The intersection danger
Automakers are racing each other to develop the first production-ready selfdriving car. With Audi being the first to break both the autonomous speed and distance records back in 2014, and Ford proclaiming it would strive to be the first to offer an autonomous vehicle “to the masses”, it’s Mercedes-Benz that has so far stolen the show with its jaw-dropping S-Class range of Intelligent Drive vehicles.
While the first advances have come in the form of adaptive cruise control, active lane keeping, and self-parking systems, full autonomy remains the holy grail, and every major automaker and software developer is now on a quest to take the first swig from that grail. But, of course, there are some challenges along the way.
The recent tech and car shows have been full of spiels and self-praise about manufacturers achieving incremental levels of autonomy, thus paving the way for the self-driving cars, including driverless taxis, in the future. While we shouldn’t downplay their achievements in pushing autonomous driving technology forward, there is still a significant technological gap between “now” and “the future”.
If anything, the recent spate of fatal accidents involving Tesla and Uber self-driving cars proves that the necessary technology for autonomous vehicles to anticipate and adapt to all possible eventualities is still some way off. A self-driving car should be equipped with sensors that will identify objects, especially pedestrians, in the most challenging conditions imaginable. That includes situations in which humans would typically fail because the very aim of autonomous vehicle technology is to be better than a human driver. These sensors work perfectly well in good weather and daylight, but they seem to have limitations under conditions that are even slightly out of the ordinary. Cameras aren’t as effective at night time, and LiDAR systems “get frustrated” in heavy rain and mist. More challenging though, these systems should be able to make splitsecond decisions to anticipate and respond to the erratic driving of non-autonomous vehicles, like taxis and learner drivers. These are challenges that go beyond just good sensors, and thus require a form of artificial intelligence that does not yet exist.
When a second’s difference can avoid a collision, getting the tech right truly is a matter of life and death.
The self-driving craze caught governments unprepared when it hit. While many have since scrambled to put laws into place to regulate the development and testing of selfdriving cars, here in South Africa we remain slightly behind the curve. Soon they will be scrambling again to ensure that accidents won’t happen and that relevant parties will be held liable in the event of such accidents.
To be fair, most of our laws and regulations are made in reaction to incidents. Very few have the foresight to act on what has yet to happen. Technology, however, rarely waits for the law to catch up and, just like with drones or UAVs, lawmakers and authorities are being challenged to work faster before tragedy strikes.
Here’s the thing: the self-driving cars of the future won’t have drivers. Some might not even have anyone inside capable of driving at all, and more than likely they won’t even have a steering wheel. Many car makers paint a picture of a future where a driver or passenger can sit back and relax, maybe chat and play, with nary a worry. Of course, they paint the ideal future, not what we have at present. But between now and then, the biggest threat to autonomous cars will be human drivers, simply because humans are easily distracted and often take calculated risks, like speeding up to cross an amber light, or push in front of other cars in fast-moving traffic. And then there are the uncalculated risks, which some drivers seem to take effortlessly and continuously, without any regard for other drivers. These challenges pale in comparison to the anxiety of sitting in the back of an autonomous car, stranded in a no man’s land of being unsure whether to trust in the automation or to be in an even higher state of alert.
Safety is always of utmost concern, and part of that involves not just drivers but also the other humans and living beings outside the vehicle. Pedestrians, like human drivers, are unpredictable, and more so when distracted or inebriated, and therefore pose the biggest challenge for autonomous vehicles. Throw into the mix stray animals, and you have a real challenge on hand. Most developers of autonomous vehicle technology agree that the autonomous challenge will not be solved before the pedestrian problem is solved, and only when pedestrians can be guaranteed that they will be safer by an order of magnitude with autonomous cars on our roads.
One thing’s for sure: The promise of autonomy is putting swagger back into the once-troubled car industry’s stride. Despite a continued threat from car sharing, ridehailing and all-electric cars, the legacy automakers now have a unified goal to work toward, and it’s something they’re all convinced we want. According to Mercedes, their F 015 fully autonomous concept will head for production in less than 15 years, and according to Tesla, Waymo and many others, we could have Level 5 autonomous vehicles by 2025. Whether that’s accurate is anyone’s guess, but we should soon at least see more autonomous features trickle through upcoming models in South Africa, and a boatload of far-out concepts in the meantime.