柔性传感器以其卓越的形状适应性、高灵敏度及功能多样性，在机器人及其相关应用中展现出了巨大的潜力，并极大地推动了机器人技术的蓬勃发展。尽管传统的柔性传感器，如电阻式和电容式，已被广泛采用，但它们通常依赖于持续的外部电源供应。具有自供电优势的新型摩擦电传感器因其可由柔性或可拉伸材料制备，已成为柔性传感器设计的理想方案之一。 然而，在机器人应用中，现有的柔性摩擦电传感器面临多方面的系统性挑战。在传感器制备阶段，刚性的金属电极材料限制了柔性摩擦电传感器的可拉伸性与保形性；在信号采集环节，基于静电计的现有信号采集方案限制了传感器的应用范围；在高阶的信号分析阶段，现有摩擦电传感器功能单一，且容易受到耦合信号的干扰。为了攻克上述挑战，本文针对柔性摩擦电传感器在机器人应用中存在的关键问题开展了深入的研究，并做出了以下主要贡献：1） 提出了一种基于液态金属聚合物丝网印刷技术的摩擦电电子皮肤制备工艺。实现了低成本、批量化制备具有高度柔顺性和可拓展的可拉伸摩擦电传感器。2） 提出了一套基于跨阻放大器的小型化信号采集电路设计方案。通过深入分析柔性摩擦电传感器的输出特性，实现最优阻抗匹配设计，不仅显著缩小了信号采集设备的重量与体积，而且实现了多路信号的低串扰、高信噪比采集。3） 提出了一类基于小波变换的多模态摩擦电传感器信号处理算法。该算法利用摩擦电传感界面的本征物理关系解耦信号中的多模信息特征，并通过包括神经网络在内的机器学习技术对特征进行分析，实现了基于单一摩擦电信号的多模态信息同步感知。 实验结果表明，以上研究成果在多维人机交互、机器人触觉感知以及软体机器人本体感知等典型机器人应用领域中展现了柔性摩擦电传感器赋予机器人的多样化感知能力。为下一步拓展柔性摩擦电传感器在机器人和具身智能领域更广泛的应用奠定了基础。
Flexible sensors, with their exceptional shape adaptability, high sensitivity, and functional diversity, have demonstrated immense potential in robotics and related applications, significantly propelling the vigorous development of robotic technology. Although traditional flexible sensors, such as resistive and capacitive sensors, have been widely adopted, they typically rely on a continuous external power supply to work. The emerging triboelectric sensors, which have the advantage of self-powering and can be made from flexible or stretchable materials, have become one of the ideal solutions for the design of flexible sensors. However, existing flexible triboelectric sensors face multifaceted systemic challenges in robotic applications. In sensor fabrication, the rigid metallic electrode materials limit the stretchability and conformability of the flexible triboelectric sensors; for the signal acquisition device, the current electrometer-based signal acquisition schemes limit the application scope of the sensors; in the advanced signal analysis, the existing flexible triboelectric sensors are functionally singular and susceptible to interference from coupled signals. To overcome the aforementioned challenges, this dissertation conducts in-depth research on the key issues existing in the application of flexible triboelectric sensors in robotics and contributes to the following aspects:1) A fabrication process for triboelectric electronic skin based on liquid metal polymer screen printing technology is proposed. This has enabled the low-cost, mass production of highly compliant and scalable stretchable triboelectric sensors.2) A miniaturized acquisition circuit design scheme based on a trans-impedance amplifier signal conditioning circuit is proposed. By thoroughly analyzing the output characteristics of flexible triboelectric sensors, an optimal impedance matching design is achieved, which not only significantly reduces the weight and volume of the signal acquisition device but also realizes the low crosstalk and high signal-to-noise ratio acquisition of multiple signals.3) A class of multimodal triboelectric sensor signal processing algorithms based on wavelet transform is proposed. This algorithm utilizes the intrinsic physical relationship of the triboelectric sensing interface to decouple multimodal information characteristics in the signal and analyzes the features using machine learning techniques, such as neural networks, to achieve simultaneous perception of multimodal information based on a single triboelectric signal. Experimental results reveal that the above research achievements have demonstrated the diverse perceptual capabilities endowed to robots by flexible triboelectric sensors in typical robotic applications such as multidimensional human-robot interaction, robotic tactile perception, and soft robot proprioception. This lays the foundation for the next step in expanding the broader application of flexible triboelectric sensors in the fields of robotics and embodied intelligence.
