SIMCOACHCORPUS
收藏资源简介:
SIMCOACHCORPUS数据集是一个独特的数据集,包含赛车模拟器驾驶的交互式数据,允许研究人员研究在有专家指导下和没有指导下的运动技能习得过程中的丰富交互现象。该数据集由29位人类参与者在赛车模拟器中围绕赛道驾驶大约九十分钟的数据组成。数据集包括身体特征,如车辆状态和输入、地图(赛道边界和赛车线)、锥形地标,这些都与专业教练的实时口头指导和每个回合结束时的额外反馈同步。数据集还包括对每个实时反馈话语的教练类别注释、学生对教练建议的遵守程度的评级,以及参与者自我报告的认知负荷和情绪状态(在研究期间通过调查收集)。数据集包含超过20,000个实时反馈话语、超过400个回合结束时的反馈话语,以及超过40小时的车辆驾驶数据。我们的自然主义数据集可以用于研究运动学习动态、探索语言现象,以及训练教学计算模型。我们展示了该数据集在上下文学习、模仿学习和主题建模中的应用。本工作中介绍的数据集将在同行评审的论文发表后公开发布。有兴趣的研究人员可以在 https://youtu.be/PgLtIH5Qbv0 注册获取早期访问权限。
The SIMCOACHCORPUS is a unique dataset containing interactive data from racing simulator driving, enabling researchers to investigate rich interactive phenomena during motor skill acquisition with and without expert guidance. This dataset comprises data from 29 human participants who completed approximately 90 minutes of track driving in a racing simulator. It includes driving-related physical data such as vehicle states and inputs, maps (track boundaries and racing lines), and cone landmarks, all synchronized with real-time verbal coaching from professional instructors and supplementary feedback provided at the end of each driving session. The dataset also features annotations of coach categories for every real-time feedback utterance, ratings of the student’s compliance with the coach’s recommendations, as well as participants’ self-reported cognitive load and emotional states collected via surveys during the study. The dataset contains over 20,000 real-time feedback utterances, more than 400 post-session feedback utterances, and over 40 hours of vehicle driving data. This naturalistic dataset can be utilized to investigate motor learning dynamics, explore linguistic phenomena, and train instructional computational models. We demonstrate the applicability of this dataset across in-context learning, imitation learning, and topic modeling. The dataset introduced in this work will be publicly released following the publication of the peer-reviewed paper. Interested researchers can register for early access at https://youtu.be/PgLtIH5Qbv0.

- 1通过丰田研究院 · 2025年



