| NSF Org: |
CCF Division of Computing and Communication Foundations |
| Recipient: |
|
| Initial Amendment Date: | June 10, 2019 |
| Latest Amendment Date: | August 17, 2022 |
| Award Number: | 1909797 |
| Award Instrument: | Standard Grant |
| Program Manager: |
Sankar Basu
sabasu@nsf.gov (703)292-7843 CCF Division of Computing and Communication Foundations CSE Direct For Computer & Info Scie & Enginr |
| Start Date: | July 1, 2019 |
| End Date: | June 30, 2022 (Estimated) |
| Total Intended Award Amount: | $500,000.00 |
| Total Awarded Amount to Date: | $500,000.00 |
| Funds Obligated to Date: |
|
| History of Investigator: |
|
| Recipient Sponsored Research Office: |
4200 FIFTH AVENUE PITTSBURGH PA US 15260-0001 (412)624-7400 |
| Sponsor Congressional District: |
|
| Primary Place of Performance: |
123 University Place Pittsburgh PA US 15213-2303 |
| Primary Place of Performance Congressional District: |
|
| Unique Entity Identifier (UEI): |
|
| Parent UEI: |
|
| NSF Program(s): | FET-Fndtns of Emerging Tech |
| Primary Program Source: |
|
| Program Reference Code(s): |
|
| Program Element Code(s): |
|
| Award Agency Code: | 4900 |
| Fund Agency Code: | 4900 |
| Assistance Listing Number(s): | 47.070 |
ABSTRACT
With the emergence of social media and high-definition video streaming, there is a growing need for a more efficient way to process streams of visual information in terms of both bandwidth and energy. Currently, conventional image sensors record visual information frame by frame, unnecessarily acquiring huge amounts of redundant data since most pixels often may not change from one frame to the next. Inspired by the human brain, this project will develop a neuromorphic vision system, which is driven by the timings of changes in the dynamics of the input signal instead of the conventional image-based stroboscopic acquisition. This work will lead to transformative advances in bio-inspired neuromorphic processing architectures, sensing, with major applications in self-driving vehicles, neural prosthetics, robotics, and general artificial intelligence. The project team will work closely with local communities to encourage participation by students from all backgrounds including underrepresented group in computing careers by fostering interest in neuromorphic computing and artificial intelligence through outreach activities including lab demonstrations, summer internships, and career workshops.
The objective of this project is to build a brain-inspired vision system by integrating a neuromorphic, event-based silicon retina with a spiking neural network (SNN). In most existing neuromorphic vision systems, the communication between the event-based camera and the computing system is still limited by the memory bottleneck, largely negating the benefits of the large bandwidth and low power consumption of the neuromorphic camera. This project will: (1) build a spiking neural network with realistic graphene-based dynamic synapses allowing advanced computational capabilities; (2) develop a brain-inspired machine learning and general computation capabilities; (3) connect the developed hardware with a neuromorphic event-based silicon retina to demonstrate real-time operating vision system with orders of magnitude better energy efficiency and bandwidth.
This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.
PUBLICATIONS PRODUCED AS A RESULT OF THIS RESEARCH
Note:
When clicking on a Digital Object Identifier (DOI) number, you will be taken to an external site maintained by the publisher. Some full text articles may not yet be available without a charge during the embargo (administrative interval).
Some links on this page may take you to non-federal websites. Their policies may differ from this site.
PROJECT OUTCOMES REPORT
Disclaimer
This Project Outcomes Report for the General Public is displayed verbatim as submitted by the Principal Investigator (PI) for this award. Any opinions, findings, and conclusions or recommendations expressed in this Report are those of the PI and do not necessarily reflect the views of the National Science Foundation; NSF has not approved or endorsed its content.
Machine learning and neural networks are fundamental technologies driving progress in several industries, from information technologies to medical applications and robotics. One problem with these technologies is that they require powerful computers to run, making them expensive to deploy and energy inefficient.
This is especially true for artificial visual systems that collect information from cameras and have to recognize the scene in a fraction of a second to drive autonomous robots and self-driving cars.
Humans are naturally more efficient at these tasks than robots. For this reason, a great effort has been spent to mimic and understand biological systems. This effort resulted in the development of artificial retinas, cochleas, and computers that could simulate brain neurons. This is the field of neuromorphic computing.
One difficulty in making computers that work like the brain is that transistors (the basic building block of computers) are fundamentally different than neurons.
In this project, the investigator used a novel device called a "memristor" that acts like synapses, the fundamental connections between neurons in the brain.
This device demonstrated many synaptic properties, like learning and storing information long-term and dynamically responding to voltage pulses. In the first iteration, the investigators built it using graphene, a material made using a single layer of carbon atoms. The investigator also built and tested different devices with other materials like tungsten oxide and produced a novel device with higher power efficiency and speed of operation.
The investigators then created a mathematical model to simulate large numbers of these devices. Using this model, the investigators tested a spiking neural network with an array of 32x32 synapses. They demonstrated that it could learn to recognize objects recorded with an artificial retina.
This network was the first of this kind to use these devices' dynamic responses. Previous solutions required complicated circuits to simulate synapses that required multiple transistors.
This work demonstrated the use of these devices as a comprehensive solution for simulating biological synapses for developing more compact and efficient visual systems for robotics, autonomous cars, and retina prosthetics.
Last Modified: 10/30/2022
Modified by: Feng Xiong
Please report errors in award information by writing to: awardsearch@nsf.gov.
