A stumpy little airport shuttle bus is New Zealand’s first autonomous vehicle. But we need a whole infrastructure and 5G cellular technology before a driverless future arrives, writes Nikki Mandow.
Driverless cars are big business.
As The Download went to press, there were no less than 27 companies working with self-driving technology in the test-bed state of California – companies like Google, Uber, Tesla and some of the big car manufacturers. Computer chip-maker Intel recently offered US$15 billion for Mobileye, an Israeli autonomous vehicle technology firm. And the California Department of Motor Vehicles has just proposed new regulations to finally prepare for moving from testing to the commercialisation of driverless cars.
In New Zealand, we had our first driverless vehicle trial in January this year – it involved a stumpy little shuttle bus at Christchurch Airport. But development of the much touted self-driving vehicle industry, along with its related benefits (anything from fewer accidents, to better mobility for blind and elderly people, to reduced road and parking congestion, to being able to legally text on the commute home) needs more than the arrival of a load of self-driving vehicles at a Kiwi port. It needs a whole load of infrastructure to support it.
Nikki Mandow talked to Chorus’s network technology strategist, Kurt Rodgers, and the New Zealand Transport Agency’s director of connected journeys, Martin McMullan, about what New Zealand needs to do to get ready for a driverless world. A warning – it’s not going to be quick and easy.
What’s the most important issue when it comes to driverless cars?
Low latency. Or, in human speak, there being only a (really) tiny delay when data gets sent or is received by the vehicle. It’s not going to work if it takes three seconds for your car to notice the boy running into the road and avoid him. You can’t have a second or two of delay receiving signals from a traffic light or an approaching bus.
Where does latency creep in?
There are lots of potential causes of latency – in-built delays in the car’s computer system, slow mobile or wireless signals, or delays caused when a vehicle is trying to communicate with a data centre that’s a long way away.
The goal for driverless cars is to get to one millisecond (one thousandth of a second) latency.
How do you get latency down?
The critical step for autonomous vehicles is the development of 5G (short for fifth generation) cellular wireless technology. We need 5G because it’s so much faster than 4G. As the UK’s National Infrastructure Commission puts it, somewhat breathlessly: “5G means seamless connectivity. Ultra-fast, ultra-reliable, ultra-high capacity transmitting at super-low latency.”
Chorus’s Kurt Rodgers was at the recent Mobile World Conference, in Barcelona, and watched demonstrations using an autonomous toy truck on a race-track. Using 4G (the present top mobile standard) the truck took a long time to respond (turning, braking etc), he says. “When they switched to 5G it drove perfectly, because 5G has much lower latency.”
So, when do we get 5G?
The problem is that 5G technology doesn’t exist yet. In fact, global technology wizards are still arguing about its specifications. No one’s anticipating 5G networks being rolled out much before 2020. (Techie people get all excited about 5G, but, to put it in perspective, there are still tens of thousands of New Zealanders using 2G phones. And you can still buy a new 3G smartphone, although 4G is better.)
Does New Zealand have the infrastructure we need for 5G?
Not yet. 5G will operate on a higher frequency (because that moves information faster) but higher frequency signals don’t travel as far as lower frequencies. And that means a lot more cell sites will need to be built to capture and transmit the data. Rodgers says we are going to need tens of thousands of small cell sites – in some places that’s going to mean a site on every street corner. That’s expensive – and we don’t yet know who’s going to build them in New Zealand.
NZTA’s Martin McMullan says it’s also critical to make sure that where possible there is ubiquitous mobile coverage on every road in New Zealand. The Government’s $50 million Mobile Black Spot Fund will go some way towards bringing at least a 3G or 4G mobile service to more of our rural roads, but we still won’t have the coverage needed for a nationwide driverless car network.
Are cell sites the only extra infrastructure we need for driverless cars?
No. To get the speeds and low latency needed for autonomous vehicles each of the new cell sites will need to be connected to fibre.
Anything else?
The last big part of the telecommunications infrastructure puzzle involves data centres – those vast storage warehouses (often in the US) that are the backbone of the cloud. There was a time when New Zealand’s internet relied largely on overseas data centres – in the US, Singapore or Australia, for example. However, the advent of video-streaming services meant overseas storage (with its delays moving information backwards and forwards) made for a slow service. New Zealand now has a small number of its own data centres, mostly in the big cities.
But that’s not going to be enough to achieve the millisecond latency needed for driverless cars, says Rodgers. Even with 5G and lots of cell sites, you’re going to need lots of smaller data centres to pick up and transmit the data from all the vehicles in (almost) real time, he says.
“I talked to the guys at Intel when I was in Barcelona and they said that to get one millisecond of latency, you need data centres every 10 to 20 kilometres. That would mean having hundreds of data centres in New Zealand. The question is who is going to make that investment, particularly when we don’t yet have any certainty about when 5G is going to happen, or what the demand will be?”
Has anything happened so far?
