Toyota's Dual Tracks Toward AV Development
April 21, 2021—During a presentation on April 13, Ryan Eustice explained how the Toyota Research Institute is developing two automation suites for vehicles with the same goal: Jidoka.
Jidoka is a Japanese term that emphasizes automation and efficiency, but with a human touch. That means that even as automation increases, it becomes an enhancement of human work rather than a displacement.
Eustice is senior vice President of automated driving at the Toyota Research Institute, and he's also the director of perceptual robotics at the University of Michigan. He spoke as part of the university's Center for Connected and Automated Transportation's 2021 Global Symposium.
“In the automotive driving space, we think about jadoka as well," he said.
As he works on the R and D segment for Toyota, he says that they're approaching automation on two tracks with the same ultimate goal of automation that enhances the occupant experience.
The Toyota Research Institute has training grounds and teams of researchers at its disposal from multiple campuses. And the work is promising. Eustice says that autonomous vehicle technology has come a long way since the DARPA Grand Challenge, when U.S. military interests asked researchers to compete in autonomous driving courses. Eustice took part in the 2007 challenge.
Many challenges remain, however. It's that understanding of nuance that humans have when driving that is tough to replicate in computer systems, as ADAPT discussed in an article about sensor perception. Eustice says that we might want an autonomous vehicle to stop for a crossing guard holding up a hand in the road. But can that perception be fooled?
“That kind of contextual reasoning is really really hard," he said. "We don't want a vehicle to stop for a teenager holding up their hand on the road.”
Through phases of computer simulation, closed course testing and public road testing, Eustice and his colleagues are working on two phases o automated systems: Guardian Mode and Chauffer Mode.
Guardian Mode has its roots in the ADAS feature suites that we're familiar with today. It's always engages, monitors the situation and intervenes when necessary, like when a car drifts out of the lane. One of the top goals for Guardian Mode is to make the automated input seamless with the human driver. You don't want the steering wheel jerked out of your hand when the computer needs to correct course.
With steer-by-wire systems, for example, the computer should be able to correct course in a way that allows the human to maintain a smooth experience as well.
“They can make that large lateral move while accounting for other objects in the environment’," Eustice said.
The second track is more like full automation. Eustice said that it's called Chauffer Mode, and that's when the occupant is merely along for the ride. That vision is farther out into the future, but Eustice says that TRI is working to develop both modes on the same technological framework, which makes scaleability and upgrades easier.
No matter the mode, the underlying result should involve an element of Jidoka. It's about creating a better transportation ecosystem for people.
Image: Toyota Research Institute