Nvidia Takes Assisted Driving to Market With DRIVE AutoPilot
Not content with being the computer and toolkit supplier to the auto industry, Nvidia has announced a full Level 2+ ADAS system, DRIVE AutoPilot, that it will make available to automotive suppliers and car companies. Don’t run off to your dealer to buy one just yet, but Nvidia predicts it will be available in production vehicles by 2020. The system makes a lot of promises, which we’ll dive into here. But I expect that in reality the full set of features will roll out over time, and some will take longer to finalize than Nvidia’s 2020 target.
Under The Hood of DRIVE AutoPilot
AutoPilot relies on Nvidia’s DRIVE AGX Xavier SoC and its DRIVE software platform. It also includes DRIVE IX to provide what Nvidia calls an intelligent driver experience. Nvidia says the AutoPilot hardware can crank through 30 trillion operations per second while consuming only 30 watts, although in practice Nvidia’s TOPS numbers don’t always match actual use cases, so we won’t know how well it performs until it ships.
AutoPilot is positioned as a lower-end alternative to DRIVE AGX Pegasus, which is being used by many companies in their Level 5 autonomous vehicle testing efforts. This makes a great deal of sense since Level 5 autonomy seems to be getting further away from large-scale deployment instead of closer, while ADAS and similar systems continue to improve and gain market share. That makes them the business opportunity for the next few years.
On the software side, the offering gives OEMs a chance to mix and match components of Nvidia’s Drive software stack. It includes features like traffic light detection and sign recognition, as well as a driver-facing interface designed to both enforce driver attention requirements and to provide the driver with a desirable cockpit experience.
A Clever First: DIY Mapping
I’m taking this claim with a grain of salt, but Nvidia says that instead of only operating in pre-mapped areas (the way Cadillac’s SuperCruise does, and most robo-taxi test vehicles do, for example), AutoPilot will be able to map and memorize a route you drive, so that it can drive it again for you. I can see that on highways or restricted access roads, but conventional wisdom for city driving with accurate behavior around pedestrians, you need 2.5cm location accuracy, correlated with HD maps. (I saw the benefits of that on the self-driving demo I got from Aptiv at the show. More on that coming in a separate article.)
DRIVE AutoPilot Advantages
For this to work, Nvidia needs to convince Tier 1 suppliers and auto companies to upgrade from their existing electronic control units to the Xavier SoC, which Nvidia says is more powerful and flexible, but of course, will also cost more. OEMs that adopt AutoPilot should also get two additional benefits. First, the SoC can be upgraded to Pegasus as capabilities improve. Second, the system can be upgraded over the air (OTA), the way Tesla does today, but that almost no other car company has pulled off. Assuming the sensor package on the vehicle is adequate (a big if), that should mean that after-market upgrades to higher levels of autonomy should be possible. That will be a very big selling point for car buyers today who basically have to buy a whole new car to get new assisted-driving features.
Continue reading
Perseverance Rover Records the Sounds of Driving on Mars
This is the first Mars rover with a microphone, and it captured the eerie sounds of Perseverance driving over the regolith during one of NASA's tests.
Chia Demand Is Driving HDD Sales, Keeping Seagate’s Factories Full
Western Digital and Seagate are seeing sharply increased demand thanks to cryptocurrency, and they're both pretty happy about it.
Tesla Built a Supercomputer to Develop Camera-Only Self-Driving Tech
Tesla is talking about what it sees as the next leap in autonomous driving that could do away with lidar and radar, leaving self-driving cars to get around with regular optical cameras only.
Tesla Rolls Out $200 Monthly Subscription for ‘Full Self-Driving’
Some vague language on Tesla's part means that vehicles marketed as having full Autopilot capabilities might need an additional $1,500 hardware upgrade to use FSD.