Neuromorphic object localization using resistive memories and ultrasonic transducers

Abstract

Real-world sensory-processing applications require compact, low-latency, and low-power computing systems. Enabled by their in-memory event-driven computing abilities, hybrid memristive-Complementary Metal-Oxide Semiconductor neuromorphic architectures provide an ideal hardware substrate for such tasks. To demonstrate the full potential of such systems, we propose and experimentally demonstrate an end-to-end sensory processing solution for a real-world object localization application. Drawing inspiration from the barn owl’s neuroanatomy, we developed a bio-inspired, event-driven object localization system that couples state-of-the-art piezoelectric micromachined ultrasound transducer sensors to a neuromorphic resistive memories-based computational map. We present measurement results from the fabricated system comprising resistive memories-based coincidence detectors, delay line circuits, and a full-custom ultrasound sensor. We use these experimental results to calibrate our system-level simulations. These simulations are then used to estimate the angular resolution and energy efficiency of the object localization model. The results reveal the potential of our approach, evaluated in orders of magnitude greater energy efficiency than a microcontroller performing the same task.

Experimental Demonstration of In‐Memory Computing in a Ferrofluid System

“…The power required by the FF to run the in‐memory computing scheme of Figure 3 is below 200 µW (Section , Supporting Information). Although the FF system cannot reach the power consumption of in‐memory solid‐state spiking neural networks architectures of nW orders, [ 27 ] the obtained value demonstrates low‐power operation, compatible with those of microcontrollers. The energy consumption of in‐memory computing, however, strongly depends on the type of processing executed.…”

Research progress in architecture and application of RRAM with computing-in-memory

“…The devices were tested with pulses to simulate sound signals, showing that the training accuracy and energy consumption of the system built using the memristor array were significantly improved, representing a significant advance in auditory localization systems with memristors. F. Moro et al 64 designed an event-driven object localization system inspired by the owl auditory cortex using a combination of state-of-the-art piezoelectric micromechanical ultrasound sensors and RRAM. The system was more efficient and power-efficient than microcontrollers performing the same task by several orders of magnitude.…”

An artificial neuromuscular junction for enhanced reflexes and oculomotor dynamics based on a ferroelectric CuInP 2 S 6 /GaN HEMT

“…Analogous to conventional GaN HEMTs, the proposed CIPS/ GaN FeHEMT also potentially attains nanosecond operation by etching the AlGaN barrier layer to minimize the surface potential by using the thinner CIPS membrane to alleviate the voltage drops at CIPS (22,37). The direct actuation of the mechanical platform using the artificial NMJ provides a wide range of neuromorphic sensing-to-action applications, including time-offlight ranging (38)(39)(40)(41)(42)(43), in-sensor/near-sensor computing (44)(45)(46)(47)(48), and human-computer interaction (49). In this study, we achieved a normalized output current of 200 mA/mm with the CIPS/GaN FeHEMT, which is notably greater than that of recently reported synaptic transistors (for more details, see table S1) (50)(51)(52)(53)(54)(55)(56)(57)(58)(59).…”