Man vs. machine: Can self-driving cars really replace human drivers?
Self-driving autonomous vehicles have been dominating the headlines this year like never before: they've gone from sci-fi fantasy to actual reality in a very short space of time, and manufacturers say some form of automobile automation will arrive by 2020 or even earlier. If you're wondering exactly how far the technology has come and how far it has to go, we're here to answer some of those questions.
The pros and cons of self-driving cars
There are plenty of potential advantages to self-driving cars, scary though the concept may be. They don't get tired at the wheel, or distracted by kids in the back seat. They're not prone to daydreaming, and thanks to the bank of sensors and equipment at their disposal, they're much better than human drivers at spotting trouble coming from all directions – danger can be spotted from farther away and reacted to more quickly.
Then there's the actual driving: computer software can calculate stopping distances, braking speeds and junction spaces with mathematical precision. Traffic should flow more smoothly, congestion is likely to be reduced and fuel efficiency rates should rise once the robot drivers take over for good. Vehicles will be able to talk to each other too, meaning earlier warnings about traffic jams or accidents.
That said, there's still room for improvement in car sensors and detection equipment. Google says its cars can now "detect hundreds of distinct objects simultaneously – pedestrians, buses, a stop sign held up by a crossing guard, or a cyclist making gestures that indicate a possible turn." But they need to be able to recognize everything a human can, and make the appropriate response, and that's going to take more time.
Like humans, they aren't very good at interpreting body language from pedestrians and cyclists, and anticipating what these erratic humans are going to try and do next; they're also susceptible to remote hacking attacks, as we saw with the Jeep Cherokee bug exposed in July. No matter how bad flesh-and-blood drivers are at observing stop signs, they can't be remotely controlled using a laptop from the other side of the world, and that's an advantage we still have.
The technology of today and tomorrow
When it comes to freeway driving in good weather, self-driving cars are already very proficient. An adapted Audi fitted out with equipment from automobile specialists Delphi recently completed a coast-to-coast drive in the United States – a human being was at the wheel at all times, but Delphi says the car did 99 percent of the driving. Six long-range radars, four short-range radars, three vision-based cameras, six lidar sensors, a localization system, intelligent software algorithms and a full suite of Advanced Drive Assistance Systems were all fitted to the car.
Nissan and NASA have also partnered together to work on autonomous driving technology. Nissan President and CEO Carlos Ghosn says self-driving capabilities will start appearing in its cars from next year, starting with the simple (like parking assistance) and moving on to the more complex. NASA of course wants to use the same kind of technology for unmanned, deep space exploration.
One area where there's still work to be done is in reducing the reliance of these cars on information pulled from pre-existing maps. As The Atlantic reports, Google's cars drive so well in Mountain View, California because they've been fed with a hugely detailed map of the area in advance. This isn't just a map of roads and buildings, it's a map showing curb heights and road sign positions – and it's essential for a smooth autonomous drive. For areas where these maps aren't yet available, the cars' capabilities are reduced.
When we drive a car, we're not just relying on the information coming through our eyes and ears – we're also relying on our previous driving experience and our knowledge of how other drivers are likely to react. That's a gap that autonomous vehicles are still trying to close, even with these vehicles clocking hundreds of thousands of miles every year.
One step at a time
As we've seen at numerous robotics competitions, making something fully autonomous is an incredibly difficult challenge. Atempting to duplicate the experience, knowledge and intuition held inside the human brain requires some complex code and a massive database of reference material. For that reason, it's likely that our cars will become gradually more sophisticated and semi-autonomous as they go, rather than making a sudden jump forward – remember that technologies such as cruise control, parking assistance and satellite navigation are already common.
That's certainly Elon Musk's approach at Tesla. He thinks it will take a couple of decades for cars to go from partially self-driving to fully self-driving, at which point it might even be illegal for humans to drive (because of the associated safety issues with letting a man or woman behind the wheel). A software update rolling out later this year is going to give certain Tesla models a limited number of autonomous driving capabilities.
One of those will be navigating slowly around a car park to pick you up; another will be managing line changes at high speeds on a highway (like an advanced version of cruise control). As Musk points out, low speeds and high speeds are where self-driving software finds operating a car easiest – it's everywhere in between where there's less predictability and still some room for improvement. The Mercedes F 015, for example, struggles in rain and hot weather, and currently requires external beacons to guide it correctly.
In the next year or two, you'll be able to buy kits to convert your car into a self-driving motor for around US$10,000. They consist of a sensor and camera pod on the roof, a computer sitting in the boot, and a device under the chassis for controlling steering and speed. Kyle Vogt, chief of the Cruise Automation company selling these kits, thinks they'll be commonplace in the next three years.
That's a lot to consider, and of course government legislation will need to catch up before you can take one of these autonomous cars out on the road, but the technology is going to be here sooner than you think – even if the reality of fully self-driving vehicles are still some 10 or 20 years away.