When Will Self-Driving Cars Become Safe Enough?

Self-driving cars are among the most frequently discussed technologies in the world today. Some people think autonomous vehicles could be life-changing for older or disabled people who can’t drive and have difficulty finding reliable transportation options.

Governments are regularly involved with rolling out legislation that ensures safe usage on the world’s roads. Progress in that effort largely depends on location. For example, people in many areas of the United States and China frequently use self-driving cars now. 

However, the United Kingdom government is still preparing for their widespread arrival, with the only usage so far restricted to tests conducted by researchers.

The hope is that authorities will allow autonomous vehicles on roads by the end of the year — with caveats. For example, cars with automated lane-keeping systems would be the first type legalized. However, they could only operate at a maximum speed of 37 miles per hour, and someone in the car must be available to take over in 10 seconds or less if required. 

Despite these specifics, many people still have lingering doubts about the safety of such vehicles.

Stories of self-driving cars crashing during road tests tend to resurface every time people bring up these technologies. Many of the most frequently cited cases are now several years old, though. Technology moves rapidly, so where are autonomous vehicles today?

As the world gets ready to welcome driverless vehicles onto the road, their safety deserves another look.

What Are the Safety Benefits of Autonomous Vehicles?

If self-driving cars work the way they should, they would be safer than traditional vehicles. That’s because human error plays a significant part in traffic accidents. People get distracted, overlook obstacles, speed, or otherwise make poor choices, leading to crashes.

Tiredness can also contribute to accidents. Statistics show that 60% of adults reported driving while sleepy. As people get more fatigued, their reaction times decrease, and they could even nod off. Similarly, individuals under the influence of drugs or alcohol often find that those substances significantly affect their spatial awareness, decision-making, and other abilities that must stay sharp when people drive. Self-driving cars don’t get tired or become affected by what they ingest.

The computers behind autonomous cars, at least in theory, can react faster than people. It can take humans up to 600 milliseconds to detect and respond to road hazards. An advanced machine vision system can process visual data far quicker than that, theoretically responding to traffic obstacles faster.

How Safe Are Self-Driving Vehicles Right Now?

Self-driving cars have significant safety benefits, but only if they work correctly and consistently. While today’s autonomous vehicles are far from perfect, they may be safer than many people realize.

Self-driving car company Waymo recently published a paper containing data from its vehicle usage in Phoenix, Arizona. From January 1, 2019, to September 30, 2020, the cars traveled 65,000 miles without a human behind the wheel or available to take over driving. Only one collision occurred. 

The California DMV has required autonomous vehicle testers to report collision statistics since 2014. Between then and 2018, there were 104 of these collision reports, and most were minor. More interestingly, 57% of them were rear-end crashes and 29% were sideswipes, both situations where other drivers are to blame.

Considering how there were more than 3,000 fatal crashes in California in 2019, many people might not consider 104 mostly non-fatal crashes across four years a large number. Even so, humans often fear the unfamiliar and unknown. Those tendencies may make them extra cautious about self-driving cars, no matter what the data indicates. 

What Needs to Improve?

While autonomous cars are surprisingly safe, there’s still room for improvement. There have been relatively few crashes involving these vehicles, but that could also be because there are fewer of them. There’s not enough large-scale data to say conclusively that they’re ready for roads yet, and some persistent challenges remain.

The key to safe self-driving cars is faster, more accurate computer vision.

That’s a challenge right now because many lidar systems are uneconomical for self-driving applications since they’re too expensive for consumers. Companies have fast, accurate vision technologies and they have cheap options, but the two don’t coexist yet.

There’s also the issue of poor weather conditions. Sensors and cameras today see significant drops in accuracy in low-visibility situations, which severely limits autonomous cars. 

Sensible 4 is a Finnish company specializing in sensors and software for these vehicles. CEO Harri Santamala clarified:

“Cameras are good in detecting and classifying objects in the vicinity of the vehicle, but they see poorly in the dark or through rain and fog, and they also have difficulty in accurately assessing distances to objects.”

Santamala continued:

“Lidar scanners work regardless of lighting conditions and are able to measure object distance and size down to millimeters, but their performance also suffers from rain and fog.”

Those comments suggest that both technological and environmental factors could affect whether a self-driving car operates safely and reliably. Additionally, researchers, engineers, and other experts are still confirming which systems give the best dependability in all weather and exploring how to accommodate identified shortcomings. 

How Can Driverless Cars Become Safer?

Better AI will lead to safer self-driving cars. The machine vision systems that drive these vehicles have to be able to recognize more objects in more situations and be able to predict their actions.

That will take training - a lot of it. To gather the necessary data in time to release these vehicles within a few years, automakers will need to ramp up their road tests.

Finding a way to balance affordability with accuracy in vision systems isn’t easy.

One possible solution to that issue might be to rely on technology outside of the car itself. For example, navigation app Waze recently added a feature that warns people about upcoming railroad crossings.

In Hong Kong, a contactless smart payment card doubles as a way for older or disabled people to extend the time provided to cross streets safely. Pedestrian traffic poles have card readers installed on them. Once a person taps their card against one, the traffic lights react appropriately. 

These examples both emerged as human-centric upgrades. However, they make it easy to envision how autonomous vehicles could communicate with IoT devices around them to get a better understanding of their environment as smart city infrastructure increases.

Navigation services in other vehicles, motion sensors, and connected traffic signals could all feed data to self-driving cars. This wealth of information can fill in the gaps of a vehicle’s machine vision to provide a more comprehensive picture of its surroundings. More IoT infrastructure will also enable edge computing, which provides the immediacy autonomous cars need when computing machine vision data.

Autonomous Vehicles Still Have a Ways to Go

The automotive industry is closer to road-safe self-driving cars than ever before, but they’re not there yet. Authorities in the United Kingdom recently launched a consultation period where relevant parties can propose changes to the existing Highway Code concerning autonomous cars and their safe usage. 

Transport Minister Rachel Maclean commented on that progress, saying, “This is a major step for the safe use of self-driving vehicles in the UK, making future journeys greener, easier and more reliable while also helping the nation to build back better. But we must ensure that this exciting new tech is deployed safely, which is why we are consulting on what the rules to enable this should look like.”

Despite significant strides, there’s still a long way to go before these vehicles are ready for the road. That said, automakers and other technology companies can speed the process.

As lidar and radar technology advances, people in the autonomous car industry should learn valuable things that help future vehicles operate safely, regardless of the weather. That knowledge, combined with more smart city spending and increased road testing hours, will help make autonomous vehicles safer, faster. When that happens, overall traffic will likewise become safer.