
Alex K. Jones, the Department Chair of Electrical Engineering and Computer Science and Klaus Schroder Endowed Professor for Engineering, has received National Science Foundation (NSF) Early-concept Grants for Exploratory Research (EAGER) award for computer memory research. The award supports early-stage research with promising ideas involving innovative approaches, new expertise, and interdisciplinary educational opportunities. Jones serves as the project’s principal investigator.
Computers in space face challenges, particularly with radiation, which can disrupt electronic parts. Most space technologies use radiation hardening, a process that designs electronics to withstand the damaging effects of radiation. However, these parts are significantly more expensive than their commercial counterparts and are often generations behind the latest technologies.
“It’s like being forced to use a flip-phone because a radiation-hardened smart phone hasn’t been designed yet,” says Jones. “I would much rather try to figure out, using software-based fault tolerance, how to use the commercial smart phone than live with the flip phone.”
The project’s overall goal is to make commercial computer memory technologies more reliable and better suited for the harsh conditions of space. First, the project aims to understand how dynamic random-access memory (DRAM), the most common type of memory used in modern computers, behaves in space-like conditions. Second, the project will develop techniques to protect DRAM during in-memory computing, a method where data is processed and stored in a computer’s main memory.
“Normally, radiation is expected to randomly disrupt electronics. In our experiments, we found that memory errors are concentrated on 1.5 ten millionths of a percent of the memory capacity,” says Jones. “We have also been demonstrating that we can use this same memory to do massively parallel processing for AI, so the question we want to answer is what happens when we do this AI processing in the presence of radiation.”
Computing systems in space require solar energy, which limits their power supply. Because of this, they cannot use hardware accelerators such as those found in data centers. However, these space systems still need memory. If Jones’ research is successful, it could make computing AI algorithms possible in space using the computer memory as an accelerator.
“As a kid, I was blown away by the almost magical computers and artificial intelligence in Star Trek,” says Jones. “Never did I imagine I’d be helping to make AI in space computing a reality.”