Flexible and Biocompatible, Organic Electronics Spiking Neuromorphic Computing
Abstract
The brain of a fruit fly is fit with only about 100,000 neurons, and yet it can perform in real time such complex and disparate tasks as flight control, 3D path planning, food and mate search, and predator avoidance while consuming microwatts of power. Additionally, it is trained by example and learns to extrapolate. Finally, when damaged, brains avoid catastrophic failure by instead suffering performance degradation proportional to the damage. In contrast, our conventional supercomputers consume megawatts of power and take minutes, hours, or longer to carry out complex, nonlinear, and nonsequential calculations. Additionally, they are explicitly programmed for a specific task, and suffer catastrophic failure when damaged.Neuromorphic Computing is a biologically inspired type ofcomputing that is becoming an attractive paradigm for next-generation computing. It aims to emulate the computational principles ofbiological neural circuits and systems to create new algorithms for patternrecognition and control. Its advantages include real time and low power computing, distributed processing, resilience against damage and component failure, and programming by example.Many neuromorphic implementations have been demonstrated, including commercial and academic efforts. Current applications include self-driving cars, image and sensory processing, and brain-machine and human-computer interfaces. At present, the overwhelming majority of those implementations rely on hard, rigid and brittle, Si-based CMOS electronics. They are not suitable for many applications that require physical flexibility, large-area and low-cost fabrication, and biocompatibility. We propose to develop complete neuromorphic computing circuit systems, fit with sensors, actuators, and a network of spiking neurons (synapses and somas), that perform information processing. The synaptic and somatic circuits will be implemented using physically flexible and biocompatible organic electronics. The entire soft neuromorphic computation circuit will be fabricated. As a proof of concept demonstration, we will fabricate a simple, mobile soft PaperBot, where the sensory data is processed using organic electronics spiking neuromorphic circuits, and an appropriate output signal is sent to an actuator driving and steering the motion of the PaperBot. This will pave the way for future smart, low cost, low power, soft distributed computing, akin to distributed computation performed by octopus skin, for instance for use in soft robots, or implanted human-machine and brain-machine interfaces.These efforts align well with ONRs goal of research on bio-inspired design of distributed sensing, actuation and control, specifically in the framework of integration with physically soft robotics. It also is in line with the ONRs focus on functional polymeric and organic materials, and DoDs initiative on flexible hybridelectronics (FHE).Approved for Public Release.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Aug 05, 2021
- Source ID
- N000142112585
Entities
People
- Robert A Nawrocki
Organizations
- Office of Naval Research
- Purdue University
- United States Navy