BLOG Autonomous Driving

How Safe Is Safe Enough for Autonomous Vehicles?

Merlin Peter
June 18, 2021

Two months ago, Tesla was in the news for killing two men near The Woodlands. The accident made national headlines because local officials investigating the accident said no one was driving the vehicle. Well, this is not the first time, autonomous driving cars are making headlines for crashes. Tesla Model 3 has also caused other fatal accidents in March 2019 while in autopilot mode, causing the death of its driver after it collided with a tractor-trailer crossing its path. 

In November 2018, the Uber crash in Tempe was the first recorded case of a pedestrian fatality involving a self-driving car. Uber self-driving cars were involved in 37 minor crashes before the crash that led to the death of Elaine Herzberg. Google’s Waymo has had 18 minor accidents from 2019 through the first nine months of 2020. 

However, a study conducted by researchers at the University of Michigan states that autonomous vehicles (AVs) got into more crashes overall: 9.1 crashes per million miles driven, compared with 4.1 crashes per million miles for conventional vehicles. However, the ones involving injury were minor compared with the injuries sustained during conventional vehicle crashes. [Source: Gov Tech

This brings us to the question: How safe is safe enough for autonomous vehicles? The answer isn’t straightforward and so we’re exploring different angles to look at this conundrum. 

What Do AV Accident Reports Conclude?

The NTSB initial reports on Uber’s Tempe crash revealed that the autonomous vehicle rightly predicted the crash but left very little time for the driver to take full control, resulting in the unfortunate fatality. Later, the investigations also showed that Elaine Herzberg was under the influence of drugs and also concluded that the driver was distracted by her cellphone device. The speculations showed that the accident could have been prevented if the driver had been attentive and applied brakes 1.9 seconds before entering the straight section or within 3.5 seconds of the [automated driving system] detecting the pedestrian. 

The charred wreckage of a crash involving a self-driving Tesla vehicle is seen on April 17, 2021. (KPRC)

As for the recent Tesla crash, Elon Musk confirmed in a tweet that the data logs showed that the autopilot was not enabled and that the car hadn’t purchased FSD. Detailed investigation reports are still underway. 

| Read Waymo’s crash reports summary here.

Most of the AV accident investigation reports indicate that self-driving cars are effective in accurately predicting an incident and handing over control, but the increased reliance on autonomous systems makes drivers less vigilant of the situation ultimately leading to fatalities.  

| View autonomous vehicle collision reports here.

Will Self-Driving Cars And Automation Reduce Accidents?

According to NHTSA’s Fatality Analysis Reporting System, 36,560 people died from motor vehicle crashes [conventional driving] in 2018. While this number slightly decreased in 2019, there were still just over 36,000 traffic fatalities. 

Moreover, car-related accidents are not only detrimental to the driver and passengers in the car but also to other vulnerable road users like cyclists, pedestrians, and motorcyclists. The approximated causes for car accidents include drunken driving, speeding, distracted driving, and driver drowsiness. 

In 2020, however, fatalities during the lockdown period plummeted on average by about a third. Some countries reported nearly 80% fewer crash victims in April 2020 compared to April the previous year.

Because 94% of serious car accidents are caused by human error, self-driving car proponents believe that removing human drivers from the equation will save many lives.

Automated vehicles are anticipated to greatly improve the safety of our roadways. Although FSD cars aren’t available for widespread public use, related technologies like adaptive cruise control, automatic emergency braking, and lane departure warning systems are common additions to new vehicles sold in the U.S.

Legislations around the world have also recognized the importance of driver monitoring systems. The EU, China, and the United States have also made these systems mandatory in vehicles. These measures may drastically decrease the dangers of driving than ever before. 

If Zero Fatalities Aren’t Realistic, How Safe Is Safe Enough? 

The focus for AVs many years ago had been to first allow cars to ply on their own. Back then the race was about developing and fielding AI systems. However, over the last decade, as the promise of self-driving cars began to take swifter turns, safety has taken precedence over any of the other economic benefits of the technology. 

