TIMID
收藏资源简介:
TIMID是由萨拉戈萨大学团队开发的机器人执行视频时序错误检测数据集,包含多机器人协同任务的模拟与真实执行视频。该数据集通过Gazebo仿真环境生成,重点捕捉互斥访问和顺序执行两类时序约束任务,包含正常执行与人为注入的协议级错误样本。数据集设计支持弱监督训练,提供视频级标签及LTL公式描述的任务规范,旨在解决复杂任务中时序逻辑违规的检测难题,为机器人执行监控提供基准测试平台。
TIMID is a robotic execution video temporal error detection dataset developed by the research team at the University of Zaragoza, which covers both simulated and real-world execution videos of multi-robot collaborative tasks. This dataset is generated using the Gazebo simulation environment, focusing on two categories of temporal constraint tasks: mutually exclusive access and sequential execution, and includes both normally executed samples and manually injected protocol-level error samples. The dataset is designed to support weakly supervised training, providing video-level labels and task specifications expressed via Linear Temporal Logic (LTL) formulas. It aims to address the challenge of detecting temporal logic violations in complex robotic tasks, serving as a benchmark platform for robot execution monitoring.
- 1TIMID: Time-Dependent Mistake Detection in Videos of Robot Executions萨拉戈萨大学·系统工程与计算机科学系; 都灵大学 · 2026年



