In my previous article, I’ve made a (not so) brief introduction to what self-driving cars are, their classification, and how do they generally work. In this entry, I’ll list the companies that are currently working on self-driving tech, how close are they to market, and what is their approach.
Which companies are trying to build self-driving cars?
In short: all of them. A few years ago, this was just a pipe dream, but with the recent advances in image processing, it’s very close to being a reality, and all the major tech companies see this as the multi-billion dollar opportunity that it is. Every automaker wants to add the “self-driving capability” sticker to their cars (mostly by fear of becoming the last ones to market), and every tech company even remotely related with cars or AI is giving it a shot too.
Waymo (previously a section inside Google) is the oldest name in this list. When self-driving cars were just a distant dream, they flooded the tech news with a video of a Toyota Prius driving itself. That was on 2009. You can check their history on their site.
Their approach was interesting, to say the least. First, they started with a commercially available vehicle (Toyota Prius) and they attached sensors and motors to it. Then, they built an actual vehicle from scratch. That “car” lacks windshields, mirrors, steering wheel and pedals, so it’s more like a small train wagon with wheels. And now, it seems they have settled for striking a deal with an automaker to modify a commercially available vehicle (Chrysler Pacifica) to add sensors and motors to it. These sensors are deeply integrated into the chassis, and so it looks much better that the original Prius prototypes.
Computer vision wasn’t as advanced back in 2009, so they built their platform relying heavily on LiDAR sensors, which are much more expensive than radar or cameras.
At the moment of writing this, Waymo has driven over 3 million miles. They also have an extensive test track. In fact, they hire people to be “human obstacles”, so their job is to cross the road when the self-driving car approaches and hope that the program works. Of course, their cars also have a trained human driver ready to take control when something goes wrong.
Standard disclaimer: I’m a huge Tesla fan. I’m constantly amazed at what Elon Musk can pull off, and I think they will have a major part in shaping the automotive and energy industry in the coming decades. I also follow Electrek for this kind of news, which is also biased towards Tesla. Only by looking at that blog, you could reach to the conclusion that Tesla is on track to beat everyone else to market and deliver a fully autonomous car by next year. That’s not going to happen, but Tesla’s approach to self-driving cars certainly has merit.
Their Model S cars had “level 2” autonomy, which means they can pretty much navigate through a highway or a regular road without human intervention (but the driver must still be ready to take control at all times). The hardware and software for that was provided by MobilEye. In October 2016, Tesla announced that all cars built from that moment on would have all the required cameras and sensors for full self-driving capability (level 5, no driver required), and they would build the software in-house. That system heavily relies in cameras (8 total) and a powerful computer with advanced computer vision algorithms. That makes it comparatively cheap to manufacture. You can see more details about it here.
Their approach is interesting because they are releasing tens of thousands of cars with full self-driving hardware, but the software is not there yet (at the moment of writing this, it has reached the same level of “level 2” autonomy of the old MobilEye-based solution). They will regularly push updates to the cars’ computers so they will become better and better at driving, until they eventually will become self-driving cars (pending tons of testing and regulatory approval, of course).
Ford, Chrysler, Daimler (Mercedes), BMW… All major automakers already offer some kind of “Level 2” autonomy as an option, in the form of automatic parking maneuvers or highway cruise control. Seeing the advances in this market, no one wants to be the last one to launch a self-driving car to market. Having a car be self-driving will be a huge selling point a few years from now. Or, more accurately, in a few years offering only “manual” cars will be a huge competitive disadvantage.
The news about the advances of the big automakers in regard of self-driving cars are few and far-between, but you can be sure that every car manufacturer, big and small, has a secret lab dedicated to this.
Uber / Baidu / Others
Uber obviously sees self-driving cars as an important step in its business. It would mean getting rid of the human drivers, which have the nasty habit of wanting to get paid. Baidu (China’s biggest tech company) is also researching autonomous cars. Apple is said to be working on a self-driving car too, but since its usual secrecy no one knows how advanced they are.
Which company is going to be the first to market?
Well, that depends on the definition of “first to market”. To get a baseline definition, I’ll define it that a company has a “self-driving car in the market” when you, as a regular person, can buy one a car from them, and then ride it from the back seat (level 5 autonomy). Or, in the case of Uber (for example), when the company deploys a fleet of driverless taxis nationwide (not restricted to a single city).
Waymo is already in what it looks like the final stages of testing. They’ve driven several million miles on real roads, the amount of times the human drivers have had to take control has been greatly reduced, and their self-driving Crysler minivans don’t even look like prototypes anymore. Based on the present, I’d say that Waymo will be the first to market. Nevertheless, their reliance on LiDAR sensors will mean that their path to market will be more complicated, because that equipment is much more costly to manufacture. Most people would gladly pay a few thousand dollars extra for their car to be self-driving, but tens of thousands would be a stretch.
Tesla is making wild promises. They teased a US coast-to-coast drive without the human driver ever touching the controls before the end of the year. That’s a lot of terrain to cover. Even discarding those claims, the fact is that their approach sounds better to me:
- Rely on cameras and the major recent advances on computer vision, so the self-driving hardware is cheaper to manufacture.
- Fit all their cars with said hardware, tens of thousands of cars with self-driving sensors are on the roads right now.
- Make the sensors work even if Autopilot (their name for “cruise control”) is off. That means that the computer will be able to process the sensor data, simulate what action would have taken if Autopilot was engaged, and send all that data to Tesla to train and refine the algorithm.
I can’t stress enough how valuable that amount of data is to train a complex algorithm. Waymo boasts that they’ve driven 3 million autonomous miles over 9 years (which is impressive). A fleet of Teslas could reach that number in months (which is mind-blowing).
I would say that Waymo and Tesla are on a league of their own. Waymo has been making consistent progress over 9 years, and no one can deny that they employ the most beilliant proffesionals on their respective fields. Tesla is the newcomer, but it’s relying in newer technology, has a clearer path to market, and it has the competitive advantage of having essentially a test fleet of hundreds of thousands of cars. I don’t see any other automaker or tech company topping those two companies at this point. So, who will be first? I honestly can’t say, I can’t even speculate. I can find solid arguments for any of those companies. I guess we’ll have to wait for a few years.
To be continued…
That’s already a lot of information to process, isn’t it? Please tell me in the comments if I missed a big player in self-driving tech, or if you think some of them should be taken more seriously.
In my next article, I veer off the actual facts and try to predict what self-driving cars will mean to several industries and our society in general, and an estimated timeline on when all that will happen. But that’s all speculative, so I’d love to hear your thoughts on that!