In the race to be the first to develop a fully autonomous vehicle, there are plenty of entrants. But this is not a sprint to the finish; it is more of a marathon. In a typical marathon, you’ll see hundreds of runners take off at the starting gun, with little evidence of who’s going to be in the lead after the first few miles. But in those first few critical miles, the race is all but decided, and after that it’s only a matter of the leaders staying in the game and maintaining their positions.

In the world of autonomous vehicle technology, or self-driving cars, there are dozens of entrants, just like in a marathon. But if you look a little closer at the crowd, you’ll see a few companies pulling ahead of the rest. These are the companies that have a real chance at Number One. Nobody knows how far away the finish line is, but if these companies can keep pushing their limits and maintaining their pace, any one of them could end up holding the trophy at the end of the run.

Interestingly, there are only a handful of companies that qualify to reach the finish line. And it’s not because they’re the only ones with the resources to do it. It’s also not because they have a technology advantage in any way. It’s simply an age-old secret of the marathon: just keep running.

So let’s see who these stars of the self-drive are, and how they’ve progressed on the road to developing truly autonomous driving systems.

Waymo, the Google Self-driving Car Project Spinoff

Waymo started the race long before most of the competition. Google’s self-driving technology efforts go back to 2009 and Sebastian Thrun’s “Stanley”, the car that won the team $2 million from the United States Department of Defense at the 2005 DARPA Grand Challenge.

Since then, the project has gone thorough many changes, but in many ways it was responsible for the evolution of laws pertaining to road transportation. Today, the State of California, among others, at least has the framework necessary to test self-driving vehicles on the road, something that Waymo has taken full advantage of.

In terms of miles driven without a driver, Waymo comes out on top with over 3 million test miles. Though there are now more than 30 companies with permits to test self-driving vehicles in California, only a handful have actually logged any miles, and Waymo leads the pack by a huge gap. That’s not about to change unless another company is able to deploy thousands of test vehicles and aggressively run driverless miles in the next few years.

Uber, the Car-sharing Company with a Shadow Hanging Over It

Uber’s self-driving technology skills are currently under the gun, with Google claiming that much of its tech has been unlawfully ported over by a key employee that was one of the early members of the Google Self-Driving Car Project.

But controversy aside, Uber has done extensive testing with its fleet of cars in California and outside. Their aggressive push has arguably even pushed Waymo to do more in a shorter amount of time.

Uber’s self-driving technology development efforts began in 2015 with a partnership with the Carnegie Mellon University. Since then, Uber’s vehicles have driven tens of thousands of miles with only a backup engineer on board to take over in case of an emergency, also called a disengagement.

The future of Uber’s autonomous technology is still under a cloud, with the courts yesterday forcing the matter into trial rather than private arbitration. If it loses, Uber could lose months of testing and a whole lot of technological advantages over its rivals in the self-driving space. But even then, it could end up licensing the same technology from Waymo to continue its efforts into driverless cars.

Tesla, the Company that Lets Users Test Autopilot

Tesla has offered Autopilot software on its various models for a long time. The first model to have Autopilot was the 2014 Model S. Although Tesla is yet to offer fully autonomous capability as part of its Autopilot software, it is committed to demonstrating a full version by the end of 2017, with a release expected in 2019.

Tesla’s Autopilot comes as part of a Tech Package that costs extra. Two major hardware upgrades have been rolled out so far, Hardware v1 (HW1) and Hardware v2 (HW2). The firmware is currently on version 8.1 for HW2 models.

Tesla’s Autopilot has been in the news a lot. Whenever something goes wrong, it is dragged into the spotlight. However, there is some misunderstanding among many drivers as to the capabilities of Autopilot, since the system has never been fully autonomous and requires the driver to take control at a moment’s notice.

Other Notables in Autonomous Vehicle Technology

Nearly every automaker today invests in some form of self-driving car capability, ranging from US companies like GM and Ford, to German auto giants like BMW and Audi. They are all at various stages of development, but not much of their testing is done in public, as evidenced by the lack of miles logged with government authorities.

