Should Night Vision be a Standard Feature on Autonomous Cars?

Home > News > Content

【Summary】Autonomous cars may have loads of high tech components, but Uber’s accident highlights a critical issue they have with seeing objects in the dark.

Original Vineeth Joel Patel    Jun 04, 2018 10:00 AM PT
Should Night Vision be a Standard Feature on Autonomous Cars?

It's been over two months since one of Uber's self-driving Volvo XC90s got into a fatal accident with a pedestrian, but everyone's still looking into the incident, attempting to get a better understanding of what exactly went wrong. 

Immediately following the accident, experts came out and claimed that the incident could've have been avoided. Cortica, a technology company that focuses on developing artificial intelligence for self-driving vehicles, stated that the autonomous Volvo spotted the pedestrian 0.9 seconds before the impact. That would have put the car 50 feet away, which is an ample amount of space to have slammed onto the brakes or to take an evasive maneuver to avoid hitting the pedestrian. 

Unfortunately, that didn't happen. The NTSB recently released its preliminary findings into its investigation in the incident and found that the autonomous SUV was not programmed to stop. The vehicle's sensors detected the pedestrian crossing the street, but the vehicle was operating on its own, which disengages the Volvo's automated braking system. 

BMW Night Vision.jpg

Uber's Accident Reveals Autonomous Cars Need Night Vision

While some are placing all of the blame on Uber and its self-driving system, the accident revealed another issue with autonomous cars – they're not very good at seeing objects at night. As Automotive News reports, the accident was re-created using heat seeking, thermal-imaging sensors that are used on luxury vehicles and the military. With night vision technology, the pedestrian is recognized by the system approximately five seconds before impact. That would've given the operator, who was looking down at the time of the incident, more than enough time to look up and make an appropriate action. 

The outlet claims that the accident has shown car researchers just how important fitting autonomous vehicles with night vision is. Auto News points towards federal data that claims pedestrian deaths are up 46 percent since 2009. Three-quarters of fatal accidents take place at night. One solution to reducing fatal accidents and making self-driving cars even safer would be to fit all of them with night vision. 

"If you have a sensor that could recognize something living, that information would be extremely useful to a computer," said Jake Fisher, director of automotive testing at Consumer Reports. "But I have not heard much about using thermal imaging to detect objects and know which ones to avoid." 

Thermal imaging may sound like a farfetched piece of technology to add onto already complicated and expensive self-driving vehicles, but some companies are already trying to bring the tech to the cars. As Auto News claims, Seek Thermal, a company that re-created Uber's accident, and headlight manufacturer Osram have both been pushing for driverless vehicles to adopt the technology. While having a few companies push for new tech isn't normally enough for the mass adoption of something, automakers and tech companies have taken notice of thermal and infrared sensors after the fatal accident. 

"The Uber accident really does reflect one of the areas in which we have the greatest number of pedestrian fatalities, which we're hoping self-driving cars can fix," stated Matthew Johnson-Roberson, an engineering professor at the University of Michigan. "Until now, a lot of the research has been focused on using daytime vision driving as the benchmark. This accident highlighted how maybe we need to expand how we think about that." 

Volvo Moose .jpg

Why Self-Driving Cars Have Trouble Seeing At Night

Just like humans, hardware components on autonomous vehicles have trouble spotting road signs, pedestrians, animals, road markings, and other items on the road at night. According to Automotive News, the darkness shrouds objects because of a lack of contrast. While this makes it difficult for humans with even the best vision to see things clearly at night, it's even harder for cameras. The components, as the report indicates, are held back by their headlights, which are only capable of illuminating the road roughly 260 feet in front of the vehicle. That gives humans and robots only a few seconds to react. 

"Human vision is already atrocious at night and we're trying to at least do as well as that and hopefully better," said Richard Wallace, an automated vehicles specialist at the Center for Automotive Research in Michigan. "Better should include night vision. Headlights are only so good and thermal infrared is a very powerful tool that the military uses." 

With the help of night vision, autonomous cars would be able to see twice as far ahead as they can now, claims the outlet. While that's good news for those trying to make self-driving cars even safer, the technology comes at a steep price. Auto News claims that thermal sensing units sell for $5,000 a piece. When you consider that LiDAR, cameras, and sensors to make cars drive on their own are already incredibly expensive, you begin to understand why it's not on driverless vehicles. 

BMW Night Vision 2.jpeg

Is Night Vision Worth It?

"We've looked at it and a lot of our customers have looked at it and it's too expensive for a very minimal benefit," said Dan Galves, a senior vice president at Intel Corporation's Mobileye. "It's not something that's really necessary because optical cameras actually do pretty well at night and you have a radar system as backup that is not affected by light."

Galves does have a point. LiDAR systems bounce laser lights to make sense of what shapes objects have and where they are located. Those lasers, though, can't determine whether those objects are living, which is why autonomous vehicles have trouble detecting animals and pedestrians. 

"For LiDAR, the question is, ‘Is it a fire hydrant or is it a 4-year old?'" said Tim LeBeau, vice president of Seek Thermal, who's trying to market the brand's infrared sensors that are currently being used by law enforcement. "With fire hydrants, you can predict what's going to happen. Four-year-olds, you cannot." 

While night vision isn't exactly new technology for cars – General Motors introduced the tech on the Cadillac DeVille back in 2000 – it's unlikely that more brands will adopt it in the near future. The issue, besides the cost, also has something to do with durability. 

"Night vision cameras – like all pieces of hardware is automated driving – have their benefits as well as their drawbacks," said Ellen Carey, a spokeswoman for Audi. "This specific technology will need to overcome challenges of cost, field of view and increased durability to meet the stringent criteria for automation-grade sensors." 

Autonomous cars, though, are supposed to be safer than human drivers. And while automakers and tech companies have been quick to reveal that cars can drive down a stretch of road on their own, finding a solution to driving at night will continue to be a pressing issue. 

Prev                  Next
Writer's other posts
    Related Content