Printed synaptic transistor-based electronic skin for robots to feel and learn.

Area of Science:

• Materials Science
• Robotics
• Neuroscience

Background:

• Next-generation robots require electronic skin (e-skin) with multimodal sensing and memory capabilities.
• Achieving uniform, large-area electronic devices with synaptic behavior is crucial for bio-like robotic sensing.

Purpose of the Study:

• To develop high-quality, uniform synaptic transistors on flexible substrates for advanced e-skin.
• To demonstrate the potential of these transistors for in-hardware learning and bio-like robotic sensation.

Main Methods:

• Fabrication of a 12-by-14 array of synaptic transistors using printed ZnO nanowires on a flexible substrate.
• Characterization of device performance, including synaptic behavior (excitatory/inhibitory post-synaptic current, plasticity, memory transition).
• Integration into a prototype computational e-skin with event-driven sensors and spiking neurons for robotic hand application.

Main Results:

• Achieved 100% yield and high uniformity in the fabricated ZnO nanowire synaptic transistors.
• Demonstrated robust bio-like synaptic behavior, including short-term to long-term memory transition.
• Successfully implemented a computational e-skin capable of associative learning and acquiring a pain reflex in a robotic hand.

Conclusions:

• The developed synaptic transistors exhibit excellent bio-like properties, paving the way for advanced in-hardware learning in robots.
• The computational e-skin prototype showcases the potential for localized peripheral nervous system-like learning, reducing data latency and cognitive load.
• This technology offers a significant advancement in creating robots with more natural and adaptive sensory experiences.