LAS VEGAS — Aptiv, the spinoff from Delphi dedicated to autonomous cars, is back at CES with a fleet of cars able to navigate through Las Vegas’ busy streets with virtually no driver involvement on public roads. Call for a ride from Lyft and you may even get an an autonomous Aptiv car picking you up, albeit with a driver on board.
In a half-hour ride in one of Aptiv/Lyft cars, it dealt with traffic lights, slower and faster cars nearby, lane changes, right and left turns, jaywalking pedestrians, and faded lane markings. Only once did the driver take over, and that was to steer around pylons in the middle of the road. This is progress.
The side of the car has short- and long-range lidar plus radar. Long-rang lidar is near the side radar, so each sensor sees almost the same thing.
Secret Sauce: Sensors Up the Wazoo, Smart Software Stack
Aptiv has been outfitting several different vehicles with its software stack and lots of sensors. According to James Ziselman, VP for engineering and program management at Aptiv, “The vehicle has 21 sensors: 9 lidar, 10 radars, and 2 cameras in front. One camera is traffic recognition, one is tri-focal … three different focal points for for the camera. All those sensors surround the vehicle for all three sensor types, 360 degrees, for all of those sensor types.
“All that data comes in and is fused, because sometimes one sensor says something and the other doesn’t,” Ziselman says. “It goes through an algorithm and some artificial intelligence to decide, ‘What is that? And what are we going to do about it?’”
Ziselman says the present-day cost of the sensors and controllers is immaterial. He fields that question a lot and says all that matters is that by the time a salable autonomous car reaches the market, the sensors will be vastly cheaper. He says the rotating mechanical lidar sensors will give way to far cheaper solid-state sensors on a chip.
The front of the car has dual long-range lidar in the grilles, short range lidar below that, plus radar, plus cameras behind the glass.
Driver Must Drive on Private Property
For what seems the safest part of the journey, pulling out of the autonomous vehicle staging area across the street from the Las Vegas Convention Center, the driver must be in charge. Nevada’s rules for autonomous cars limit the autonomous driving/testing part to public roads. Once the driver crosses the sidewalk and turns onto Paradise Road, the car takes over.
Our test vehicle is an alpine white BMW 5 Series with bright orange five-spoke alloy wheels, Aptiv markings on the flanks, and a Lyft sticker in the front windshield. There’s an LCD for the rear passengers to confirm the ride when it’s being used as a Lyft vehicle. The big center stack screen in front has an engineering view overlaying radar and lidar mapping and obstructions. The car announces autonomous mode as well as lane changes and turns.
The Aptiv/Lift BMW is hands-off on public roads, but a driver has to be behind the wheel: Nevada rules. He’s not even allowed to talk to road-test passengers in the back until the car is off public streets.
Big Progress in a Year
A year ago Delphi (and others) showed autonomous cars that cautiously navigated Las Vegas streets. Some could and some couldn’t make left turns. Some only worked on highways, some only worked on local streets. A lot has happened since then. The car seems surer of itself, if you were to give it human attributes. With the 2018 Aptiv autonomous-drive tools, you can see how we’re on a glide slope that would provide Level 4, possibly Level 5 autonomy sometime around 2020-2021. Level 5 is autonomous drive everywhere; Level 4 might call for driver intervention, but only with a good long warning period. The most common example would be where a car drives you 200 miles into ski country, but you’ll need to drive the last 20 miles to the condo once you’re on rural roads.
BMW’s center stack LCD shows the lidar and radar mapping of surroundings.
Delphi Splits in Half
Last fall, Delphi split into two companies. In part, this makes Aptiv more of a tech stock. Aptiv is focused on new mobility solutions, smart vehicle architecture, connected cars, and connected cities. To compete against other tech companies in the new-transportation sector, it has armored itself via the academic world. It acquired Ottomatika, a spinoff from Pittsburgh’s Carnegie Mellon engineering research group, for its software expertise in the driverless-car decision-making realm. Then it acquired nuTonomy, a spinoff from MIT, for its software expertise in the process of driving the car.
The automotive powertrain, advanced propulsion, and aftermarket solutions provider part of the old company retains the Delphi name.