The latest news from academia, regulators
research labs and other things of interest
Posted: Apr 08, 2011
DARPA works to build electronic neural architectures that can learn, adapt, and respond
(Nanowerk News) 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."