Todd Coleman leads project to enhance current computer networking technology
When UI researcher Todd Coleman searched for a way to enhance communication between computer networks, he turned to the world’s most efficient communications
"During my post-doctorate work in neuroscience, I studied the probabilistic structure of how neurons in the brain communicate based on spikes," said Coleman, an associate professor of electrical and computer engineering and resident researcher in the Coordinated Science Laboratory. "I discovered that you could use the same theory and apply it to manmade systems."
Coleman recently received a $357,500 grant from the Air Force Office of Scientific Research to study enhancing communication between networks based on his neuroscience findings. The project will examine how networks can convey information not only through packets, which carry information between computers, but through the timing of their arrivals.
Through his research, Coleman will develop practical codes to enhance current computer networking technology. The project will also use sophisticated statistical analysis tools, including point process theory, nonparametric statistical estimation and convex optimization to perform network anomaly detection and other network inference methods.
"It will potentially provide another degree of freedom for people to manage communication networks and for protocols to efficiently control networks," Coleman said. "If you’re trying to go from point A to point B on the freeway, for example, you’d have two lanes instead of one."
Coleman’s research builds on findings outlined in "Bits through Queues" published by Verdu & Anantharam. While those results were theoretical, Coleman’s research will move the concept into the practical realm.
The concept will apply to both wired and wireless networks. Nodes on land and sky will be able to communicate with each other. The technology would improve throughput delay trade-offs that engineers currently have to consider. For the same number of bits per second, computers would on average move bits from one place to another in a more efficient manner.
The research will have applications in defense and security.
"This has been an untapped area, mostly because people didn’t know how to do it," Coleman said. "By understanding the probabilistic dynamics of router queues in networks, one can develop simple, structured graphical interpretations of those dynamics that lead to low-complexity iterative inference algorithms."