Chorus has just signed an agreement with start-up company VertIS to convert existing telephone exchanges into data centres. At the moment, the demand for these local data centres is around caching Netflix and other video-streaming services, but they could be used for driverless cars. The new buzzword is “fog computing” (also known as edge computing) where instead of data being stored in the cloud (a far-away data centre) it is stored much closer to the ground (ie., near the individual user). Hence “fog”.
How is Google able to test driverless cars without 5G and all the extra infrastructure?
Think of a driverless car as a robot controlled by an artificial intelligence brain. With the Google car, and other test cars, all the computing power sits in the vehicle. The sensors “see” what’s going on and the car makes decisions. But having all that processing power inside each car takes up valuable space and is very expensive. Self-driving car prototypes collect about a gigabyte of data per second – that’s the equivalent of a feature length, high-definition film’s worth of data every five seconds. The first Google cars were said to contain upwards of US$150,000 of technology, although that cost is coming down. Still, the more computing power they can shift out of the car, the cheaper the car.
What about non-telco infrastructure – what else do we need?
NZTA’s Martin McMullan says getting the transport and road sectors ready for driverless cars is a big task and will involve different sectors working together. “The transport industry is a $5.4 trillion business and it’s going to be digital in the next decade. We’ll need good, robust partnerships with public and private organisations, including those in the cellular and technology space.”
McMullan says New Zealand’s regulations are conducive to testing driverless cars and he expects to see more happening here.
Won’t we need a sensor network on our roads?
McMullan says because driverless cars are equipped with so many sensors this will reduce the need for on-road sensors. In fact, the cars of the future will be able to help NZTA with its work, he says. For example, an autonomous car that spots a pothole in a road could automatically send a message to NZTA.
Do we need an electric vehicle network to have a driverless car network?
Not necessarily, McMullan says. Ford, for example, is testing an autonomous car with a conventional petrol engine. But most pundits predict that the development of driverless cars will progress in parallel with the development of electric cars – in tandem with the infrastructure of charging stations these will need.
“The likely path will be that electric vehicles will become connected vehicles; connected vehicles will become autonomous vehicles, and they, in turn, will become autonomous connected vehicles.”
The billion-dollar question: when?
Overseas experts are divided about when the first autonomous cars will go on sale. Optimists talk about fully automated cars being on the freeways and some city streets in the US (and maybe some parts of Europe and Asia) around 2020, or at least some time between 2020 and 2025. Putting them into dense urban areas might take a bit longer, as will the full automation that would allow blind people to hop in a car and have it take them to their destination.
And in New Zealand? It’s not an exact science, but McMullan estimates we are three to four years behind the US in terms of deployment of driverless cars. If you look at a related technology, the Tesla electric car, Tesla launched its Model S worldwide in 2012, but the company has only just signalled its New Zealand launch. The first cars should be on sale by June this year, along with an Auckland store that is being set up.
That’s a long time coming. Will we get some clever cars before then?
Some top-end cars are already itching to prove they are better at driving than you. They include some Tesla models, BMWs and Mercedes. Several offer cruise control – where on a motorway the car takes over and maintains a safe speed and distance from other vehicles, braking and accelerating when required. Some will even auto-correct if you change lanes without signalling and some can change lanes on their own. Several can also help you park in a tight space. And cars that can park themselves (you get out and it does the parking for you) are already available.
Boring… most of us can cope with parking
OK, yes, but this ability could allow for narrower parking spots (because you wouldn’t have to allow space to open the car door and squeeze out between cars) and could also cut nervous parkers’ stress levels. Tesla already offers downloadable software that allows its Model S to park itself, while Nissan promises a self-parking mobile app next year for its LEAF electric car. But the next step is more exciting. Imagine being able to leave your car in a designated drop-off spot, at the entrance to a parking building, say, and it then finds its own spot and parks for you. And when you return, you just let the car know and it comes to pick you up. Now that is cool.
Six steps to a fully driverless car The National Highway Traffic Safety Administration, and most automated car developers, use a six-stage progression to describe automated driving systems, moving from complete driver control to full autonomy.
Level 0:
The human controls it all – like we've been doing all along.
Level 1:
Most functions are still controlled by the driver, but a specific function (like steering or accelerating) can be done automatically by the car.
Level 2:
The car takes over two or more functions. For example, there might be adaptive cruise control combined with technology that keeps a car in its lane. Level 2 cars are already on the road.
Level 3:
Drivers are still behind the wheel at all times, but they abdicate control under certain conditions — on motorways, for example, or at certain speeds. The driver must be ready to take over at any time, with the car allowing some time to make the transition. Such cars are being tested but are not yet on the market. Still, they might be available before 2020.
Level 4:
The vehicle is fully autonomous under most but not all driving scenarios. All the driver needs to do is say where he or she wants to go, then sit back and enjoy the ride. Level 4 might exclude conditions like gravel roads or extreme weather. Experts say level 4 cars are likely to appear sometime between 2020 and 2025, probably nearer 2025 in New Zealand.
Level 5:
This refers to a fully autonomous system that expects the vehicle’s performance to equal that of a human driver in every driving scenario. At this “automated chauffeur” level, a blind person or a child could hop in the car. We are probably looking at 2030 or beyond for this level.