Public trust may be a significant obstacle for the future of driverless cars

In August, a fire truck stopped abruptly on its way to an emergency in a major US city when a self-driving car crashed into its side.

The driverless taxi was part of a rollout across San Francisco that caused chaos on public roads.

So what went wrong? And are incidents like this eroding public trust in autonomous vehicles?

Since 2022, Cruise, a subsidiary of General Motors, has been gradually deploying driverless taxis in San Francisco.

In August, he received permission to operate in the city at all hours, including during the most dangerous peak periods.

In two days, Cruise driverless taxis recorded three incidents: one that failed to stop in front of the fire truck, another taxi that drove onto wet concrete at a construction site, and a third incident where 10 taxis lost connection due to a nearby music festival. .

These the incidents could have been avoided if autonomous technologies had undergone more design and testing.

Loading…

How do self-driving cars work?

Some argue that autonomous vehicles are a safer alternative to conventional driving because they eliminate the chance of human error.

However, recent incidents in San Francisco have many people wondering if there is really any benefit to using autonomous vehicles.

The race to be first in this new industry means driverless vehicles are being deployed prematurely.

When autonomous systems are developed, they are tested against data intended to reflect the environment in which they will operate.

For self-driving cars, the dataset should reflect typical road conditions, such as traffic lights, pedestrian crossings, or emergency vehicles.

The self-driving taxis were showcased at the World Robot Conference 2023, held in Beijing in August.(Getty Images: Zhang Yu/China News Service/VCG)

But it’s impossible to find a dataset that captures every possible scenario a self-driving car could face.

Think about all the unusual experiences a person can have while driving: animals running across the road, other drivers unexpectedly swerving into your lane, and cars running red lights.

Such scenarios explain at least some driverless car accidents, but not the driverless taxi that crashed into a fire truck in San Francisco after failing to yield the right of way at an intersection.

Yielding to an emergency vehicle is not unusual.

In that incident, the driverless taxi entered the intersection on a green light and collided with the fire truck that was heading to an emergency, with lights flashing and a siren blaring.

Although it had the ability to detect emergency sirens, the driverless taxi did not detect the fire truck in time to yield the right of way, Cruise executive Greg Dieterich wrote in a statement. blog post:

The AV (autonomous vehicle) identified the emergency vehicle almost immediately upon its appearance… The Cruise AV did indeed identify the risk of collision and triggered a braking maneuver, reducing its speed, but ultimately did not unable to avoid the collision.

The boundaries of this specific intersection make visual identification more difficult – for both humans and AVs – because it is significantly obstructed by buildings, meaning it is not possible to see objects at the corner until that they are physically very close to the intersection.

Cruise did not say how he would prevent such accidents from happening again.

Such flaws could have been resolved if more time had been required before driverless vehicles hit public roads.

The consequence of this opportunity is a loss of public confidence in autonomous vehicles.

Taxi drivers and other groups have protested regulators’ approval to allow driverless taxis to operate in San Francisco.(Getty Images: David Paul Morris/Bloomberg)

Why is loss of trust important?

Trust in technology can be difficult to build and even harder to regain once lost.

A survey carried out in September shows that three-quarters of Americans are worried or somewhat worried about driverless cars.

Most oppose allowing self-driving taxis in their city.

The situation in Australia could be similar. A recent survey found that 47% of Australians are concerned about artificial intelligence or self-driving features.

Incidents involving self-driving cars threaten to have long-term effects on how people view these vehicles on public roads.

Even before the recent deployment of driverless taxis in San Francisco, incidents involving self-driving cars regularly made headlines. In 2018, an Uber test vehicle crashed into and killed a pedestrian.

In 2021, a Tesla on autopilot almost hit a pedestrian. And in 2022, a Tesla ran a red light and killed two people.

Despite the setbacks in San Francisco, Cruise recently expanded its driverless ride service to two other cities: Phoenix, Arizona and Austin, Texas.

In late September, 20 driverless Cruise cars caused traffic jams in Austin, generating outrage and even more bad publicity.

A trial-and-error approach to driverless cars on public roads doesn’t seem like the best way forward.

Can driverless cars work in Australia?

Deploying driverless cars in Australia won’t be easy.

While semi-autonomous cars capable of driving themselves in certain situations are becoming more common, Australia has yet to see major trials of fully autonomous cars on public roads.

Australia’s most advanced driverless car, ZOE.2, recently took to the outback roads of Mount Isa.

This trial with a single autonomous vehicle traveling on uncongested roads is a step towards driverless taxis in Australia – but there is still a long way to go.

And because there is a larger market for driverless cars elsewhere, these autonomous technologies are being designed overseas, without considering Australian road rules.

Australia is made up of a patchwork of official road rules and standard road practices that differ from state to state.

In Victoria, motorists can make U-turns at all intersections unless there are solid lines or a “no U-turn” sign. In New South Wales, a U-turn is not permitted unless otherwise signed.

States and territories don’t just have different road rules. They also have different habits.

Driverless cars trained on US road rules could struggle to navigate Australian streets.(ABC News: Chris Sonesson)

The difference in how people merge lanes is a telling example of how driving expectations can change from one jurisdiction to another.

For example, I was taught to merge at the end of a merge lane, but in Queensland I observed people merge well before the end of the lane. Each state tends to have its own idiosyncratic approach.

These types of unregulated labels are difficult to capture in a standalone system, particularly if that system is not designed in the country in which it intends to operate.

If driverless cars are to succeed in Australia, they must be designed and tested specifically for Australian roads.

It will take time to achieve this to standards that will satisfy regulators and the public.

Are driverless cars really a safer alternative?

The California Department of Motor Vehicles estimates that annual traffic collisions will decrease by at least a third with the use of autonomous vehicles.

What’s missing from this estimate are the assumptions made to reach that one-third figure.

Does this assume that autonomous vehicles do not share the road with human drivers, who can behave erratically and confusingly?

If so, this is an unlikely scenario, at least for the next two decades.

A more likely outcome would be a combination of driverless cars operating alongside conventionally driven cars. And if this is the case, some mutual misunderstanding between drivers and driverless systems is to be expected.

We should expect more crashes.

The prejudices people form around driverless cars will affect how they drive alongside them. And once formed, these prejudices will be difficult to shake.

The race to be first has already given the public numerous examples of how these technologies don’t work on public roads.

If driverless cars are to be part of our future, we need to think about how they can be deployed safely.

Zena Assaad is a senior research fellow in the School of Engineering at the Australian National University. She is also one of ABC RN’s Top 5 Scientists for 2023.