Rendering of autonomous vehicle accident-avoidance scenario. Utopia is one of those goals often dreamed of, but rarely achieved. But in attempting to create the technology aimed at achieving autonomous vehicles that can “think” like humans, Israeli tech company Mobileye is chasing that dream. Erez Dagan, Executive Vice President Products and Strategy, Mobileye, speaking at 2019 Society of Automotive Engineers World Congress Experience in Detroit, April 10, 2019 Speaking at this week’s Society of Automotive Engineers World Congress Experience in Detroit, Mobileye Executive Vice President Products and Strategy Erez Dagan laid out the path to reaching that level of artificial intelligence and the challenges impeding the journey.
While automakers and their technology partners are pouring billions of dollars and intellectual resources into developing so-called Level 5, or completely self-driving vehicles, Dagan argues those vehicles will never be adequately safe until they can do something only humans have the capacity to do–anticipate and react to unusual circumstances, declaring, “when it comes to safety, machine learning is not the tool.”
He calls that ability buying into a “social contract” where the self-driving vehicle goes about its business without causing trouble.
“In order to follow this social contract we need AVs to follow some digital interpretation of it, which is explicit, concise, metric, machine interpretable and is of traceable logic,” he said. “It has to be sound. It has to organically fit with common sense interpretation of humans of this contract, and it has to be pragmatic, It cannot be overly conservative and harm the traffic flow.” YOU MAY ALSO LIKE
Mobileye’s “holistic safety concept for autonomous vehicles is titled Responsibility Sensitive Safety, or RSS. Its five principles are:
1-Preserve a safety corridor, allowing you to never hit someone from behind.
2-Respect others safety corridors. Do not cut in recklessly
3-Right of way. It’s given, not taken.
4-Visibility awareness: drive cautiously in areas of limited visibility or occlusions.
5-Avoidance may not become a cause. If you can avoid a crash without causing another, you must.
That’s Mobileye’s Utopia in a nutshell, but to get there, Dagan says application of RSS’s goals needs to begin with technology available today called advanced driver assistance systems or ADAS. Those are things like adaptive cruise control, lane departure and collision avoidance systems. Mobileye is engaged in a crowdsourced, global mapping project to gather data that can be used in autonomous vehicles to help avoid accidents. Mobileye
Mobileye is in the midst of a crowdsourced global mapping project using cameras in ADAS-equipped vehicles to provide valuable intelligence about driving conditions around the world.
“Our crowdsourced road book provides us with foresight of upcoming negotiation points, lanes topology and lanes priority,” he said “Both these things gives the vehicle kind of super human perception. If there is a right turn that commonly invokes emergency braking, you want to know that.”
That information is being used to reach what Dagan calls Level 2-plus autonomy, explaining, “it’s kind of the first lovechild of ADAS and autonomous driving development.
based on our crowd sourced mapping venture. It allows us to deliver what we call pseudo autonomy. Level 2-plus is a vehicle that doesn’t change the liability structure. The driver is still responsible to monitor and be available. His mind is on the drive but we can get a very expressive and very rich envelope of operation based on technologies trickled down from autonomous driving. If we trickle down RSS to advanced driver assistance vehicles, we will get the revolution in saving lives.”
With Level 2-plus as a stepping stone to full autonomy, the goals become more steep such as developing the ability for a vehicle to not only react to dangerous situations but to help prevent them from occurring at all, said Dagan.
The road to autonomous Utopia is fraught with challenges, but Dagan believes that destination will be reachable if the industry can just “digitize the principles of cautious driving.”