ATW #2524 – Using AI To Make ADAS More Effective
October 7th, 2021 at 11:00am
Listen to “Autoline This Week #2524 – Using AI To Make ADAS More Effective” on Spreaker.
Internet Premiere: 10/07 @ 11:00am ET
Detroit Public TV: 10/17 @ 10:30am ET
PANEL:
Jim Quesenberry, Director R&D, Magna International
Mark Crowley, University of Waterloo
Ross McKenzie, University of Waterloo
John McElroy, Autoline.tv
Automotive safety technology can prevent accidents and save lives. But it can also be maddeningly annoying. All those beeps, buzzes and warning chimes. So researchers are carefully collecting data on how humans drive, and using artificial intelligence to make those safety systems less intrusive and more effective.
Follow us on social media:
Thanks to our partner, the Consulate General of Canada in Detroit.
To learn more about Canada’s unparalleled environment to
develop the mobility technologies of the future,
click here & contact us at DTROTTD@international.gc.ca
Subscribe to the free podcast:
Share
Permalink
October 8th, 2021 at 5:58 pm
What a great/interesting program. Thanks so much for getting these gentlemen together to speak of their research.
October 9th, 2021 at 8:10 pm
The thought occurred to me while watching this discussion “I live in a right hand drive country – is this research “reversible”. By this I mean is the ADAS software completely able to adapt to RHD from an LHD base set of code, or is a separate code base required for the 25% of the world that is RHD.
As my American friends tell me, they drive on the right side of the road and I drive on the wrong side.
October 13th, 2021 at 12:07 pm
Yes it was a great discussion and John McElroy has such a good way of asking the right questions and keeping it fun.
To Warwick’s question, that’s a very interesting problem! Modern AI and Machine Learning algorithms don’t rely entirely on their “programming” because the program tells the computer how to learn what to do from data itself.
So, in order to solve the problem you raise, one way to do it is retrain the systems on data from drivers in countries with different rules of the road. Even more exciting, would be to experiment with “flipping” the AI’s model, and see how well it could do without ever having seen someone driving that way. This is kind of how human’s do it, with A LOT of common sense and other rules built in to our heads.