Add Category
Self-driving cars are better at routine tasks but are five times more likely to be involved in an accident in low-light, new research has found.
The comparison of accident data from 2,100 self-driving vehicles (SDVs) and 35,133 human-driven vehicles (HDVs) found the automated cars were generally safer when staying in lane or adjusting to traffic.
The peer-reviewed study, conducted by researchers at the University of Central Florida, also found SDVs were less likely to be involved in rear-end and sideswipe accidents.
"However the odds ratio of an [self-driving vehicle] accident happening under dawn/dusk or turning conditions is 5.250 and 1.988 times higher, respectively," the report said.
"The possible reasons might be a lack of situational awareness in complex driving scenarios and limited driving experience of autonomous vehicles [AVs].
"Improving the safety of Advanced Driving Systems [ADS] … necessitates a holistic approach that involved advanced sensors, robust algorithms, and smart design considerations."
The team also noted their results were limited by the need for more data and the complex nature of some accidents on the road.
"It would also be crucial in the future to incorporate data about right of way at intersections, encompassing yield signs, stop signs, priority signals, and traffic lights," the report said.
Swinburne’s Professor of Urban Mobility Hussein Dia told the ABC previous studies had been based on "speculations and estimations".
"One of the good things about this study is that it's based on real world data," he said.
"It is a very difficult problem to crack, autonomous driving.
"But it's good to see good progress and it's good to see these results because they give people confidence that we are on the right track."
Driverless vehicles have also been the subject of multiple controversies and high-profile accidents around the world.
In 2023 General Motors was forced to recall all 950 of its self-driving cars for software updates, after an accident involving a pedestrian in San Francisco.
The crash in October prompted California regulators to revoke the company's license, stating the cars posed a danger to public safety.
Research by the National Highway Traffic Safety Administration (NHTSA) in the US has previously found self-driving vehicles were involved in double the number of accidents compared to traditional vehicles.
A 2022 NHTSA study found 69.94 per cent of reported crashes involving ADS were Teslas, 273 of 392 crashes.
Paul Roberts, an associate professor and deputy director of the Western Australian Centre for Road Safety Research, said the technology was "constantly evolving".
"We're talking about software that learns as well, so the more exposure it has to road conditions, including crashes, the better it's going to get," he said.
"I think this report really highlights the sort of things that we knew already. There are limitations on the sensors even now … they don't like rain and fog, a lot of the sensors don't work amazingly in those conditions.
"The other thing where autonomous systems were not doing as well were things to do with turning.
"If you think about it, when you're turning there's a lot of situational awareness. Ideally you want to know not only is there anything in your way, but is there someone behind you that might crash into you as well.
"And [whether there] is someone in front of you that might be in some way inhibiting your turning motion. Not just other vehicles but pedestrians, cyclists."
Australia's most advanced fully-autonomous car, the ZOE.2, completed its first trial in the outback in mid-2023.
ZOE.2 was developed via a partnership between the Queensland Department of Transport and Main Roads, Queensland University of Technology and the federal government's transport research arm iMOVE.
However there are still multiple barriers to getting self-driving cars on the road across Australia, including an inability to adjust to changing traffic conditions, hacking concerns and lack of support.
Less than half of respondents to a study carried out by Melbourne's Swinburne University of Technology felt positively about self-driving vehicles.
Around 47 per cent of the 562 people surveyed saw them as a desirable trend.
However 51 per cent believed having an automated car would give them time to complete other tasks while driving.
Professor Dia, who carried out the survey along with fellow Swinburne professor Ali Martin, said trust in the technology was one of the "big barriers".
"The ultimate dream we're chasing is something called Level 5 autonomy, this is when a vehicle can go anywhere, anytime and under all weather conditions," he said.
"The best we have done so far is Level 4 [which is] what you see in videos from the US and China, these robo-taxis which travel within confined conditions.
"Five years ago everyone wanted to develop self-driving technologies, and there were these silly dreams, you know, 'we're going to have these small pods and we're just going to sleep while they transport us everywhere'.
"This is what I call a nightmare scenario, I hope it doesn't happen.
"The way we see them is really supplemental. In the future I'm hoping most people wouldn't own a car and most of us would just order them on an app to take us from A to B.
"This is what we're trying to achieve, less car ownership, more reliance on shared assets and shared vehicles."
For Dr Roberts, the next step forward is standardisation across different systems.
"We've got no standardisation across those kind of technologies, and that will undoubtedly have unintended consequences," he said.
"The really simple example of variability in human-machine interface in driving is one where you get into a Japanese car that has the indicator's on one side [then] into a European car the indicator's on the other.
"Most people would have had that experience, and that's what happens when you've got lack of standardisation.
"It doesn't really matter if you turn your windscreen wipers on rather than indicating, but it does matter if it means you don't really know what your vehicle is doing in terms of its autonomous systems."