WCX 2019: Mobileye’s Quest for Super-Human Perception

April 10, 2019

Kami Buchholz

Ken Washington, SAE International’s Automotive Section vice president (left) hosted a question and answer session with Mobileye’s Erez Dagan after Dagan’s keynote speech. (Brooke)

Human drivers typically approach driving by following traffic rules and exercising a sense of caution in certain driving situations and road conditions. So, it makes sense that autonomous vehicles function by anticipating possible driving hazards and risks with so-called ‘super-human perception.’

“We asked ourselves how can we define a safety concept for autonomous driving that allows us to blend with human traffic, and to do that we had to model how humans deal with safety,” Erez Dagan, senior vice president and general manager of strategy, research & development at Mobileye, told Automotive Engineering following his April 10 keynote address at WCX 2019.

To a large extent, humans obey traffic rules and follow a framework of cautious driving. Human ‘negotiation’ in traffic, involving multiple inputs to make a decision—such as when merging from a blocked lane into a busy moving lane—“is far from flawless,” Dagan noted. That leaves drivers risk-exposed.

“Even if all agents comply with all of the traffic rules, there is still a risk of accidents,” Dagan said. There are underlying reasons why accidents can occur, ranging from drivers disobeying traffic laws and rules to lapses in judgement.

Mobileye’s development to creating a safer road environment for vehicles is RSS, a model-based and holistic technology for safe decision making. “The model gives us a very clear boundary line from what it is to be agile and what it is to drive dangerously. Once you have that boundary line, you can optimize that agility,” Dagan said.

Applying math-based models to autonomously driven vehicles is only part of the development landscape. Advanced Driver Assist Systems (ADAS) are part of the picture, and from Mobileye’s vantage point, so is crowd-source mapping. “We’re using the proliferation of millions of ADAS vehicles out there to aggregate and dynamically capture a high-definition map of the road,” Dagan said.

Those high-definition maps, using crowd-sourced data collected since 2014, can also reflect dynamic changes, such as a change in speed limit for a construction zone or pedestrian hotspots.