Roads can be safer by analysing consumer behaviour, start-up co-founder Andre Azevedo tells Louis Bedigian.
The auto industry is very focused on how autonomy, ADAS and other technologies will improve road safety. If a driver fails to brake fast enough, for example, the car could take over and prevent an accident. If another driver starts to veer out of his lane, the vehicle could bring him back automatically. These features are the tip of the iceberg for what will be possible in the not too distant future. Many automakers hope to bring partial or full autonomy to the masses over the next few years.
All of these technologies have one thing in common: they are dedicated to reducing or eliminating the need for human involvement. This leads to a very important question: what can be done to improve road safety for those who want to stay behind the wheel?
HealthyRoad, an automotive start-up out of Porto, Portugal, might have the answer. The company is developing facial biometric detection software to monitor drivers, determine their state (ex: are they drowsy or distracted?) and send alerts when necessary. “The goal here is to understand the focus of the driver,” said its co-founder Andre Azevedo. “If he’s looking at the road all the time or if he’s distracted, texting, or falls asleep because he’s too fatigued.”
The company hopes to get a full picture of the driver’s state by implementing emotion detection and heart rate measurement. Wearables could feasibly provide some of that info but the start-up’s tech does not require the user to buy an extra device. It can monitor drivers simply by plugging its algorithm into a camera.
Said Azevedo: “The goal is to understand the stress, fatigue and emotion of drivers in the future. Today our product is wholly our algorithm, giving intelligence to cameras. We are not looking to integrate with wearables but, if it’s necessary, we can do that.”
Identifying fatigue is only half the battle. Once a problem is detected, the algorithm must take action to prevent drivers from hurting themselves or others. The tech currently alerts drowsy or distracted drivers with a beep but the company plans to partner with automakers (including BMW, which is already on board) for deeper integration.
“We are pushing to integrate the technology in the vehicles,” said Azevedo. “Your steering could vibrate to automatically enable autopilot because you fell asleep. It could stop the car and try to wake you up without the risk of crashing. Those are the kinds of things we are looking for.”
The company’s long-term plans extend well beyond safety and envisions a future where fully autonomous vehicles make it possible for drivers to kick back and relax. At that time, comfort – not safety – may be consumers’ primary concern during their daily commute. Its algorithm would then be used to quickly identify drivers and their preferences, providing a custom experience for each passenger. This may include different driving modes that cater to a passenger’s state of mind. One such mode could make it easier for a fatigued passenger to fall asleep. Said Azevedo: “In the future the car will have the capability to be intelligent and understand what passengers want automatically and make suggestions.”
Deeper learning, better recognition
HealthyRoad already uses deep learning to determine who is behind the wheel. This allows the algorithm to differentiate between drivers. “The goal is to have one software that works with, and recognizes, everyone,” said Azevedo. “We use deep learning to achieve this kind of high accuracy.”
Co-founder Filipe Monteiro added that traditional computer vision algorithms were tested before deep learning was used. He said: “We reached good accuracy but never as good as when we started using machine learning on that, with a lot of data that we acquired from our test vehicles. We use the deep learning on the eye state detection, emotion detection and also on facial recognition.”
The benefits of connectivity
The technology is completely offline but the company is already thinking about the benefits of adding some connectivity features. One potential use case involves fleet managers who receive up-to-date information about the mental state of their drivers. “It’s this kind of information that we can share with a fleet manager so that he can better control the fleet,” said Azevedo. “It’s better to have the right employee in the right moment driving that vehicle.”
Fleets are not the company’s target market, however. While the company acknowledges that its algorithm could be beneficial to anyone operating multiple vehicles, the tech is being built for automakers and their suppliers.
Beyond fleet information, the company would also like to alert drivers when a dangerous vehicle approaches. “If you are fatigued, why not send that data to the cloud?” Azevedo added. “Let all the other cars know that information so that when a drowsy driver is passing, everyone is aware.” Azevedo speculated that an alternate route could be provided so that risky drivers could be avoided altogether.
06 Sep 2017 - 07 Sep 2017, CHICAGO, USA
The Connected Car Insurance USA Conference & Exhibition is the largest and most informative forum for executives from across the connected car and auto insurance industries.