Tech Transfer Spotlights
Soar Technology
The idea of machines endowed with
human-like intelligence has been around for decades, but the concept is
reaching new heights, literally, through the work of UM Professor of
Engineering John Laird and his USC colleague Paul Rosenbloom. As
students at Carnegie Mellon, the two engineers developed an early
version of Soar, a computer language for modeling a full range of human
mental processes. In the 1990s, the technology attracted sizable
Department of Defense (DOD) research grants, many of which were
channeled to the UM Artificial Intelligence (AI) Lab, where Laird served
as director.
It was this government funding that
enabled Laird, Rosenbloom, and a team of UM and USC engineering
researchers to develop TacAirSoar, an agent-based simulation system.
According to Soar Technology CEO James Rosbe, "TacAirSoar 'agents' are
synthetic pilots that fly aircraft in simulations at the same level of
skill as an expert human pilot, making human-like decisions on
maneuvers, and communicating and reacting to changing events--adaptively
and intelligently."
Soar Technology was established in 1998
to support and expand DOD applications of intelligent autonomous agent
technology beyond what was possible within the University. Because of
the terms of its unique government research funding, Soar Technology did
not require a formal license for the DOD market. Even so, Tech Transfer
assisted in a variety of ways: coaching the founders before and after
start-up, providing an MBA student intern to conduct market research,
and facilitating license negotiations.
Now, however, the company is beginning to
broaden its focus to address non-DOD markets and, in 2003, became an
official UM start-up, having licensed TacAirSoar and two other
technologies from the UM AI Lab (EPIC and GLEAN, developed by Professor
David Kieras). Rosbe notes that, with the technology access afforded by
the UM license, the company plans to diversify into intelligent agent
applications in medical research, health care and logistics.
Copyright, Regents of the University of
Michigan
|