In the United States, GM and Ford are heavily invested in the technology, both pumping in hundreds of millions of dollars to support academic research as well as startups that are developing autonomous vehicle technologies. GM, for example, invested $1 billion in Cruise Automation to bring some of their forward-looking technologies into its fold. And Ford matched the ante this year as well, with an equal-sized investment over a five-year period in Argo AI, which has some of Google’s and Uber’s self-driving veterans behind the wheel.

Across the pond, German automakers are also investing equally heavily on new technologies, as are Japanese carmakers like Toyota. GM, BMW and Toyota recently invested in Nauto, which makes a two-way dashboard camera – dashcam – that monitors both the driver and the car.

The Growth of Cloud Computing and Edge Computing

But the front-end technology of the self-driving car is just one part of the equation. Autonomous vehicles generate tremendous amounts of data, and that data needs to be processed somewhere.

Today, that information is processed by hyper-computing servers at cloud data centers around the globe. Believe it or not, the average self-driving car generates about 25GB of data per hour. And that’s not all. All of this data has to be processed in near-real-time for the system to work. The time between a self-driving car perceiving an obstacle, processing the information and reacting to the situation has to be in milliseconds or nanoseconds.

That kind of number-crunching also requires a stable connection to the Internet through which the cloud servers that do the heavy lifting can be accessed. And that’s why companies like Microsoft are moving towards something called Edge Computing. That’s essentially a decentralization of the compute power to the periphery. That periphery could be the self-driving car itself, or a series of intelligent hubs along its route that handle the data transfer and compute power required to make those split-second decisions critical to autonomous vehicle technology.

Edge Computing is still in its nascent stages, but Microsoft has taken a bold step to change the company’s entire vision betting on this major shift in cloud computing. They’re calling it an Intelligent Cloud with an Intelligent Edge. And that’s going to be the formula for artificial intelligence applications from here on out. A cloud server is no longer sufficient because dodgy connections can render it useless in a mission-critical situation such as a self-driving car on a journey with human passengers.

The Internals

The other element of autonomous technology is the processing power of the cloud servers themselves. Such immense power can’t be handled by traditional CPUs, so now we have an emergent family of products that are closer to and beyond GPUs. Companies like NVIDIA, Intel, Qualcomm and AMD are at the forefront of such developments, giving rise to new generations of products for on-board and off-site application.

For example, NVIDIA’s Drive PX2 is a 256-core processor that boasts up to 1.5 teraflops, and is intended for “deep learning-based self-driving AI cockpit systems,” according to the company. Other companies are developing similar, powerful processors for on-board use. Intel IoT is also focused on Edge Computing at the source of data generation, meaning on-board the vehicle itself.

On the data center side, there are chips like Intel’s Xeon family of processors designed for superior compute power on the server side of the equation.

The chipmaking industry is one of the foundations of self-driving technology, simply because of the massive compute requirements for even the simplest of operations such as lane discipline on a highway. The more complex the situation, the more the data required and generated, and the more the compute power required.

The Many Moving Parts of Self-driving Car Technology

All of these developments are happening in parallel, and in a way, the speed at which the underlying technology develops governs the speed at which autonomous technology evolves.

One of the biggest points to note here is the fact that cars have very little to do with respect to self-driving capability. None of these modified cars have anything different about the way they operate. The drive trains are still the same, the body, the interiors, they’re all the same. The difference is in replacing a human driver with a ton of technology that can do the same job.

Looking at it from that perspective, we’re not going to see something out of a science fiction movie when we cross the finish line in the race to develop the first fully autonomous car. It’s going to be your everyday Audi or Volvo or Minivan or whatever. The difference is in the way the automobile operates itself, as opposed to requiring a driver to operate it.

That seemingly minor detail is what the whole field of autonomous cars is all about. It is about taking the human element out of driving, but retaining the human skills that are required for safe navigation.

