Why autonomous cars are the driveway to future

Why autonomous cars are the driveway to future

SHARE

As the idea of a self-driving or autonomous car starts to become less and less far-fetched, will tech-hungry consumers be willing to cede control to a machine? By Ishveena Singh

Autonomous cars and geospatial

It’s been a pretty long day at work. You’re at the wheel, tired. You don’t realize when your eyes close and you drift off your lane. Somebody honks. You wake up, startled. This time, you got lucky.

Accelerate to 2025 — the year by which all major automakers plan to get their autonomous vehicles on the road. You get in your car, punch in the destination, and sit back. The engine revs into action. The car’s computer connects to the Cloud to get real-time traffic data. Sensors and software discern objects like pedestrians, cyclists and vehicles, and navigate safely around them. You, meanwhile, are free to catch up on those zzz’s, work on the business report you have to present that afternoon, or simply enjoy the latest Netflix show on your tablet.
Champions of self-driving cars envisage a time when road accidents will become an unfortunate thing of the past. Like the smallpox. With ever-alert computers taking over from reckless drivers, motor vehicle crashes would plummet. Car-sharing would become commonplace. Fewer automobiles on the road would translate into more fuel savings, less carbon emissions, and lesser pressure on the city infrastructure. There would be no need to put up with irritable chauffeurs. The elderly, differently abled and visually impaired would become free from their physical limitations. The society would attain vehicular nirvana… Or would it?

The pot of gold

The debate on that may have just started, but at least one thing is clear. No company wants to be left behind in the road to a pot of gold filled with endless possibilities. Audi is aiming to beat all industry predictions and put its A8 driverless limousine in the showrooms in 2017. Meanwhile, Tesla has already equipped its Model S electric sedan with a patch called Autopilot, which allows the car to operate autonomously under certain conditions. Tesla CEO Elon Musk, who describes this technology as a “really good chauffeur”, expects the company’s fully autonomous vehicles to make their debut as early as 2018. Google plans to partner with several different companies to bring its much talked about self-driving car technology to the market. Its timeline is set for anywhere between 2018 and 2020.

And if Google is pushing for autonomous driving technology, how can “China’s Google” be far behind? Baidu has partnered with BMW to roll out a self-driving car prototype before the end of this year. General Motors has said it would introduce a fleet of autonomous Chevy Volts on its Warren Technical Center campus in Michigan in late 2016, but, these would only for employee use. Japanese carmaker Nissan has confirmed at the Tokyo Motor Show last month that it is “well on track” with plans to “equip innovative autonomous drive technology on multiple vehicles” by 2020. Jaguar Land Rover predicts its driverless vehicles to hit the roads in 2024. Daimler has its eyes set on 2025.

Even ride-hailing service Uber is eyeing a slice of the smart car pie. The company has made public its plans to open a research and development center for driverless cars in Pittsburgh. And while Apple might not comment on numerous rumors that the company wants to follow Google’s footsteps and build its own electric car, chief executive Tim Cook does acknowledges the massive transformation happening in the auto industry. Speaking at the 2015 WSJDLive conference in California, Cook pointed out, “When I look at the automobile, what I see is that software becomes an increasingly important part of the car of the future. You see that autonomous driving becomes much more important.”

Crawling the world for autonomous cars

Software, indeed, will drive the car of the future. And powering that software would be an ultra-precise digitization of the physical world. No, we are not talking about a normal digital map which shows you road intersections on your smartphone. These super-accurate maps are packed with tiny details, such as, the position and height of every single curb, measured all the way down to centimeters and inches. As James Etheridge, head of media relations at Nokia’s mapping service HERE, says “the map is no longer an imprint of the world frozen in time; it’s becoming a conduit for the dynamic data generated by vehicles, people and businesses”.

