AsianScientist (Jan. 19, 2018) – On a recent road trip some friends and I took in Thailand, we navigated dark roads in foggy conditions, drove on back lanes that, if Google Maps is to be believed, do not exist, and practiced the art of overtaking on a single lane highway against oncoming traffic.
When I say “we,” I actually mean our designated human driver, who did a stellar job in our humble rental Toyota. For him, these challenges were nothing out of the ordinary—they were simply par for the course when driving long distances.
Since a road trip also presents the perfect opportunity for bringing up all manner of random topics, we found ourselves thinking about one question in particular: how much longer will it be before we can make the same trip in a self-driving car?
There is no doubt that self-driving cars are the future of transportation. By taking human error out of the equation, they have great potential to make our roads much safer. But full automation—involving zero human input—is still years away: on engineering association SAE International’s six-level classification scheme, with zero being no automation at all and five being full automation, even Tesla’s much-hyped Autopilot system is regarded as only a level two.
Road sense
Despite our limited computing capacities and stubborn propensity for texting and driving at the same time, humans still do a lot of things better than the current generation of autonomous vehicles.
For one, our eyes, ears and brains are better at understanding the world than the sensors self-driving cars are packed with. While self-driving cars require clear lane markings, humans can quite effortlessly negotiate roads on which these markings have faded or disappeared altogether. Self-driving cars can detect large animals like deer, but get confused by objects that move like nothing they have ever seen before; humans, on the other hand, can tell at a glance that said strangely moving objects are in fact kangaroos, the number one cause of animal-related traffic accidents in Australia.
Of course, the senses of self-driving cars are improving all the time. Longer-range, higher-resolution LiDAR and radar systems and better computer vision algorithms are making them more adept at seeing in the dark and navigating snow, fog and other bad weather conditions, for example. But it will be some time before these cars will be able to adapt as well as humans do to all sorts of unexpected and bizarre situations.
Passing the test of passing
Perhaps the most difficult road trip skill that humans have to master is the ability to overtake slower-moving vehicles. As anyone who has been stuck behind an unwieldy RV on a single-lane dual carriageway will understand, this is a necessary maneuver for getting to your destination in good time.
The move can be harrowing and unpredictable—drivers must wait for a large-enough gap in oncoming traffic before crossing over to the ‘wrong’ side of the road to overtake, often at high speed or on twisty roads. Still, with experience, most drivers can learn to do so safely.
This is easier said than done for self-driving systems. While some cars now offer automatic lane changing and overtaking functions, these are there only to assist human drivers, and are most definitely not driverless options.
Tesla’s system, for example, requires drivers to first check if it is safe to change lanes or overtake, and then manually engage the turn signal, before it will kick in. The jury is still out on how effective and safe these semi-autonomous features are, with a number of reviewers expressing serious doubts. On this count, at least, humans still leave autonomous vehicles in the dust.
Calling all cars
Human drivers aren’t hermetically sealed off from each other; they can communicate with gestures or by winding down their windscreens to speak. While we don’t recommend using these methods to communicate feelings of road rage, eye contact, pointing or nodding can help give other drivers an indication of your intended actions—you’re letting them cut in front of you, for example.
Similarly, researchers believe that vehicle-to-vehicle communication between autonomous cars is the one strategy in particular that could vastly improve their performance. A wirelessly connected network of cars would gain a more complete picture of its surroundings, and thus better situational awareness. When overtaking, for example, a car could tap into the sensors of other vehicles in front of it to better determine whether or not the coast is clear.
So, when can we look forward to a driverless utopia? According to a report from the New York-based Regional Plan Association, fully autonomous vehicles capable of vehicle-to-vehicle communication should be plying city streets by around 2040. But this will only happen if technology advances as companies promise it will, and if cities invest in the necessary infrastructure—new traffic management strategies, for example—to support a driverless society. In the meantime, don’t throw away your driver’s license just yet—you’re still going to be needing it.
This article is from a monthly column called The Bug Report. Click here to see the other articles in this series.
———
Copyright: Asian Scientist Magazine; Photo: Shutterstock.
Disclaimer: This article does not necessarily reflect the views of AsianScientist or its staff.
