“My family environment was pretty STEM oriented.” This comment from Dr. Alexandre (Alex) Bayen may be understating things a bit. He grew up in Paris, France, with a mathematician as his father, physicist as his grandfather, a chemist as great grandfather, and neurophysicist as a grandmother. Alex spent his childhood rollerblading across the marble floors of the University where his father taught and visiting the electrostatic room that his grandfather designed in the Paris Science Museum (Palais de la Découverte), for which he served as the Director. Alex frequently heard math riddles from his father, made trips to the geology museum, and was surrounded by thousands of books on all topics at home.
Now Dr. Bayen is the Liao-Cho Professor in the Electrical Engineering and Computer Sciences Department at UC Berkeley. He leads the Transportation Initiative at Lawrence Berkeley National Laboratory (LBNL), as Faculty Scientist. Dr. Bayen’s work moves along (pun intended) transportation analysis through the tools of Artificial Intelligence – AI – to improve people’s daily commutes.
Traffic is a relatable woe to pretty much everyone all over the world. When you’re stuck behind a line of cars in a huge back up you may feel the urge to wave a magic wand and smooth out the traffic flow. But modeling and understanding traffic is incredibly complex. Current tools for transportation agencies, planners, and engineers struggle to result in the analysis needed. Alex and his team are working on creating that magic wand to understand and fix traffic issues through artificial intelligence, and its cousin deep learning, through their program called CIRCLES – Congestion Impact Reduction via CAV-in-the-loop Lagrangian Energy Smoothing.
CIRCLES taps into AI tools of deep reinforcement learning, where researchers program computers to learn the behavior of complex systems over time. Cloud computing enables the UC Berkeley and LBNL team to run giant simulations at scale and replicate an accurate simulation of traffic, seeing where energy is currently wasted and can be saved, using a software framework called FLOW that was developed by Bayen’s team of students and post-doctoral researchers.
The team believes that self-driving cars of the future can use this data and drive at the right places and speeds (likely slower) in order to smooth the flow of traffic. “This will be shocking to people initially but it will be the next generation of traffic control,” Dr. Bayen said.
AI makes this work possible. “AI has gone through winters, and now a rebirth,” Dr. Bayen said. “There will be a constant need for AI and a workforce for it.”
And how exactly did Dr. Bayen end up in this AI workforce? After his STEM-packed childhood in France, he got his undergraduate degree at Ecole Polytechnique in applied mathematics, from which he graduated as a Second Lieutenant. He was fascinated by what was happening in the United States’ Silicon Valley at the time – right before the dot com crash – and applied to join the fun by coming to Stanford to study aeronautics and astronautics and earn his M.Sc. and Ph.D degrees from the school. “There were things happening in U.S. campuses that there was no match for in Europe. It was a different order of magnitude and scale at the time,” he said.
At Stanford, Dr. Bayen was studying control theory, optimization, and machine learning. He focused on air traffic control applications, and served as a Visiting Researcher at NASA Ames Research Center from 2000-2003 after his graduation from Stanford. He then went back to France’s Department of Defense for a year, before leaving with the rank of Major. He took a position at UC Berkeley in 2005 and has been there ever since. He joined the Lawrence Berkeley National Laboratory staff in 2014 (many research staff at the Lab are UC Berkeley faculty).
For others looking to work in AI, whether it’s in deep neural networks, optimization, control theory, applied math or other layers of AI, Dr. Bayen recommends studying electrical engineering and computer sciences. Some schools have joint degrees for this purpose. A degree in data sciences is also a good bet to work in the field, he advises. Many people find the AI field through its mathematical approach, Dr. Bayen said, and there’s a great deal of statistics skills needed to work on AI.
Dr. Bayen wants to see the AI workforce reflect the diversity of the nation, ensuring that the work result isn’t unintentionally limited due to oversights from a lack of approaches. “I think that diverse teams produce more creative solutions because they look at a problem from different angles,” he said.
“I really care about diversity, inclusion, and equity,” said Dr. Bayen. He’s proud of his work at the entry to the AI workforce pipeline, as he’s overseen the establishment of the FIRST LEGO League Competition in Piedmont, California. He coached an all-girls team in the first year that made it to the finals and won Best Robot Award in the regionals. He’s even dipped into his discretionary funds to host a regional FIRST Robotics championship tournament at LBNL. He’s loved mentoring diverse teams and bringing in more girls and minority students into computing and machine learning and robotics.
“This experience supersedes everything else I’ve done professionally, with the enjoyment I’ve had to do these competitions,” he said.
Learn more about the Department of Energy’s work in AI and the AI workforce behind it here.