Are you aware that in the US alone, every year, more than 35,000 people die because of motor vehicle accidents?
That’s a scary figure no doubt and one that may lead you to believe that driverless cars are perhaps safer for us. Since self-driven cars will be able to react much faster compared to humans, the chances of lives getting saved are much higher.
What’s so great about driverless cars?
Driverless cars are not driven by humans; so there’s no fear of driving under the influence of alcohol, texting while driving, or driving recklessly because of fatigue. What’s more; they can keep seniors mobile and independent! That’s a huge advantage for the elderly whose cognitive and physical abilities are likely to get compromised with age.
But, the real question is, how safe are driverless vehicles? It’s important to know that one of the biggest problems for the automotive industry has been about handling unprecedented situations arising from edge cases. Today, there are 2 new standards, namely, the ISO 26262 and UL 4600, for addressing such cases. But, regulatory agencies are not forcing compliance with these.
What things should you consider when stating that driverless cars can be safer?
- Levels of Automation: What exactly is meant by self-driving? That’s the first question to ask yourself when debating about safety issues of driverless cars. According to the Society of Automotive Engineers, there can be 6 levels of automation. For levels 3-5, the driver is not required to look at the road; he may read a book or watch a movie but should take control when warned by the vehicle.
On the other hand, consider the Tesla; this car comes with driver assistance capabilities to keep the car inside its designated lane, speed up, and brake as and when needed. But, the driver cannot relax and read a book. He has to monitor the road at all times and be ready to take control of the car when required. Such vehicles are Level 2 vehicles and not ADSs (Automated Driving Systems). So, not all driverless cars can run without the driver’s involvement.
- Commonsense reasoning: When we drive, we use our common sense to swerve or navigate the road better. For instance, what do you do when a deer suddenly darts in front of your car? You swerve automatically. These edge cases aren’t taught in any driving school; these are decisions that are based on our instincts and common sense. This helps us to predict actions. For instance, when a ball rolls out suddenly onto the road, you can expect children to be chasing it. You automatically change the driving behavior in anticipation. But, how do you build this reasoning into cars? ADS developers now have to focus on anticipating and coding every possible situation that can arise. Almost each one of us has had an unusual driving experience sometime in our life; how can we account for billions of edge cases? This is where it’s hard to understand how the ADSs can anticipate and prepare the car for every such situation.
- Ability to “see” like people: Can ADS cars see like you or me? Not really because computer vision systems have been prone to mistakes. They can be cheated in ways humans cannot. Researchers have proved that very minor changes in a speed-limit sign can be misinterpreted and read wrong by machine learning systems in cars. Did you know that Tesla’s autopilot feature has been tricked by hackers who used bright stickers to give the image of a fake lane? Such changes can easily fool cars, but not when you have humans behind the wheel.
All this tells us why testing has become more important than ever. There are solid reasons to suggest that ADSs can be safer compared to human drivers and equally solid arguments against it. Given that 94% of crashes are because of manmade errors, automated vehicles are certainly safer. These vehicles have the means to eliminate human errors and protect the driver and passengers, even pedestrians and cyclists. Moreover, ADSs will never get tired like humans, or text while driving. That makes us think they must be safer compared to human drivers. At the same time, no one yet knows how to integrate common sense into a computer. ADSs can only handle unforeseen edge cases which have been already coded into their software, not otherwise. Accidents are possible if the situations haven’t been anticipated by ADS engineers.