Uber's New Female Executive is Defining the Future of Driverless Car Technology

While driverless cars may not be ready to for mass use just yet, tech limitations won’t prevent them from reaching that point…

By now, it’s becoming very clear that the technology behind driverless cars is viable, and constantly improving. While they may not be ready to for mass use just yet, tech limitations won’t prevent them from reaching that point.

However, there’s reason to believe that price may. Manufacturing a driverless car is expensive. For instance, virtually all autonomous cars rely on technology called Lidar. It’s a laser-based system that allows cars to “see” the world around them via principles similar to those behind radar. Ideally, this will cut down on human error and lead to many fewer car accident injuries and on-road collisions. Unfortunately, high-end units can go for as much as $85,000 in the US. Equipping a fleet of vehicles with them is a costly task.

That being said, there are some who believe the best way to reduce costs is to simply break away from Lidar technology altogether. Raquel Urtasun, an A.I. expert who recently joined Uber to lead a research facility, suggests equipping driverless cars with cameras, instead of Lidar. Her team is transforming what the future of driverless cars could look like, and here’s how.

Less Expensive Lidar on the Horizon

Luckily, that may not always be the case. Back in the 1990s, innovators quickly developed affordable supercomputer tech in order to build stronger, more affordable PCs. Right now, startups are striving to do the same for Lidar.

One company, Ouster, recently introduced its own alternative to the premium Lidar systems currently on the market for driverless cars. The company claims its product performs at almost the same level of reliably that the high-end models do. It also costs a mere $12,000. If the claims are true, this product could substantially reduce the costs involved in manufacturing autonomous vehicles.

New Technologies Might Clear the Way

At its early stage, the results from Raquel Urtasun’s team have been promising. Urtasun believes that ordinary cameras may be able to provide essentially the same degree of high-quality 3D imaging that Lidar sensors provide. She demonstrated early results at a New York computer-vision conference, showing how regular cameras can be used to generate stereo images.

Running in real-time, the system would theoretically compete with Lidar within 40 meters of the vehicle. While these early results don’t quite match Lidar in some areas – Lidar typically has a stronger range of vision – they do suggest an alternative that would be much more affordable if implemented.

While some startups continue to pursue low-cost Lidar solutions, Urtasun maintains that making autonomous cars available on the mass market won’t happen until the industry as a whole embraces a less-expensive option. As she says, “If you want to build a reliable self-driving car right now we should be using all possible sensors. Longer term the questions is how can we build a fleet of self-driving cars that are not expensive.”

Prior to her work with Uber, Urtasun has been involved in related projects that may lay the groundwork for a camera-based autonomous vehicle sensor. Her lab at the University of Toronto has developed software which can creates maps of roads, sidewalks, and similar features, using only aerial and ground-level pictures. If Urtasun and her team are given the time and resources necessary to develop this technology further, it may replace Lidar as the go-top sensor for autonomous cars in the future.

Of course, there will be those who disagree, believing Lidar is simply more reliable. As of now, it’s not clear who is right. What is obvious, however, is the fact that many experts like Urtasun are working diligently to ensure autonomous vehicles will soon be much more affordable than they have been in the past. If these problems are solved, driverless cars may be on the road sooner than you expect.

Written by Catherine Metcalf