Graphene muscle with artificial intelligence

Ning-Qin Deng,He Tian,Fan Wu,Ye Tian,Xiao-Shi Li,Yang Xu,Yi Yang,Tian-Ling Ren
DOI: https://doi.org/10.1109/edtm47692.2020.9117928
2020-01-01
Abstract:Artificial intelligence (AI) shows powerful applications. Most of previous AI devices and systems were focused on electronic devices with pure electrical input to realize pattern recognition or language recognition. However, human can receive many other kinds of information (such as mechanical input). The research of AI with mechanical input remains elusive. Human muscle has the most important function with intelligence: dynamic training process that muscle lines can be broken first and recover to be stronger after exercise. Here, we show a graphene artificial muscle with mechanical intelligence. Multiple muscle lines can be directly printed on flexible substrate. After long run of training process, it shows three stages: decrement, enhancement and stabilized. Moreover, single artificial muscle line is built with laser-scribed graphene and shows "injured" muscle line in the first 1200 s and then start to "recover" after that. A new concept called "Force Learning" is also investigated in this work with accuracy reaching similar to 96%. Our work shows the great potential of building artificial muscle lines by laser-scribed graphene with the learning ability, which can offer widely applications in robot control, mechanical neuromorphic engineering.
What problem does this paper attempt to address?