The Original Google-X man - Stanford Professor Sebastian ThrunGoogle’s self-driving car team’s mapping lead Andrew Chatham explains what goes into the maps they are developing for their vehicles. “[Our maps] are any geographic information that we can tell the car in advance to make its job easier. We tell it how high the traffic signals are off the ground, the exact position of the curbs, so the car knows where not to drive. We’d also include information that you can’t even see, like implied speed limits,” he says. In the United States, Google had mapped 2,000 miles of road this way till mid-2014. The US road network sits at around 4 million miles. And Google is looking to map every single street where its car might want to operate.

Ever since Uber started dabbling with the idea of driverless cars, it has also become serious about developing its own mapping platform. It acquired San Jose, Calif., based mapping startup deCarta in March this year. After that, Uber purchased around 100 of Bing Maps’ employees, along with some of its mapping assets. It was also in the bidding to acquire HERE, but lost out to a consortium of German carmakers. And now, it has been learned that Uber has been hiring contractors to drive its mapping cars and capture 3D images of local streets, a la Google Street View.

Tesla is also creating high-precision digital maps of the earth using GPS. And it is acquiring that data through its drivers. Every Model S car, with or without Autopilot, is connected to the Cloud. So, the company is using data from each of its cars to develop maps. In the meantime, GM is also researching precision mapping to guide its future cars.

Now, it’s not like that the cars of the future are some dumb machines. They come with their own set of intelligent armor: radars, sensors, cameras, et al. But, as Etheridge points out, “A car’s camera and radar can’t see through another car or around a bend or a building.” And even without obstruction, the range of those sensors is limited to 100-200 meters at best. You need a much longer electronic horizon to make the right decisions in real-time. “An airbag deploying 500 meters ahead? Likely an accident, which you can be routed around. Tires slipping on a warm day? Perhaps heavy water build-up or a spill which other cars should be warned about, so they can slow down while approaching the area? Cars have an incredibly rich array of sensors generating a ton of data, which currently just sits in the car. The next few years will see us begin to harness that data in the Cloud and do useful things with it.”

Couch on the road

There’s a branch of artificial intelligence (AI) called deep learning, which trains computers to understand patterns in large reams of visual data. Deep learning can be extremely significant for self-driving cars’ safety systems. It can be used to program the software to recognize different kind of automobiles, including emergency vehicles. It can enable your computer on wheels to detect speed limit signs. Or figure out that there’s a truck on the left, so it should not try to change lanes and cut it off. Essentially, an autonomous vehicle is trained to work just like our brains would — by accepting sensory input and acting accordingly. But, how many unexpected situations can you actually feed into the system?

The first time a Google car spotted a couch in the middle of the road, it could not figure out what was going on. The human safety driver had to take over. But soon, the software on all the cars was upgraded to handle such a situation. And all self-driving cars got to learn from one car’s mistake, even those that have not been manufactured yet.

When it comes to AI, the learning capacity of computers is endless. Tesla’s Musk calls this a “fleet learning network”, where all cars contribute to a shared database. “When one car learns something, all learn,” Musk says. Take I-405 in California, for instance. It’s a highway where lanes are terribly marked. But, Tesla’s Autopilot functions well on this section also because it has all the requisite information from Model S drivers who pass through this specific stretch of road.

Chipmaker Nvidia knows that interpreting tons of data an autonomous vehicles generates every second requires an obscene amount of processing power. Which is why, the company has shifted its focus from video games to using deep learning technique to push into the world of driverless cars. Danny Shapiro, Director of Nvidia’s automotive unit, affirms: “This notion of being able to build a brain for a self-driving car has really accelerated the demand for our technology.” So much so, the company’s automotive unit posted 85% annual growth in sales in the last fiscal year. And just last month, Nvidia’s director of deep learning, Jonathan Cohen, was poached by Apple — yet another sign that the tech titan is getting serious about autonomous cars.

The winter is coming

Apple and Google have plenty of cash to burn — their combined bank worth is estimated to be around $270 billion. And Musk has even quipped that if Apple makes a self-driving car, it would finally be able to “offer a significant innovation”. But, what is making traditional automakers ignore an inevitable peril of autonomous driving: Collapse in future car sales?

