We were told self-driving cars were just around the corner, but the road to true automation is proving far from smooth

By Andy Swales

“November or December of this year, we should be able to go from a parking lot in California to a parking lot in New York, no controls touched at any point during the entire journey.”

So said Tesla boss Elon Musk in April 2017 while attending a TED conference to spruik – among other things – his cars’ “autopilot” technology with the promise of a hands-free transcontinental trip by the year’s end. Don’t worry, though, if you can’t recall that milestone moment in autonomous motoring: it never happened.

Like so many other predictions on “driverless cars”, Mr Musk’s words proved far more bold than accurate.

And it wasn’t just the Tesla chief overstating the tech’s state of advancement.

Tesla has promised much but has often failed to deliver

“Since 2015, we have experienced a torturous crawl of slow progress in vehicle automation dominated by the spectre of endless ‘edge cases’,” says a recent paper by consultant Hatch examining policy and regulation reforms for automated driving.

“An alarming number of firms that made ‘driverless is coming soon’ promises have succumbed to acquisition or bankruptcy. The steady stream of ‘next year’ and ‘five years away’ prognostications became a running joke.

“From 2005 to 2015, the progress curve of vehicle automation appeared to rise sharply. Pundits gauged this as ‘exponential’.

“The marketers of ‘driverless’ repeatedly promised its arrival before 2020; 2030 was sometimes identified as the essential end of private ownership, with a tsunami of ‘mobility-as-a-service’ vehicles replacing private [cars] as well as public transportation.”

Indeed, in 2016, the co-founder of ride-share service Lyft, John Zimmer, reckoned that “by 2025, private car ownership will all but end in major US cities”. In 2018, General Motors said it was ready to mass-produce its Cruise cars with no steering wheels, pedals or other manual controls.

Meanwhile, the insurance industry was racing to understand what all this might mean for motor cover, particularly around liability. How to ensure coverage when a car might be more at fault for an accident than its occupants? What if there are no more accidents?

In 2018, insuranceNEWS.com.au covered a report by Barry Nilsson Lawyers warning that “from an insurance perspective, a whole new business model needs to be considered if the driverless vehicle lives up to the promise of removing over 90% of collisions”.

Of course, that promise may yet be fulfilled. But it’s fair to say the revolution on our roads has failed to follow some of the more ambitious timelines.

That pre-2020 period represented the “peak of inflated expectations” for autonomous vehicles (AVs), according to Canadian transport expert Andrew Miller.

“Before the pandemic, people were … hyping up this technology despite the fact it was nascent and unable to deliver a fraction of what was promised,” he tells Insurance News.

Amid that hype – and with investors ploughing huge sums into companies such as Tesla, Cruise and Waymo – the AV makers were missing deadlines and suffering setback after setback during testing and pilot schemes. These included a series of highly publicised accidents.

Late last year, Cruise recalled 950 “robo-taxis” from schemes in US cities after a hit-and-run victim in San Francisco was thrown into the path of one of its autonomous vehicles, which struck the crash victim again and then dragged them before finally stopping.

Dr Miller co-wrote last year’s Hatch paper, titled The Driverless Endgame, which posited that we are now in the “trough of disillusionment” on autonomous driving. According to the Garnet Hype Cycle model from which these terms derive, that means a period in which “interest wanes as experiments and implementations fail to deliver. Producers of the technology shake out or fail.

Investments continue only if the surviving providers improve their products to the satisfaction of early adopters.”

Dr Miller – who now runs Paladin Consulting and is co-author of the forthcoming book End of Driving (second edition) – tells Insurance News: “I think the argument we made in The Driverless Endgame still stands, and the Gartner Hype Cycle remains an effective description of where we are.”

Examples of the “disillusionment” include a claim last year by engineer and former Lockheed Martin safety whistleblower Michael DeKort that AV makers “can never spend the money or the time, or sacrifice the lives” to develop their vehicles enough to achieve safe full automation.

“They’ll make progress, which is why they are where they are,” he told US finance news outlet The Street. “But they will not get far enough to where they’re better than a human.”

In March this year, motor news outlet CarExpert’s founder Alborz Fallah argued the recent demise of Apple’s AV development project meant “in essence, the death of actual autonomous driving for the foreseeable future”.

“Apple was seen by many as perhaps the only company that could bring about real autonomous driving, with a seemingly unlimited pool of money and a market that it deemed big and profitable enough to enter,” he wrote.

What do we mean by “actual” or “real” automation? There are six levels of autonomy, as defined by global standards organisation the Society of Automotive Engineers.

At levels 0 through 2, the driver is in charge, with some assistance from features such as automatic braking or adaptive cruise control. At level 3 a driver can be essentially hands-free for periods but must take control when required.

