The advent of the artificial intelligence (AI) and Internet of Things (IoTs) era has spurred a surge in the analysis of voluminous data gathered from myriad distributed sensors. This endeavor is primarily aimed at executing sophisticated recognition functions, which frequently demand excessive energy consumption. As a result, the development of a streamlined design capable of performing these functions with comparable efficiency continues to pose a significant challenge. Herein, a rigiflex pillar-membrane triboelectric nanogenerator (PM-TENG) is proposed for universal stereoscopic recognition by machine learning. An integral design is adopted to generate dynamic sensing signals in time series, which can obtain abundant and high-resolution information of stereoscopic structures. By combining the advantages of both rigid steel pillars and flexible/elastic membranes, the proposed rigiflex PM-TENG contains information from multiple sensing pillared pixels and focuses on the study of dynamic changes during the whole contact cycle. The proposed rigiflex TENG can effectively recognize objects across nine categories by leveraging machine learning technique, achieving an accuracy rate of 96.39 %. This system offers substantial potential for application in assembly lines for production control management in future smart factories and unattended warehouse workshops.

中文翻译：

---

机器学习增强型 rigiflex 柱膜摩擦纳米发电机用于通用立体识别


人工智能 (AI) 和物联网 (IoT) 时代的到来刺激了对从无数分布式传感器收集的大量数据进行分析的激增。这项工作主要旨在执行复杂的识别功能，这些功能经常需要过多的能量消耗。因此，开发能够以相当的效率执行这些功能的流线型设计仍然是一个重大挑战。在此，提出了一种 rigiflex 柱膜摩擦纳米发电机（PM-TENG），用于通过机器学习进行通用立体识别。采用整体设计产生时间序列的动态传感信号，可以获得丰富、高分辨率的立体结构信息。通过结合刚性钢柱和柔性/弹性膜的优点，所提出的rigiflex PM-TENG包含来自多个传感柱像素的信息，并重点研究整个接触周期期间的动态变化。所提出的rigiflex TENG可以利用机器学习技术有效地识别九个类别的物体，准确率达到96.39%。该系统在未来智能工厂和无人值守仓库车间的生产控制管理装配线中具有巨大的应用潜力。