Everyone wants to know when fully self-driving cars will be ready for the road. The companies developing the technology usually give a timeline of several years to decades away.
Delphi and Mobileye are no different, and during a media roundtable this week in San Francisco Delphi’s VP of services Glen De Vos envisioned a long-term timeframe for the widespread use of the technology. For example, DeVos expects that fully autonomous driving will first be deployed on commercial vehicles since businesses will be able to save on the expense of drivers and the downtime they require.
But at the same time, Delphi also announced this week that it’s partnering with Mobileye, the dominant automotive camera supplier, to fast-track getting self-driving technology into production by 2019 via a system the partners call Centralized Sensing Localization and Planning (CSLP).
CSLP uses an array of sensors to allow a vehicle to know its location within 10 centimeters, even without GPS connectivity. This helps a car navigate complex intersections even when there are no lane markings. It also identifies vehicles by their basic shape, senses whether another car is stationary or parked, and has what Delphi calls “semantic understanding” to predict the path of other vehicles so that a self-driving car can “behave more human-like in its driving behavior and determine the best path forward.
A crucial aspect of the CSLP system is circumventing what Delphi and Mobileye call the “map-heavy” approach used by companies like Google and many automakers that allows a self-driving vehicle to not only know its location but also its surroundings in order to achieve full autonomy. Instead, Delphi and Mobileye’s CSLP system will employ a “map-lite” approach by leveraging a technology Mobileye introduced at CES 2016 called Road Experience Management (REM).
Autonomous vehicles that heavily rely on detailed maps require a robust data connection to constantly feed accurate mapping data to the car. But the REM system uploads data in small bursts that can be easily handled by the 4G LTE connectivity already found in many vehicles.
Another benefit of REM technology is it can be “seamlessly integrated with existing vehicle platforms,” De Vos mentioned during the media roundtable. But perhaps the primary advantage of the REM technology is that it will essentially take real-time pictures of roads.
Mapping software can be out of date due to changes to the streets made after the maps were created or temporary conditions such as road construction. The CSLP system will not only capture and take into account permanent infrastructure ranging from intersections to road signs, but will also record construction and other short-term changes to the roadway. And this data can be crowd-sourced among millions of cars equipped with the system.
I had chance to see the CSLP system in action on the streets of Mountain View, California near Delphi’s Silicon Valley Labs, where its being tested in addition to Pittsburgh and in Singapore. And I couldn’t help compare the drive to one I took just over a year ago in the same location in one of Google’s autonomous Lexus RX 350s.
But unlike Google vehicles, one of which I spotted on the street during our drive, there are no noticeable sensors on the Delphi Audi, unless you look very closely. Compared to a self-driving Google Lexus, the Delphi Audi handled situations such as lane closures and stopped vehicles in adjacent lanes without hesitation, although the Delphi test driver did have to take over the controls twice: When we needed to speed up to make a right turn ahead of a city bus, and after we made the turn and got stuck behind a garbage truck.
Regardless, I was very impressed with the system, and was surprised to find out that it was operating without the benefit of Mobileye’s REM technology. This means it will only get better, and that self-driving cars could be available to the public much sooner than we think.
The post How Delphi and Mobileye Are Fast-Tracking Autonomous Car Technology appeared first on C3 Report – Connected Car Council and Conferences.