外观
论文arXiv 2609.159101 个月前
SlipSense：用于低延迟和泛化滑移检测的多模态触觉学习
原文链接：SlipSense: Multimodal Tactile Learning for Low-Latency and Generalized Slip Detection（arXiv 2609.15910 · 发布日期：2026-09-14）
SlipSense: Multimodal Tactile Learning for Low-Latency and Generalized Slip Detection
SlipSense用压阻阵列与加速度计多模态融合，实现低延迟滑移检测，并在37物体数据集和跨平台迁移中验证。
核心要点
- 滑移检测对灵巧操作关键，现有系统对检测延迟和跨平台泛化刻画不足。
- 基于TacV5融合压阻阵列与3轴MEMS加速度计，进行模态编码、传感器内融合和跨模态注意力。
- 在140万帧、37物体数据集上，Macro F1为96.7%，误报率低于1.6%，76%滑移在23.1 ms内检出。
- 仅用UMI数据训练时，可零样本泛化到Tesollo灵巧手，跨未见物体、传感器单元和平台。
验证范围与局限：验证范围包括37个物体数据集，以及从UMI数据训练到Tesollo灵巧手的零样本迁移。
英文摘要 (Abstract)
Slip detection is fundamental to dexterous manipulation, yet existing systems often lack precise characterization of detection latency and cross-platform generalization. We present SlipSense, a multimodal tactile slip-detection framework built on TacV5, a compact sensor integrating a $32 \times 32$ piezoresistive array operating at 240 Hz and a 3-axis MEMS accelerometer operating at 8 kHz. The piezoresistive array captures spatial pressure distributions, while the accelerometer captures friction-induced vibrations, providing complementary slip cues. The framework performs modality-specific encoding, intra-sensor fusion, and cross-modal attention with causal temporal prediction at 240 Hz. Experiments on a dataset of 1.4 million frames spanning 37 objects demonstrate the complementarity of the two modalities. SlipSense achieves 96.7% Macro F1 with a false-positive rate below 1.6%, detecting 76% of slip events within 23.1 ms. When trained solely on UMI data, SlipSense generalizes zero-shot to a Tesollo dexterous hand, transferring across unseen objects, distinct sensor units, and robotic platforms without retraining.
作者：Tong Jian, Aditya Thurvas Senthil Kumar, Xinyi Li, Ziling Chen, Tianyu Dai, Ali Sengül, Matteo Grimaldi, Wenjie Lu, Saleh Nabi, Tao Yu
来源
本页事实对应的公开页面，以原文为准。