How Renault Is Learning from the Best Human Test Drivers to Create the World’s Safest Autonomous Cars
Today, we announced that we’ve developed an autonomous control system that can handle challenging driving scenarios and, as an industry first, can demonstrate road obstacle avoidance as well as professional test drivers.
Despite popular belief, the reality is that human beings are pretty amazing drivers, with less than 1 fatality per 100,000,000 kilometers in most developed countries. Reaching and exceeding that benchmark is essential to improve safety and realize our dreams of autonomous cars, providing more productivity during our morning commutes and robo-vehicle services in city centers. And to do so, we must learn from the best of the best.
Despite all the recent technological improvements, in some challenging scenarios humans can still compete with many of today’s autonomous driving systems. This is why we test against professional test drivers. To our engineers, they are an inspiration and a benchmark — teaching us new ways to think about a problem, and setting the bar on the performance we need to demonstrate.
While we can perform as well as these professional drivers today, we’re not content in stopping there. One truly fascinating thing is that, while our control system is currently matching their performance on our obstacle avoidance scenarios, we aren’t completing the evasive maneuvers in the same way. While our superior reaction time gives us an edge, we suspect we can do even better by working closely with these experts.
Since 2014, we’ve worked with Professor Chris Gerdes, the former Chief Innovation Officer of the U.S. Department of Transportation, and his team at Stanford University. They’ve helped accelerate our understanding of these difficult problems, and together we’ll continue enhancing autonomous driving safety-critical performance even further.
The technology was originally developed on “Callie,” Renault’s autonomous Zoe prototype in Silicon Valley. But thanks to our strong ties to France and Japan, it can help accelerate deployment for several products in Renault-Nissan-Mitsubishi. That means that the 15 Renault models and 40 Alliance vehicle models with different levels of autonomous driving capabilities could benefit from this solution, with eyes-off / mind-off technology or robo-vehicle capabilities.