Heterogeneous Structure Omnidirectional Strain Sensor Arrays With Cognitively Learned Neural Networks.

Mechanically stretchable strain sensors gain tremendous attention for bioinspired skin sensation systems and artificially intelligent tactile sensors. However, high-accuracy detection of both strain intensity and direction with simple device/array structures is still insufficient. To overcome this limitation, an omnidirectional strain perception platform utilizing a stretchable strain sensor array with triangular-sensor-assembly (three sensors tilted by 45°) coupled with machine learning (ML) -based neural network classification algorithm, is proposed. The strain sensor, which is constructed with strain-insensitive electrode regions and strain-sensitive channel region, can minimize the undesirable electrical intrusion from the electrodes by strain, leading to a heterogeneous surface structure for more reliable strain sensing characteristics. The strain sensor exhibits decent sensitivity with gauge factor (GF) of ≈8, a moderate sensing range (≈0-35%), and relatively good reliability (3000 stretching cycles). More importantly, by employing a multiclass-multioutput behavior-learned cognition algorithm, the stretchable sensor array with triangular-sensor-assembly exhibits highly accurate recognition of both direction and intensity of an arbitrary strain by interpretating the correlated signals from the three-unit sensors. The omnidirectional strain perception platform with its neural network algorithm exhibits overall strain intensity and direction accuracy around 98% ± 2% over a strain range of ≈0-30% in various surface stimuli environments.

Retina-Inspired Color-Cognitive Learning via Chromatically Controllable Mixed Quantum Dot Synaptic Transistor Arrays.

Artificial photonic synapses are emerging as a promising implementation to emulate the human visual cognitive system by consolidating a series of processes for sensing and memorizing visual information into one system. In particular, mimicking retinal functions such as multispectral color perception and controllable nonvolatility is important for realizing artificial visual systems. However, many studies to date have focused on monochromatic-light-based photonic synapses, and thus, the emulation of color discrimination capability remains an important challenge for visual intelligence. Here, an artificial multispectral color recognition system by employing heterojunction photosynaptic transistors consisting of ratio-controllable mixed quantum dot (M-QD) photoabsorbers and metal-oxide semiconducting channels is proposed. The biological photoreceptor inspires M-QD photoabsorbers with a precisely designed red (R), green (G), and blue (B)-QD ratio, enabling full-range visible color recognition with high photo-to-electric conversion efficiency. In addition, adjustable synaptic plasticity by modulating gate bias allows multiple nonvolatile-to-volatile memory conversion, leading to chromatic control in the artificial photonic synapse. To ensure the viability of the developed proof of concept, a 7 × 7 pixelated photonic synapse array capable of performing outstanding color image recognition based on adjustable wavelength-dependent volatility conversion is demonstrated.

Large-Area Pixelized Optoelectronic Neuromorphic Devices with Multispectral Light-Modulated Bidirectional Synaptic Circuits.

The complete hardware implementation of an optoelectronic neuromorphic computing system is considered as one of the most promising solutions to realize energy-efficient artificial intelligence. Here, a fully light-driven and scalable optoelectronic neuromorphic circuit with metal-chalcogenide/metal-oxide heterostructure phototransistor and photovoltaic divider is proposed. To achieve wavelength-selective neural operation and hardware-based pattern recognition, multispectral light modulated bidirectional synaptic circuits are utilized as an individual pixel for highly accurate and large-area neuromorphic computing system. The wavelength selective control of photo-generated charges at the heterostructure interface enables the bidirectional synaptic modulation behaviors including the excitatory and inhibitory modulations. More importantly, a 7 × 7 neuromorphic pixel circuit array is demonstrated to show the viability of implementing highly accurate hardware-based pattern training. In both the pixel training and pattern recognition simulation, the neuromorphic circuit array with the bidirectional synaptic modulation exhibits lower training errors and higher recognition rates, respectively.