Sebastian Thrun, a computer scientist at Stanford University and a former leader of Google’s self-driving-car project, believes that self-driving technology will challenge the very notion of car ownership. “There will be fewer cars on the road — perhaps just 30% of the cars we have today,” he insists.

A University of Utah research foresees a more disruptive future. It predicts that an autonomous taxi with dynamic ride-sharing has the potential to replace 10 private cars. And with fleets of driverless cars offering greater mobility with far fewer vehicles, car sales would take a nose-dive. For several manufacturers, this would spell death or acquisition. Just like smartphones upended Nokia and Kodak. Traditional automakers would need to shift from hardware to software model, and the value in the industry would come from services, rather than the product.

But, the auto industry is not the only one that needs to worry. The terrible driver’s dream car could doom the auto insurance industry. Robert W. Peterson, a professor at Santa Clara University School of Law, notes, “Over 90% of accidents today are caused by human error. There is every reason to believe that self-driving cars will reduce frequency and severity of accidents, so insurance costs should fall, perhaps dramatically.”

The various facets and forces that must come together to enable self-driving

The various facets and forces that must come together to enable self-driving

Well, there would still be non-crash related situations: theft, vandalism or a tree falling on your car. But, maybe insurance companies should think of themselves in the position of record companies before the iPod came out. A Brookings Institution study predicts that autonomous vehicles will complicate the already complex entanglements between insurance providers, plaintiffs, drivers/owners named as defendants, and manufacturers. And if personal liability tumbles, liability for auto manufacturers will go up. More so, if an enthusiastic hacker decides to have a little fun with the car’s security system. Post-sale safety, the study noted, will focus on software upgrades. Manufacturers that become aware of potentially risky software issues will need to provide upgrades as soon as possible, but, at the same time, they will also have to ensure that the new version is properly tested.

There are ethical questions too that need to be answered. What if an autonomous vehicle finds itself confronted by an unavoidable accident? Should the car be programmed to hit another vehicle or a pedestrian? Or should it just crash itself into a wall, potentially hurting its occupants? And who do you hold liable in such a situation? The Toulouse School of Economics in France conducted three surveys in this matter. The research showed that people are “relatively comfortable” with the idea that driverless car should be “programmed to minimize the death toll in case of unavoidable harm.”

Automation will also be bad news for the taxi industry. According to a Columbia University research, a fleet of 9,000 autonomous vehicles have the potential of replacing New York’s all 13,000 taxis. No wonder that the technology has piqued the interest of Uber — its biggest cost is paying the taxi drivers. Uber CEO Travis Kalanick has even admitted to this in an interview. “When there is no other dude in the car, the cost of taking an Uber anywhere becomes cheaper than owning a vehicle,” he has said.

A brave new world

So, what about those love the feel of flooring the accelerator, or the sound of the engine revving? Perhaps they are hoping that the technology will never work. Or that the widespread deployment will take so much time that they won’t be around to see it happen.

In fact, a research by The University of Michigan Transportation Research Institute indicates that a large proportion of American adults (68%) would be either very or moderately concerned about riding in a fully self-driving vehicle.

Michael Sivak, the co-author on the study, points out autonomous vehicles will increase people’s susceptibility to motion sickness as well. He says, “Basically, the problem is caused by the activities that people would like to do in self-driving vehicles (which they are unable to do while actively driving), such as, working on a laptop, watching movies, playing video games, etc. These types of activities are known to increase the frequency and severity of motion sickness.”

Technological advances will come. And if the cars get a chance to prove they really are both reliable and safe, cultural adoption will also follow. But, the biggest hurdle before driverless cars today is the regulatory one. Lack of government support could be a significant obstacle to adoption. Engineers need to know what the government is going to come down on and what it will allow. In the absence of national guidelines, there is going to be a lot of confusion over the rights and liabilities involved with the use of autonomous driving technology.

But, like Musk says, “It’s going to be interesting, ultimately, to see how cities handle these disruption waves, which are going to be coming faster and faster. Some cities are going to allow it, and then they’re going to be the bastion of the future, and the other cities are going to look like they’re in the Middle Ages.”