Human-inspired robotic texture exploration触摸速度和动作怎样影响纹理识别
How do touch speed and exploratory movements change what people and robots learn about texture?通过人类与机器人实验，比较触摸速度、接触力和探索动作对纹理感知的影响。
Studies velocity modulation and compound exploratory motions to improve texture discrimination under realistic contact variation.我通过人类触摸实验和神经形态机器人实验，分析速度与复合动作对纹理识别的影响。
- Role项目角色
- First author · Equal-contribution researcher第一作者 · 共同一作
- Period项目时间
- 2022–present
01 / problem问题
Why movement matters研究问题
Texture is not sensed passively: velocity, force and exploratory motion change the tactile evidence available to a person or robot.触摸速度、接触力和探索动作会改变人或机器人采集到的纹理信息。
02 / contribution工作
My contribution我负责的工作
I first-authored the human velocity-modulation study and contributed equally to the neuromorphic robotic movement study.我以第一作者完成了人类速度调节研究，并以共同一作身份参与神经形态机器人的探索动作研究。
- Developed NeuroTac event preprocessing and a lightweight spiking CNN, and deployed inference to SynSense Speck2f using Sinabs.完成 NeuroTac 事件预处理与轻量化 SCNN 开发，并通过 Sinabs 将推理部署到 SynSense Speck2f。
- Designed TacTip/NeuroTac housings, optical supports and robot flange adapters in Fusion 360, then iterated PolyJet/FDM prototypes.使用 Fusion 360 设计 TacTip/NeuroTac 壳体、光学支架与机械臂法兰适配件，迭代 PolyJet/FDM 原型。
03 / approach方法
Human-to-robot method实验设计
A human study tested 10 natural textures with 40 participants across six velocities and four force conditions. The robotic study compared six distinct or combined exploratory motions.人类实验招募 40 名参与者，在 6 种速度和 4 种接触力下探索 10 种自然纹理。机器人实验比较 6 种单独或组合的探索动作。
- Public acquisition code and research materials for the human single-finger exploration experiments are available on GitHub.人类单指探索实验的采集代码与研究材料已公开在 GitHub。
04 / outcome结果
Measured outcomes实验结果
Human perceptual dimensions predicted preferred speed with R²=0.77. Under varying speeds and contact depths, sliding plus rotation reached 87.33% texture accuracy versus 79.67% for sliding, a gain of 7.66 percentage points. Separate Speck2f measurements reported 8.04 mW mean on-chip operating power for sliding-plus-rotation inference.人类感知维度对偏好触摸速度的预测达到 R²=0.77。变速度与接触深度条件下，滑动加旋转的纹理识别率为 87.33%，纯滑动为 79.67%，相差 7.66 个百分点。独立的 Speck2f 功耗测量中，滑动加旋转输入的片上平均运行功耗为 8.04 mW。
Evidence论文与材料