Auto Newbies Finally Accept Driverless-Only Roads

Automotive disrupters who once predicted the imminent adoption of autonomous technology are now considering geo-fencing driverless cars as automakers predicted several years ago.

The latest to come to this conclusion is Chris Heiser, CEO of Renovo Auto, a company working on autonomous vehicle technology, who told TU-Automotive that while Level 4 cars on general roads are still a decade or more away, yet by limiting the scope to pre-defined, geo-fenced areas, Level 4 cars are possible today.

His comments come more than four years after Honda UK’s former CEO Philip Crossman told us dedicated driverless highways is the best way to get the technology adopted sooner rather than later.

Heiser, said: “I think the trends that we’re seeing now is that Level 4, depending on how you define it, is getting further and further away. But we’re actually seeing it get closer if you limit the scope of what you try and do with that level, limiting it to pre-defined areas and low speed environments.”

Level 4 vehicles are the long-term aim for but it’s not easy to design a system which can handle the edge cases and variables that driving a car brings. “One of the current challenges is the sensing system. For example, if you change the exposure of a camera and have a look at the same object. Does the machine vision system think that’s the same object now when it’s darker, or lighter, or more blue, or more red? Where the humans kind of have an advantage is that we have these kind of layers of perception, we have our visual systems where you can look at it and say ‘that is a bicycle, not a shrub’. We also have lower level systems that say, hey, that’s a thing, let’s not hit it. That’s the abstract way we look at the world, and we’re not there with AI yet,” said Heiser.

By limiting the edge cases and variables, Level 4 vehicles could be possible. “Level 4 is really complex; we’ve been working in this space for a while now and it’s a hard problem, with lots of moving parts and unpredictable variables, especially in urban environments. On the flip side, you have these great applications like low speed, fixed geo fence areas, campuses, retirement communities, military bases, mining, all of these things are much simpler from what’s called an ODD – operational design domain. And as you constrain that ODD, the complexity of the system goes down. That makes it possible to deploy Level 4 vehicles today. We’re working with a company, named Voyage, which is deploying ‘robot taxis’ to retirement communities. Very low speed, reasonably simple environments. It’s allowing Voyage to get to production quicker and gather more relevant data than someone trying to do taxis in downtown Los Angeles or something,” Heiser told us.


Leave a comment

Your email address will not be published. Required fields are marked *