Tactile sensing is essentially required for dexterous manipulation in robotic applications. Mimicking human perception of softness identification in a non-invasive fashion, thus achieving satisfactory interaction with fragile objects remains a grand challenge. Here, a scatheless measuring methodology based on the multisensory electronic skins to quantify the elastic coefficient of soft materials is reported. This recognition approach lies in the preliminary classification of softness by piezoelectric signals with a modified machine learning algorithm, contributing to an appropriate contact force assignment for subsequent quantitative measurements via strain sensing feedback. The integration of multifunctional sensing system allows the manipulator to hold capabilities of self-sensing and adaptive grasping motility in response to objects with the various softness (i.e., kPa-MPa). As a proof-of-concept demonstration, the biomimetic manipulator cooperates with the robotic arm to realize the intelligent sorting of oranges varying in freshness, paving the way for the development of microsurgery robots, human-machine interfacing, and advanced prosthetics.

机器人应用中的灵巧操作本质上需要触觉感应。以非侵入性方式模拟人类对柔软度识别的感知，从而实现与易碎物体的令人满意的交互仍然是一个巨大的挑战。在这里，报告了一种基于多感官电子皮肤的无损伤测量方法，用于量化软材料的弹性系数。这种识别方法在于使用改进的机器学习算法通过压电信号对柔软度进行初步分类，有助于通过应变传感反馈为随后的定量测量分配适当的接触力。多功能传感系统的集成使机械手具有自感知和自适应抓取运动的能力，以响应具有各种柔软度（即kPa-MPa）的物体。作为概念验证演示，仿生机械手与机械臂配合，实现了对新鲜度不同的橙子的智能分拣，为显微手术机器人、人机界面和先进假肢的发展铺平了道路。