用于机器人智能抓取的多参量三维触觉传感器研究

Tactile technology for robots has been obtained more and more attention in the world. It is an important research hotspot in the field of robotics just like visual technology and auditory technology in recent years. Tactile sensor is one of the core components of robots (especially service robots), and it is also a necessary technical premise for robots to realize man-machine collaboration..From a bionic perspective, the human skin has the ability to perceive a variety of tactile signals (contact pressure, shear force, slip sensation and temperature, etc.) simultaneously in the same position. At the present, the reported tactile sensors mostly have sole function, and there are no tactile sensors that can perform as much functions as human skin..This project focuses on the tactile perception needs of the application of intelligent robot hand grasping.Based on the bionic principle, the structure design, fabrication method and integrated packaging technology of the device are studied.It is proposed to develop an integrated tactile sensor array that integrates temperature, proximity, pressure, shear force and sliding sensing functions and can be detected synchronously.To solve the current robot tactile perception function is incomplete tactile information and other problems.On this basis, a fusion method of multi-parameter tactile signals is explored to improve the intelligent grasping ability of the robot's dexterous hand.

机器人触觉技术受到全世界范围内的广泛重视，是近年来机器人领域继视觉技术和听觉技术之后的重大研究热点。触觉传感器是机器人（尤其是服务机器人）的核心部件之一，更是机器人实现人机协同的必要技术前提。.从仿生角度来说，人类皮肤具有实时同位感知多种触觉信号（接触压力，剪切力，滑觉，温度等）的能力。目前已报道的触觉传感器功能单一，尚没有能实现类似人类皮肤功能的触觉传感器。.本项目围绕智能机器手抓握应用的触觉感知需求，基于仿生原理，从器件结构设计、制备方法和集成封装工艺等方面展开研究。拟开发集温度、临近、压力、剪切力及滑动感知功能一体的，可以同位同步检测的集成触觉传感器阵列，解决目前机器人触觉感知功能不健全，触觉信息不完整等问题。在此基础上，探索多参量触觉信号的融合方法，提升机器人灵巧手的智能抓握能力。

多功能的电子皮肤是丰富机器人、灵巧手对环境感知能力的技术基础。本项目在多模态触觉器件方面，分别开发了高性能的压力，温度，湿度传感器，为多参量触觉的实现提供了器件基础；探索并实现了多种触觉传感器的器件集成及信息融合，包括压力-摩擦电，电感-摩擦电等多传感器的集成，并在此基础上基于机器学习方法对多参量的触觉信息实现信息融合；并且基于等电势法的行列复用电路，实现了多通道、高精度、低延时的多模态触觉信号采集；针对视触融合信息不同质，不同步的痛点，开发了一种可视化的柔性压力传感阵列，并基于可视化的触觉信号结合机器学习方法实现对多物体的高精度辨识。