苏州大学《AFM》：激光诱导石墨烯触觉芯片构筑双向触觉通信，用于机器人遥操作

当前的机器人遥操作大多依赖基于视觉的操作环境探测与反馈，对接触层面的”触觉”信息捕捉不足；如何建立主从两端双向的触觉感知与反馈，始终是提升机器人综合态势感知能力、突破”隔空操作”临场感瓶颈的关键难题。基于此，本文，苏州大学李永泽、陈涛 教授、何剑 教授团队在《Advanced Functional Materials》期刊发表题为”Distributed Haptic Chips With Materials’ Multifunctionalities for Dual-Way Tactile Communication Enabled Terrain Recognition and Reconstruction in Robotic Teleoperation”的研究，借助激光诱导石墨烯（LIG）与硅胶两类材料，提出一种触觉芯片，可为遥操作系统中主端与从端同时提供多维度、多模态的触觉感知与反馈。

此外，该芯片集成摩擦电与压阻传感阵列，具备三维动态与静态触觉感知能力；其巧妙之处在于，摩擦起电的”空间”被复用为主端气动触觉反馈，而压阻传感器中的 LIG 层还可兼作热刺激加热器，实现”一材多用”；同时集成的压电芯片进一步提供高频触觉感知与反馈。经系统集成，多枚触觉芯片可构建分布式触觉通信网络；搭载该芯片的六足机器人在行走与攀爬过程中，可基本识别地形、姿态与纹理，并能在主端触觉芯片阵列上被选择性”再生”呈现，触觉芯片本身也能直接发出控制指令。本研究提出了一种以触觉芯片为载体的双向触觉通信范式，作为视觉系统的互补控制接口，为特殊任务场景下机器人的遥操作提供了高临场感、多模态的新思路。

图1、 展示触觉芯片的总体设计构想与材料体系。以激光诱导石墨烯（LIG）与硅胶的多功能特性为核心，勾勒出面向遥操作主从双端的”感知—反馈”一体化芯片框架，点明摩擦电、压阻、压电多模态集成的设计思路。

图2、Characterizations of sensing functions of the haptic chip. (a) Working principle of haptic chip with torque transmission cross. (b) Characterization data (peak-peak voltage) of triboelectric sensor for comparing the design parameters of the torque transmission cross with a constant contact force (20N), i) the output signals recorded from the varied pillar heights of 3, 4, 5, and 6 mm, ii) the output signals recorded from the varied pillar diameters of 4, 5, 6, and 7 mm, iii) the output signals recorded from the varied cross thicknesses of 1, 2, 3, and 4 mm. (c) Characterization data of piezoelectric sensor, i) the experimental data of sensitivity and linearity, ii) the hysteresis data, and iii) the data of response time. (d) i) Characterization of resistance variations of the LIG layer for the heating and cooling processes provided by an external heater, ii) Characterization of resistance variations of the LIG layer for the heating and cooling processes done by self-heating as thermal feedback. (e) Experiments of verifying stability and repeatability of pressure sensing before and after the heating process, with a constant pressure of 10N.

图3、Characterization of haptic feedback functions of haptic chip. (a) Schematics of pneumatic actuation of haptic chips with a piezoelectric vibrator located at the center. (b) The experimental data of pneumatic actuation-induced displacements with different materials of actuation membranes (0.1 mm thickness) and a constant input pressure of 40 kPa. (c) Relationship between the thickness of actuation membranes made by Dragon Skin FX Pro and the corresponding displacements with an input pressure of 40 kPa; the sidewall is replaced by Dragon Skin 30. (d) Relationship between the input pressures of the actuation membrane (0.1 mm thickness) made by Dragon Skin FX Pro and the corresponding output forces. (e) Photos of the haptic chip array at the user side, and laser profiles of regenerating perceptions of two terrains, including hollow and bump. (f) Schematics and the experimental data of using piezoresistive sensors to monitor the pneumatic actuation process; the red data curve shows the stepwise inflation and deflation process, the blue data curve shows the repeated inflation and deflation cycle with a constant input pressure (40 kPa). (g) Schematics and the characterization data of a piezoelectric vibrator operating with encapsulation and without encapsulation. (h) Schematics and the data of programmable multi-channel vibrational feedback from array-based piezoelectric vibrators monitored by attaching piezoelectric sensors on vibrators. (i) Schematics of compression-induced thermal feedback and thermal images showing the selective activation of the corresponding region via pressing of the rod. (j) Characterization data of thermal feedback temperatures generated by the corresponding input power. (k) Response time of thermal feedback for the input power of 0.05 and 0.09 W.

