Electronic chip represents key building block of neuromorphic architecture |
Today’s
warfighters possess the ability to meet the dynamic demands of the
battlefield by relying on their knowledge and training to make the right
decisions in demanding complex situations. In contrast, unmanned
systems and electronic devices, while able to collect and process
information, are limited in their efficiency and flexibility, and
current computer systems can only process information according to their
programming. What if warfighters could access an entirely new class of
electronic systems that can meet the demands of dynamic environments.
DARPA’s
Systems of Neuromorphic Adaptive Plastic Scalable Electronics (SyNAPSE)
program aims to fundamentally alter conventional designs by developing
biological-scale neuromorphic electronic systems that mimic important
functions of a human brain. Applications for neuromorphic electronics
include robotic and manned systems, and sensory and integration
applications such as image processing.
The
goal of SyNAPSE is to create electronic systems, inspired by the human
brain, that can understand, adapt, and respond to information in
fundamentally different ways than traditional computers. While current
computers are organized into distinct processor and memory units that
function in accordance with their programming, the brain is organized as
an intimate and distributed web of very simple processors (neurons) and
memory (synapses) that spontaneously communicate and learn their
functions. Using knowledge of the brain’s organization as a platform,
SyNAPSE is developing integrated circuits with high densities of
electronic devices and integrated communication networks that
approximate the function and connectivity of neurons and synapses.
This
program has also developed tools to support this specific area of
hardware development such as circuit design tools, large-scale computer
simulations of hardware function, and virtual training environments that
can test and benchmark these systems.
The
first phase of SyNAPSE developed nanometer-scale synaptic components
capable of adapting the connection strength between two electronic
neurons, similar to what occurs in biological systems, and simulating
utility of these components in core microcircuits that support the
overall system architecture. The next phase specified large-scale system
architecture, simulated core dynamical behaviors of large networks, and
demonstrated microcircuits of electronic synapse and neurons that
spontaneously organize in simple environments.
“So
far, SyNAPSE has successfully demonstrated all the core hardware,
architecture, simulation, and evaluation capabilities needed for a new
generation of intelligent electronic machines,” said Todd Hylton,
DARPA’s SyNAPSE program manager.
Goals
for the upcoming phase include chip-fabrication process development,
design and validation of single-chip systems, and demonstration of these
systems in virtual environments that emphasize behavioral tasks related
to navigation and perception. “Now that all the building blocks are
available,” said Hylton, “our next task is to start building functioning
systems out of them.”