We’ve covered how Tesla’s approach to self-driving cars differs from Google’s. The former is incrementally introducing more advanced autonomous features to eventually get to full autonomy, while the latter thinks it is safer to make the jump to completely autonomous driving once the technology is at maturity.
The two companies also have different approaches to hardware. Google relies on LiDAR, while Tesla thinks it’s unnecessary and that a combination of cameras and radars should do the job. Only time will tell which strategy will turn out to be the most successful, but in the meantime, Tesla shared a rare data point that gives us some perspective on the two programs.
Unlike Tesla, Google shares monthly reports on its self-driving car program (please Tesla, take example). The company releases its cumulative miles driven on autonomous and manual modes, as well as accident reports (when there are any) and sometimes they write about new technologies they are implementing.
We often share those reports on Electrek and our sister-site 9to5Google.
Our colleague Stephen Hall keeps track of the autonomous miles driven by Google’s SDCs:
As you can see, it recently reached the 1.5 million miles mark last month. That’s since the beginning of the program in 2009.
While Tesla doesn’t share similar data for its Autopilot program, for some reason, over the weekend the company decided to release the total miles driven with Autopilot by its fleet since its introduction in October 2015.
[tweet https://twitter.com/TeslaMotors/status/718845153318834176 align=’center’]
The company confirmed that 47 million miles were driven while the Autopilot was activated. When first introducing the new features, CEO Elon Musk said that Tesla’s fleet was adding about 1 million miles of data every day, which indicates that the company might be pulling data even when the system is not actively controlling the car.
Of course, Google’s miles were city miles, while the Autopilot is for highway driving which gives a distinct advantage to Tesla in term of racking up data in miles. Although Google is lagging behind for real-world miles, the company is quick to add “simulated-miles” with over 3 millions a day, without leaving the lab.
They also pull data differently based on their hardware (LiDAR vs Cameras), but the main difference is that Tesla has about 70,000 vehicles equipped with Autopilot hardware on the roads, while Google currently operates only 54 prototypes. Google uses both its cute prototypes (picture above) and Lexus RX450h SUVs equipped with Google’s system. The vehicles are located in Mountain View, CA, Austin, TX, and Kirkland, WA. While Tesla’s fleet is spread out across the world.
It certainly makes for an interesting race to fully autonomous vehicles. Last year, Elon Musk said that he expects Tesla’s level-4 fully self-driving technology to be ready in about 2 years, while Google is vague about its timeline to bring its system to market, but it expects the cars will be on the road by 2020.
The Sooner the better. Just this weekend, we watched a Model S with Autopilot autonomously avoid a collision with a truck [Video].
FTC: We use income earning auto affiliate links. More.
While I’m a huge Tesla supporter, you can’t compare Tesla’s Autopilot to Google’s self driving cars. They’re two completely different technologies. Google is a fully autonomous vehicle while Tesla’s (current technology, anyway) is an advanced cruise control. Autopilot is not making nearly the calculations that Google’s systems are making. Google has collected exponentially more data in that 1.5 million miles than Tesla has even begun to.
Google’s self-driving cars aren’t “fully autonomous”. Without human drivers, they’d lose control approximately every 1,500 miles, according to Google’s own reports. They’re not actually much more sophisticated in reality. The “fully autonomous” versus “advanced cruise control” distinction is a myth perpetrated by Google’s Chris Urmson as a marketing ploy. Don’t buy into it. Sensors on a Tesla and sensors on a Google car are similarly able to collect and model data. (And simulate decisionmaking, whether or not the car is currently in control.)
That’s quite wrong. Google’s car is precisely fully autonomous even though it might rarely be at a loss. By contrast, Model S would crash every few minutes. What’s missing in this article is how frequently Tesla drivers have to catch it. Google’s system is very strong, Tesla’s (mobileye) is very very weak. They are worlds apart. I’m a big fan of Tesla and I’m all for autonomous drive, I’m an AI scientist, but Tesla’s current system is surprisingly weak.
Tesla’s autopilot receives one million miles of training per day by it’s drivers alone. I’m fairly confident they already have enough data to accurately drive between the supercharger points (inc. suburbs)
I haven’t done the math, but 47 million miles of truly-hands-off-driving would be hard for me to believe in such a relative short period of time. I’m not saying its not accurate, but I would definitely want to see a breakdown.
That being said, I’ve always wondered why Google is proceeding so slowly. They had 53 autonomous vehicles in the November 2015 report, but 4 months later, they only have 54. Come on Google! Get more cities up and running and get more vehicles. You’ve got endless amounts of cash and could add 100+ vehicles to their fleet in no time flat. What’s the hold up?
With 70,000 autopilot equipped cars driving only 10,000 miles a year each, that’s 1,000,000 miles a day if the autopilot is on only half the time. The mode has been available over 150 days.
Mostly because Google’s tech is nowhere near as far along as they’d like you to believe. Lying is good for business.
What Google is doing is exponentially more difficult than simply keeping in a well-marked highway lane and enabling active cruise control (acc). In my BMW i3 BEV I have used ACC for 18 months and it works great. In Europe, BMW has the lane keeper sensors available. Done. I love Tesla and reserved a Model 3, but “autopilot” is nothing new. And comparing it as somehow better than the Google effort is disingenuous.