Bio-Inspired Semiconductor Technology

The U.S. military is dependent on electronics for situational awareness. Power consumption of electronics must be minimal for practical deployment. Much of the incoming video, acoustic, and sensory data is insignificant and does not need to be analyzed by computationally intensive algorithms operating on conventional processors. This project aims to develop an ultra-low-power parallel processor capable of processing video, acoustic, and sensory data and identifying when features of interest are present. The bio-inspired neuromorphic processor will serve as a decision processor to control the operational state of a multicore processor that is only powered on when interesting data are present. In Phase I, the architecture of the neuromorphic processor was defined, and the hardware and software tasks were partitioned. System design analysis and circuit simulation environments were put in place, and the nonvolatile memristors for use in the neuromorphic processor were characterized. Phase II will involve evaluation and specification of design elements and parameters, development of a prototype hardware system to demonstrate proof of concept, specification of a processor chip implementation, and development of software for programming of memristors, as well as ongoing improvements to memristor technology.