Abstract：

Tactile feedback is essential for generating tactile perception, which plays a critical role in haptic exploration and dexterity during object manipulation. The nervous system develops specialized tactile neural circuits for his purpose, shaped by experience. However, current electronic skins for tactile feedback fall short of achieving human-level tactile perception, such as the ability to rapidly recognize objects through touch and grasp. The use of backpropagation in training artificial neural networks is not biologically plausible and, therefore, unsuitable for incremental learning in physical dynamic environment.

In this talk, we present our advancements in developing a neuromorphic tactile system that enables self_x0002_powered sensing and rapid object classification with a temporal resolution of 1 ms for information encoding (1). Additionally, we introduce a novel neuromorphic learning approach that is 10 to 1,000 times faster than traditional backpropagation methods. (2). These breakthroughs hold significant potential for applications in robotics and prosthetics, as outlined in a perspective paper (3) and highlighted in Swedish media coverage (4–5).

1. Chen, et al., Spike-Timing Based Coding in Neuromimetic Tactile system Enables Dynamic Object Classification, Science, 384, 660-665-2024.

2. Xing, et al., ArXiv: 2410.00745. 2024

3. Micera, Toward more naturalistic tactile sensors, Science 384, 624, 2024

4. https://www.etc.se/halsa/framtidens-robotar-faar-hud-och-kaensel

5. https://www.dn.se/brandstudio/uppsala-universitet/forskare-har-skapat-artificiell-hud-med_x0002_beroringskansla/

      主讲人简介：

张志滨，瑞典乌普萨拉大学电子工程系研究员，博导，柔性电子和神经形态工程课题组(FENE)组长。2001年6月于中科院上海应用物理研究所博士毕业, 2002-2010年瑞典皇家理工学院博后，研究员。2013年于乌普萨拉大学获得电子学独立研究员资格 Docent in Electronics. 长期从事碳纳米管，石墨烯以及基于柔性，可印刷技术的可穿戴电子器件研究。近期该课题组主要在柔性传感器，电子皮肤，神经形态电子器件和电路设计，仿生机器学习和人工感知系统领域的研究。

FENE group webpage:

https://www.uu.se/institution/elektroteknik/forskning/fasta-tillstandets-elektronik/FENE