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

Abstract

An electronic skin (e-skin) for the next generation of robots is expected to have biological skin-like multimodal sensing, signal encoding, and preprocessing. To this end, it is imperative to have high-quality, uniformly responding electronic devices distributed over large areas and capable of delivering synaptic behavior with long- and short-term memory. Here, we present an approach to realize synaptic transistors (12-by-14 array) using ZnO nanowires printed on flexible substrate with 100% yield and high uniformity. The presented devices show synaptic behavior under pulse stimuli, exhibiting excitatory (inhibitory) post-synaptic current, spiking rate-dependent plasticity, and short-term to long-term memory transition. The as-realized transistors demonstrate excellent bio-like synaptic behavior and show great potential for in-hardware learning. This is demonstrated through a prototype computational e-skin, comprising event-driven sensors, synaptic transistors, and spiking neurons that bestow biological skin-like haptic sensations to a robotic hand. With associative learning, the presented computational e-skin could gradually acquire a human body–like pain reflex. The learnt behavior could be strengthened through practice. Such a peripheral nervous system–like localized learning could substantially reduce the data latency and decrease the cognitive load on the robotic platform.

Sensorimotor Correlation Using Printed Synaptic Transistors and Conditioning PCB

“…The self-aligned, automated setup minimises the location-to-location and batch-to-batch non-uniformity in the printed NWs and holds a great promise for the development of highly uniform, large-area electronics using printed NWs. The synaptic behaviour of the as-fabricated device has been thoroughly explained in the previous work [25]. Nevertheless, for the sake of completion, the test showing the synaptic behaviour of the transistor is shown in Fig.…”

“…Specifically, this requires a block to provide the sensory input and a block for the motor output, both in a neuromorphic manner. The working diagram can be found in our previous paper [25]. In this paper, we focus on the actual implementation (i.e., circuit diagram) of the whole conditioning circuit.…”

“…Several works have been carried out to realise the abovedescribed concept using emerging synaptic devices [23][24][25]. For example, we have demonstrated the "in-skin" learning of the pain-like reflex using printed ZnO nanowires (NWs) based synaptic transistors.…”

“…For example, we have demonstrated the "in-skin" learning of the pain-like reflex using printed ZnO nanowires (NWs) based synaptic transistors. While our previous paper has mainly focused on the scientific aspects (e.g., device physics, hardware neuropathway demonstration along with associative learning and the theory of use and disuse, etc) [25], the technological details such as the arrangement of the conditioning circuit, especially in a PCB manner and the fabrication of the synaptic transistors were not reported in detail. For better reproducing of our results, herein we present these technical details for the realisation of the adaptive sensory-motor correlation.…”

Deep Active Cross-Modal Visuo-Tactile Transfer Learning for Robotic Object Recognition

“…Visuo-tactile cross-modal perception and learning is a challenging problem due to the weak-pairing between visual and tactile data: (a) variation in density of information from each modality, (b) scale gap as vision sensors can capture the global scene while tactile sensors capture local object geometry, (c) temporal misalignment as vision sensors capture data in one-shot while tactile sensors capture data sequentially, and (d) tactile data are inherently action conditioned as data depends on the type of action that is performed [14], [15]. Few works in literature address the problem of visuo-tactile cross-modal learning.…”