Levels of Autonomy

The varying levels of autonomy clearly indicate what one can expect from autonomous driving systems at different levels. Self-driving cars aren’t a reality yet and there are many predicaments concerning co-sharing driving tasks in L4, the co-existence of self-driving cars alongside conventional ones in L5 along with security vulnerabilities, legalities, regulation, and more. 

Alongside these dilemmas, testing safety is also a big challenge. Gathering enough data to prove self-driving cars are safe require billions of miles to be driven. It’s an expensive task, which is why researchers are finding other ways to validate driverless car safety, such as computer simulations and test tracks.

But when it comes to the safety of self-driving cars, it’s always going to be a relative measure. Safety thresholds vastly differ for all individuals — and so in the case of self-driving cars, it will never be a utopian scenario of zero fatalities. Most people believe that driverless cars should at least be as safe as conventional ones and they probably need to do better. A study conducted in 2017 by the RAND Corporation found that deploying cars that are 10 percent safer than human-driven cars will reduce fatalities and save more lives than waiting for systems to be 70 or 90 percent better. The relative decrease in fatal accidents could be reason enough to continue pursuits for L5 autonomy. 

Automakers have the increased responsibility of educating their users about the functionalities without deception. Last year, when Telsa sent out its "full self-driving" software to a small group of owners to test it on public roads, they buried the disclaimer that the $8,000 system doesn't make the vehicles autonomous and drivers still have to supervise it.

Experts say that irresponsible marketing could make the roads more dangerous when these systems become more publically available.

Self-Driving Cars and Autonomous Systems: Pre-Covid vs Post-Covid Public Opinions

Pre-Covid 19 Take On SDCs

Previously, people had many apprehensions about the effectiveness of autonomous vehicles and self-driving cars in particular. A survey conducted by Morning Consult in 2017 revealed that only 16% of respondents were very likely to ride as a passenger in an autonomous vehicle, while 28% of respondents stated that they were not likely at all. Additionally, only 22% of respondents believed self-driving cars were safer than the average human driver, while 35% of them believed self-driving cars are less safe than the average human driver.

Furthermore, the US citizens were also concerned about driver’s vulnerability to hackers, laws around autonomous driving safety, and legal responsibilities during a self-driving car crash. 

Post-Covid 19 Take On SDCs

Last year, Lyft and Aptiv successfully provided 100,000 commercial robotaxi rides in Las Vegas. 98% of these paying passengers have given 5 stars to their self-driving ride experience, with most stating this first-of-a-kind experience is something they are eager to try again.

An Autonomous Delivery Systems: Consumer Awareness & Favorability Study conducted by the Consumer Technology Association (CTA) in 2020, found that U.S. consumers’ enthusiasm for autonomous delivery technologies – including drones, self-driving vehicles, and robots – is growing during the COVID-19 pandemic. 

The study found that 26% of consumers viewed autonomous delivery technologies more favorably than before the coronavirus health crisis. Additionally, 49% of US consumers rated autonomous delivery technologies as somewhat or very favorable.

“With the increased need for contactless delivery, consumers are getting the first-hand experience and becoming more familiar and comfortable with autonomous systems such as self-driving vehicles and drones. Our data shows that these emerging technologies are playing an important role in delivering essential groceries, household items, and medical supplies.”

~ Lesley Rohrbaugh, Director of Research, CTA

We’re Keeping Our Eye On The Road Ahead

It’s true cars today have impressive technology, but L5 autonomous vehicles that are safer will be a remarkable breakthrough in the mobility domain. We’re seeing huge investments flowing into the research and experimentation of self-driving vehicles. Tech giants like Waymo, Tesla, Uber, Audi have raised billions of dollars for expanding self-driving technologies. 

Starts ups like Cruise, Rivian, Nuro, and Zoox have also had multiple rounds of investments coming in. More recently, Cruise got the California PUC permit to test AV with passenger under state's first AV pilot program. It became the first company to acquire both Cal DMV [to test vehicles on road] and Cal PUC driverless permit [to deploy passengers].

LiDAR companies are also being heavily funded. They are continuously improving sensor capabilities to assist better AV systems. We are looking forward to developments in the field. 

As with any technological advancement, there are many conundrums surrounding the promise of a driverless future — we’re excited to see it unfold. 

Tell us what you think at hello@playment.in