
收藏资源简介:
NatSGLD数据集是由美国马里兰大学帕克分校的研究团队创建的,包含了18名参与者在11项活动中与机器人进行的1143次多模态指令交互(包括言语和手势)。该数据集通过Wizard of Oz方法收集,记录了人类的多模态指令、示范轨迹和线性时序逻辑公式,为机器人提供了丰富的交互模态和强大的监督信号,可应用于多模态指令跟随、计划识别、基于示范的人类可指导的强化学习等领域。
The NatSGLD dataset was created by a research team from the University of Maryland, College Park, United States. It contains 1,143 multimodal instructional interactions (including speech and gestures) between 18 participants and robots across 11 activities. Collected via the Wizard of Oz methodology, this dataset records human multimodal instructions, demonstration trajectories, and linear temporal logic formulas. It provides rich interaction modalities and robust supervisory signals for robots, and can be applied to domains such as multimodal instruction following, plan recognition, and demonstration-based human-guided reinforcement learning.

- 1NatSGLD: A Dataset with Speech, Gesture, Logic, and Demonstration for Robot Learning in Natural Human-Robot Interaction美国马里兰大学帕克分校 · 2025年