图4、Multi-channel sensing of haptic chip for perceiving contact status. (a) Schematics and piezoresistive sensing signals from four units to detect the contacting angles. (b) Schematics and triboelectric sensing signals to detect the surface roughness via sliding. (c) Schematics and fusion of triboelectric and piezoresistive sensing signals to detect the hardness of the contacting surface. (d) i) Schematics of detecting shear force directions by four channels, the experimental data of detecting shear force directions by ii) triboelectric outputs and iii) piezoresistive outputs. (e) Demonstrations of using multi-channel sensing for i) achieving hexapod control by detecting control commands at the user side, and ii) realizing the detection of contact status via haptic chips at the feet of the hexapod.

图5、Demonstration of recognizing surface materials via distributed haptic chips. (a) i) Schematics of selected haptic chips and sensing channels for collecting tactile information, ii) flow diagram of the process for intelligent recognition. (b) i)Photos of the tested surfaces, including acrylic, wood, gravel, carpet, stone, grass, and metal, ii) sample data collected from the corresponding surfaces during walking, iii) t-SNE visualization of recognition results, iv) confusion matrix with an overall accuracy of 98.57%, v) screenshot of demonstration video of surface materials recognition.

图6、Demonstration of recognizing terrains via distributed haptic chips. (a) Schematics of selected haptic chips and sensing channels for collecting tactile information. (b) i) Schematics of the tested terrains, including bump, flat surface, hollow, and ramp, ii) sample data collected from the corresponding terrains during walking, iii) t-SNE visualization of recognition results, iv) confusion matrix with an overall accuracy of 95.24%, v) screenshot of demonstration video of terrain recognition. (c) Subjective experiments for evaluating recognition accuracy of haptic feedback: i) example of regenerated terrain pattern and its real geometry for pneumatic haptic feedback, ii) five pneumatic feedback patterns for terrain regeneration, iii) confusion matrix of subjective tests shows the accuracy of 93.6%.

图7、Demonstration of tactile communication-enhanced closed-loop hexapod manipulation. (a) Flow diagram of motion control of the hexapod via haptic chip array and haptic feedback with the recognized information from distributed haptic chips at the hexapod. (b) Screenshots, photos, and real-time recorded data during i) motion control of the hexapod and haptic feedback of the hexapod-recognized information at the user side, and ii) recognition of terrain via multi-channel sensing data at the hexapod side. (c) Thermal feedback assisted heat source recognition, i) setup of subjective experiments for identifying hidden heat sources, ii) Test conditions with randomly arranged sequences, and iii) statistical data of true (activated) and perceived activations of the hot plate.

综上，该工作以激光诱导石墨烯与硅胶的多功能材料为基础，构建出可同时服务遥操作主从双端的触觉芯片：摩擦电—压阻阵列实现三维动静态触觉感知，摩擦起电空间复用为气动反馈、LIG 层兼作热刺激，压电芯片补足高频反馈；多芯片进一步组网为分布式触觉通信网络，并成功赋予六足机器人地形、姿态与纹理识别及主端”再生”能力。该触觉芯片系统作为视觉通道的互补接口，为特殊任务下机器人的高临场感遥操作提供了新范式。

文献：https://doi.org/10.1002/adfm.78287