I’ve been writing about the driverless car technology revolution for close to a decade now. Between 2015 and 2018 significant progress appeared to be being made. Spurred on by the traditional automakers including General Motors and Ford of the USA, Germany’s Mercedes-Benz, BMW and Volkswagen and Sweden’s Volvo entering the driverless technology race, the early frontrunners like Google’s Waymo and Uber looked like they were putting their foot down.
Uber and Waymo vehicles kitted out with cameras, arrays of sensors and banks of powerful processors were completing hundreds of thousands of miles of test drives in designated parts of California, Texas and Arizona. Safety drivers were still always present, able to take the wheel at a moment’s notice if required. But the PR updates being released by the companies leading the race towards the driverless revolution that would give us back hundreds if not thousands of hours a year spent behind a wheel were upbeat.
The driverless vehicles, or more accurately vehicles with drivers that didn’t have a lot to do. Except always be aware and ready to act quickly and decisively, were gathering and uploading huge reams of data from their trips. That data would help iron out the kinks that would allow driverless car tech to make the leap from ‘getting there’ to ‘arrived’ and convince regulators and lawmakers to take the handbrake off the autonomous vehicles revolution.
There was, however, a major setback in 2018 when a Volvo XC90 car owned by Uber and fitted out with its driverless technology stack hit and killed a 49-year-old woman in Arizona. The Uber vehicle was on one of the many test drives being conducted in the region and while a safety driver was present he had between watching the streaming service Hulu until half a second before the accident occurred.
A subsequent investigation into the incident concluded Uber wasn’t criminally liable for what was put down to a tragic accident. The woman had been jaywalking at night, crossing a road at an illegal point not designated for pedestrian crossings. The conclusion was that a human driver would have been extremely unlikely to have been able to react in time under similar circumstances.
However, the National Transportation Safety Board (NTSB) concluded:
“the system design did not include a consideration for jaywalking pedestrians”.
Police investigating the incident also said they felt the tragic outcome was “entirely avoidable”, and would most likely not have happened if the safety driver had been alert rather than watching TV.
There had been multiple other incidents involving Uber self-driving car technology test vehicles. Reuters coverage from 2019 wrote:
“The NTSB reported at least two prior crashes in which Uber test vehicles may not have identified roadway hazards. The NTSB said between September 2016 and March 2018, there were 37 crashes of Uber vehicles in autonomous mode, including 33 that involved another vehicle striking test vehicles.”
Uber suspended all test drives for some months following the incident and made systems and software updates. But in 2019 Uber chief executive Elon Musk confidently claimed Uber’s driverless technology would be ready for the streets by the next year. Only irrational fear and overly cautious regulators would now hold driverless car technology back from making the roads a safer place for everyone.
In 2020, Musk made similarly bold claims once again, telling the World Artificial Intelligence Conference in Shanghai:
“I feel like we are very close. I remain confident that we will have the basic functionality for level five autonomy complete this year. There are no fundamental challenges remaining. There are many small problems. And then there’s the challenge of solving all those small problems and putting the whole system together.”
But these “many small problems” are proving they collectively represent a huge obstacle. Waymo, the Google autonomous driving unit, and Cruise, that of the USA’s biggest traditional automaker General Motors, have recently received permission to run self-driving taxi services in San Francisco but under tightly controlled conditions.
Cruise can offer its own employees 100% autonomous rides with no safety driver present. But they have to take place between 10 pm and 6 am and the car’s speed can’t exceed 30 mph. Waymo vehicles still need a safety driver to be permanently present.
The difference with this latest ‘milestone’ is that nobody really expects it to lead anywhere fast. It’s not being heralded as a sign we are now on the cusp of the driverless future. It’s just another set of tests on a long, long, journey. We’ll eventually get to fully driverless vehicles but it’s not on the horizon. That last 1%, or even a fraction of 1%, of circumstances that still confuse driverless systems is now, it is clear, extremely difficult to overcome.
The perfect example is San Francisco’s 15th Avenue, which has become a meme for final hurdle driverless technology has to overcome. In October Waymo vehicles kept turning down a cul-de-sac before performing a U-turn at the dead-end and going back out again. The next Waymo car to pass by would do exactly the same thing. They were all being confused by a temporary no entry sign telling vehicles not to enter a nearby street.
It was, in the event, a harmless and amusing fly in the ointment of the test drives. But just imagine what would have happened if every 10th car on the street was a Waymo driverless car. Within minutes the cul-de-sac would have been crammed full of confused robo-cars preventing each other from doing the u-turn needed to get back on track. It would be chaos.
It’s clear that if driverless technology were given the green light tomorrow every busy city in the world would have multiple daily instances of variations on the San Francisco cul-de-sac mishap. Musk and others heavily invested in driverless technology argue more driverless vehicles on the roads will quickly gather the data for the AI powering driverless car technology to iron out the last minor kinks in systems.
But lawmakers and regulators aren’t going to allow that to happen. The kinks will have to be ironed out first. It’s becoming clear that we’ll have to readjust our expectations of the driverless revolution. It will almost certainly eventually happen and for most of us probably within our lifetimes. But it won’t be next year or even this decade that we’ll have apps we can hit to call a robo-taxi to pick us up and take us to work or home from the theatre.
Autonomous vehicles technology is most likely to come onto our roads in stages, starting with autopilot mode on motorways. Driverless (but with a safety driver always at the wheel) lorries, on motorways, and light last-mile delivery vehicles in suburbs can be expected to come first. Robo-taxi services that operate along designated routes and under strictly controlled conditions will also feature.
But the real driverless car technology revolution allowing us to sell off our personal vehicles and switch to on-demand apps for daily journeys seems as far away in late 2021 as it did five years ago. The difference is those in charge of the companies pursuing driverless technology are no longer announcing dates they expect their services to be operational by. The realisation the last fraction of a percent of driverless technology decisions that are wrong when human intuition would probably have been right might take quite a long time to reduce to almost zero has belatedly dawned.