The entire chain of whatever we spoke about is critical, each in its own way. The car needs to have the right hardware to collect the data, and process some of it; the connectivity to a cloud data center or an intelligent hub is essential; the immense compute power is a prerequisite; and all of this has to come together in a symphony software and hardware to do what we take for granted – drive a car.

Another important component of this – probably the most critical one of all – is the artificial intelligence programming that goes into the mix. Without it, nothing else in the realm of self-driving cars would be possible.

Advancements in artificial intelligence algorithms are at the forefront of such technologies, and companies like Google already have the experience and resources to take this to the next level. The acquisition of UK-based DeepMind certainly gave them an edge in AI, but they’ve been using intelligent algorithms right from their early days of search engine development. The introduction of neural networks and deep learning to that mix has merely given them a jump start to developing software that befits such a lofty purpose as autonomous vehicle technology.

And Google, or its parent company, Alphabet, is not the only one engaged in deep learning. Every tech major from Microsoft to IBM to Amazon Web Services is pushing the boundaries of AI capability.

As I said, all of these moving parts have to be able to work together to achieve the final goal. That hasn’t happened yet, and it’s not about to happen overnight. There are still several hurdles that need to be jumped over, not only on the hardware and software front, but on the regulatory front. Moreover, humans need to be fully confident in the technology before it can be adopted en masse.

Humans are a finicky species, and we’re not going to embrace such a critical thing as self-driving car technology before we’ve seen it live and successful for a considerable period of time. The paradox here is that people won’t adopt it as a mainstream technology until they’ve seen it working on a large scale in the real world; and that can’t happen until people start adopting it on a large scale.

So, not only are we currently bound by the technological and regulatory limitations, but we are also bound by the fear of the unknown. We still don’t know what it would be like to have a city full of autonomous cars. In fact, we don’t even know what it would be like to have 10% or 20% of the cars on the road being autonomous while others are still driven by human beings.

We’re focusing so much on how the self-driving car will react to human drivers on the road that we’ve lost sight of learning about how human drivers will react to autonomous cars on the road. Will they “bully” autonomous cars into giving right of way? Will they take advantage of the safety stops and bend traffic rules to get what they want, like a parking spot?

We don’t know the answers to any of these things, so it’s going to take time for autonomous car technology to make its way into our daily lives.

The Potential Winner of the Self-driving Car Technology Race

As I mentioned at the beginning of the article, this is a marathon, not a sprint to the finish. In a marathon, the leaders are usually spotted early on. But that’s not a rule. Any one of these companies could pick up the pace and deliver the first fully autonomous vehicle technology before any of the others. For now, it is the tech majors against the auto majors, but everyone wins in the end.

The first successful piece of technology is likely to be heavily licensed once it has proven its mettle, and that’s no small feat. A fully autonomous system has to jump through numerous regulatory hoops, and the lack of comprehensive laws governing autonomous vehicles is only going to make the going even harder.

There’s no resource shortage among these top contenders in the realm of driverless automobile technology. Any company with the technology will have access to untold millions to get it to market. The problem here is with regulators, and the fact that most laws pertaining to automobiles date back to horse-drawn carriages on public roads. And then there are also the other ‘human’ problems we talked about earlier.

Is Waymo going to be the first across the elusive finish line? They’re certainly in the lead right now, but that’s no guarantee that they’ll be breaking the ribbon. It could be Uber or Tesla, or it could be any one of the dozens of companies that are aggressively developing this tech.

The only upside to all the waiting is that we’re guaranteed to have a working self-driving car system in the next few years. Considering the sheer amount of resources being poured into this technology, that is inevitable. The only question is, how long after that will it be before we, as humans, are willing to make them part of our daily lives?

Share...Share on FacebookTweet about this on TwitterShare on LinkedInShare on RedditShare on StumbleUponShare on TumblrPrint this pageEmail this to someone

LEAVE A REPLY

Please enter your comment!
Please enter your name here