At level 4, the vehicle drives itself most of the time. Level 5 is fully autonomous – no steering wheel or other controls required.

Levels 0-2 are increasingly common, and Mercedes is selling some level 3 systems. There are also real-world – albeit limited – examples of level 4.

Making progress: Google-owned Waymo’s robo-taxis are now offering automated rides on the streets of San Francisco and a clutch of other American cities

Among them are Cruise, which is soon to return its robo-taxis to US cities in a partnership with Uber, and Google-owned Waymo, which is operating robo-taxis in a clutch of cities including San Francisco.

Professor Michael Milford, from Queensland University of Technology’s School of Electrical Engineering and Robotics, says the US robo-taxi schemes show the technology is progressing.

“There have been some incidents. Cars and fleets have been pulled off the road. Some of the companies have folded, but fundamentally, the scale of testing of legitimate autonomous robo-taxis has continued to grow … and the capabilities of the technology and the use cases of the technology seem to be getting legitimately pretty close,” he tells Insurance News.

The biggest handbrake on a more widespread rollout of highly autonomous vehicles, Professor Milford says, is financial.

“The problem is [the AV makers] don’t earn money, or don’t earn enough money to justify their huge development costs, and their pretty expensive ongoing costs.

“And so the big question is whether there’s enough real financial appetite from the big companies which are pushing these [vehicles] to keep scaling up their deployments. That’s really the crux of the problem.”

He says the near future is more likely to involve “niche” programs.

“The funny thing about hype dying down is a lot of the not so exciting, not so trillion-dollar ideas for autonomous vehicles actually get a bit more resourcing and get a bit closer to reality.

“So things like autonomous shuttles in retirement home communities, or last-mile transport, or in new development communities … They are actually getting more attention now than they did five or 10 years ago, because five or 10 years ago everyone was just assuming Google would have [autonomous] cars everywhere. And obviously that hasn’t happened.”

Professor Milford says widescale private ownership of highly autonomous cars is “unlikely any time soon”.

“There is an asterisk. Tesla, obviously, is the dark horse … and continues consistently to promise amazing outcomes and then doesn’t deliver.

“But that doesn’t mean that won’t happen one day … I think it’s highly unlikely, but they’re taking quite a different approach.

“They have cars and the data at a much wider scale than most other companies, so there’s always the long-shot chance that they actually pull it off.”

He says the “most likely scenario of having reasonably widespread deployment” is the expansion of centrally controlled ride-share fleets like Waymo’s.

“If it was in Australia, there’d be a couple-hundred autonomous robo-taxis operated by a big tech company in lucrative areas, so Sydney, Melbourne and Brisbane, for example.”

We may never see full automation on the roads, but it seems likely that partially machine-controlled driving systems will continue to roll out.

Australia is taking steps to prepare. The federal government recently ended a consultation on its plans for an Automated Vehicle Safety Law that would provide the framework for their use.

And programs such as NSW’s “talking traffic lights” pilot are testing the tech that may be needed to enable smarter cars.

That scheme involves driverless cars interacting with traffic lights to receive information such as intersection layout, timing of signal changes and potential hazards such as pedestrians and cyclists.

And insurers are still considering what it all means for them. A McKinsey report earlier this year warned the industry that “by 2030, roughly half of new vehicles will be electric, nearly all new vehicles will be connected, and some (maybe one in six) will have level 3-plus autonomous driving capabilities”.

Global broker Howden says: “For the foreseeable future, driving will be a hybrid system where both human-driven vehicles and AVs share the road.

“In such a system, AVs cannot eliminate accidents altogether as the AI that controls them operates on assumptions that are occasionally incorrect. AVs complicate the accident and claims landscape by making it more challenging to determine responsibility.”

The Hatch paper – in urging governments to act now on regulations and issues such as liability – says the shift towards automation will be gradual, but it is happening.

“Automated driving is a gradual, generational change and will likely take more than two generations to embed and saturate,” it says.

“The social response to driving automation is nascent, but it has begun. By now, it is impossible for transportation authorities to turn away from vehicle automation and its social, economic and regulatory impacts, however constrained they may be.

“Too much of the technology – its sensors and its collaborative, near-automated driving decisions – has escaped the lab.”

Dr Miller tells Insurance News the US ride-share rollout is a sign of progress.

In the current trough of disillusionment, “people disdain automated driving, despite the fact that robo-taxis are thriving in China, are in revenue service in San Francisco, and are increasingly available across California and the southwestern USA.

“In other words, now is the time to begin to get excited, but everyone did seven years ago, are [now] exhausted, and incorrectly think nothing is happening.”

Also in this edition: