Paris Auto Show
Safety Is a Priority with the Autonomous Vehicle, Says Mobileye Vice President
Lior Sethon spoke Tuesday at a conference on innovation held by Calcalist at the Mondial de l'Automobile auto show in Paris
15:0603.10.18
90% of car accidents can be attributed to the human factor, according to Lior Sethon, Mobileye vice president and its deputy general manager of the aftermarket division. Sethon spoke on Tuesday at a conference on innovation held jointly by Calcalist and events firm Connecting Leaders Club at the Mondial de l'Automobile auto show in Paris. Once the fully autonomous car hits the market, he explained, accidents will go down by 90%, maybe more.
Jerusalem-headquartered Mobileye develops sensor-based driving assistance systems using artificial vision technology. The company's systems are installed in over 27 million vehicles, according to Sethon, with over 25 OEMs using them. In 2017, Intel paid $15.3 billion for the company, turning it into its global center for activity in the field of autonomous cars.
"Our approach is to have the technological building blocks in place first," Sethon explained. "It has to be scalable economically, and we need to formulate safety—the industry needs to create transparency for the end users." Mobileye thinks the primary sensor of any self-driving vehicle should be the vision sensor, he added, coupled with different types of sensors such as radar and LiDAR added for redundancy, and a very strong recognition algorithm.
Sethon also touched on Mobileye’s current project of collecting data for HD maps. "For autonomous vehicles, you need to have very precise maps for the road. It's another redundancy, and also a layer of memory for the vehicle, enabling the creation of localization and planning for drives." The data collected from the vehicles and anonymized could later be sent to and used by cities, governments, and autonomous fleets, he explained.
Another part is teaching the software to deal with humans on the road. "Drivers use their intuition and are constantly negotiating. And this is what we are teaching the software–how to negotiate with human drivers."
