从高带宽单点传感中学习触觉感知
Learning tactile perception from high-bandwidth single-point sensing
- Wormsensing
- Hugging Face
机构由 AI 辅助整理，请以论文原文为准。
AI总结：
本文提出SpectRobot框架，将单点高带宽触觉信号转为频谱图，利用标准视觉编码器处理，实现遮挡下的机器人操作，并证明时间历史与带宽的重要性，且兼容多种传感技术。
AI中文摘要：
触觉传感正越来越多地被纳入基于学习的机器人操作中，然而许多现有方法依赖于空间分布的传感器。在此，我们介绍了SpectRobot，一个将单点触觉信号转换为紧凑的时频频谱图的框架。这些频谱图将高带宽的触觉历史编码为固定大小的图像类表示。它们可以被标准的视觉编码器处理，并集成到最初为视觉开发的学习流程中，同时保留了传统相机无法获取的时间和频率信息。SpectRobot并非通过增加触觉元件阵列的空间密度，而是利用稀疏、高带宽单点测量中包含的丰富动态信息。在我们的实现中，传感器安装在远离接触表面的位置，同时保持与接触表面的机械耦合，从而减少直接磨损暴露，并可能提高在恶劣环境中的鲁棒性以及灵巧机器人的长期部署能力。我们的实验表明：（1）机器人可以利用单点振动信号解决视觉遮挡的操作任务；（2）时间历史对策略性能有强烈影响，而传感带宽控制可用的频谱信息，测量范围延伸至100 kHz；（3）相同的表示可以用于不同的触觉传感技术，包括基于加速度、力或应变的传感。我们进一步表明，以前与研究级仪器相关的能力可以通过现成的商用硬件实现。我们相信，更广泛地获取高带宽触觉传感可以促进接触动态融入具身学习系统，并且对于某些任务，可以提供增加触觉传感空间密度的替代方案或补充。
英文摘要：
Tactile sensing is increasingly being incorporated into learning-based robotic manipulation, yet many existing approaches rely on spatially distributed sensors. Here we introduce {SpectRobot}, a framework that transforms single-point tactile signals into compact time-frequency spectrograms. These spectrograms encode high-bandwidth tactile histories as fixed-size image-like representations. They can be processed by standard vision encoders and integrated into learning pipelines originally developed for vision, while preserving temporal and frequency information unavailable to conventional cameras. Rather than increasing spatial density through arrays of tactile elements, SpectRobot exploits the rich dynamics contained in sparse, high-bandwidth single-point measurements. In our implementation, the sensors are mounted away from the contact surface while remaining mechanically coupled to it, reducing direct exposure to wear and potentially improving robustness in harsh environments and for long-term deployment on dexterous robots. Our experiments demonstrate that: (1) a robot can exploit single-point vibration signals to solve a visually occluded manipulation task; (2) temporal history strongly influences policy performance, while sensing bandwidth controls the spectral information available, with measurements extending to 100~kHz; and (3) the same representation can be used across different tactile sensing technologies mediated by acceleration, force, or strain. We further show that capabilities previously associated with research-grade instrumentation can be accessed using readily available, off-the-shelf hardware. We believe that broader access to high-bandwidth tactile sensing could facilitate the integration of contact dynamics into embodied learning systems and, for some tasks, offer an alternative or complement to increasing the spatial density of tactile sensing.
