The advances of technology in our world have continued to increase and self-driving cars are the next logical step for our society. This poses a fundamental question: do we know how safe autonomous vehicles can be?
SAE International is an organization who describes a categorization for "levels of driving automation". It defines six levels of automation for cars, ranging from Level 0 (no driving automation) to Level 5 (full automation), transitioning gradually from "driver support features" to "automated driving features". Check our Exponential Chats episode on the topic to learn more about how these levels of automation are classified.
Even though car makers are starting to offer at least some self-driving features in most of their models, and have gone through great lengths to test their safety, none of them have gotten to the point of full automation.
The new advances in cameras and sensors are helping to build more safety into self-driving cars. Car makers such as Subaru, for example, use simple techniques to deliver on basic lane keep assist, adaptive cruise control, and emergency breaking such as a Stereo Camera set that is capable of perceiving depth and distance much better than a single camera. However, determining if an object is an obstruction is still a big question. What if it is a paper bag? Even human drivers would need be able to tell if it is an empty bag or is it a bag full of rocks. Both human driver and AI would either go around the obstacle if it was safe to do so or come to a stop.
The style of a driver can be a big factor in the safety aspects of a self-driving car and even in an environment that the car has been tested, it may not have experienced all possible permutations of possible conditions.
All driving conditions need to be accounted for. If you are driving on roads with different tractional surfaces like rain, snow, or black ice, then most human drivers would drive much slower (at least we sure should). However, if you are driving in poor visibility such as fog, darkness, or dust, a human driver most likely would slow down or even stop. Some self-driving cars are able to use different more advanced sensors which allows the vehicle to safely drive under conditions of low visibility or detection. These sensors and the technology they use help the sensors see better than humans. Some sensors, such as LiDAR, are able to project laser light and determine distance and depth of the area around the car. It can construct a 3D vision of a driving route which is safer than what the human driver could do.
We have all heard about the dangers of distracted drivers. Maybe you got distracted or you were driving behind a distracted driver. Drivers who are distracted account for at least 30% of the traffic accidents in the US. Drivers that are texting, arguing, on their cell phone, or just fatigue can benefit from giving more of the driving workload to the vehicle. Drivers are still required to sit in the drivers seat up to Level 2 and Level 3 of self-driving automation. Level 4 require much less, if any, and Level 5 is designed to require no human driver at all.
We are currently in a transition mode, where only a small pocket of society will embrace fully autonomous cars and the safety of them. Our current road system is in many different levels of conditions. Some roads are smoothly paved, lane markings are clearly seen, and road signage is correct. The reality is that not all of our roads are in this condition. A self-driving car could still be driven, but the driver is required to take responsibility for additional driving tasks.
What are the risks and the threats we expose ourselves to being hacked or having our private information stolen? Self-driving cars are also vulnerable to attacks from a variety of sources which could seriously jeopardize the safety of the passengers, drivers, and other cars near it. Self-driving cars are a bigger extension of a computer which controls a device, namely your car. These computers and the electronics in the systems of the car are also vulnerable to outside threats similar to what a computer is exposed to.
Cyber attackers could get access to information you have given your car access to. Many cars have access to a garage door code that will open their garage door when they are close to their homes, plus all their contacts and saved geographical locations for easier navigation such as home and work addresses. On top of that, an attacker could get access to information such as who you called, where the car has been, and a slew of other information you would not want a thief to have.
Another danger could come in the form of cyber terrorists where an outside person or groups could cause random car accidents or deadly car crashes. Nation states might be in a situation where they could shut down all self-driving vehicles if they wanted to.
So exactly how long do we think that all self-driving cars would be the norm? If you look at the pattern of how long it takes for a whole tier of cars to be replaced, then this might be an indication of the timeframe. If you think about how station wagon cars were very popular for many years, it was probably about 20 or 25 years ago when they started to be replaced with SUV’s. This was a gradual process, but it became a functional necessity for many families.
In the next ten years, we may start to see self-driving cars become more popular and more common. The saturation of the drivers owning and driving a self-driving car may reach a 50% level, but it may be at least another decade before we reach a 90% level. In the meantime, we will have to share the road with a hybrid of cars with all levels of self-driving capabilities.
Where does this leave us today in the automotive market? Can we buy a car today that is currently at the highest level of autonomy? Unfortunately, not right now. Self-driving and more autonomy of vehicles and devices are coming toward us quickly and it is not going away. We are always going to have “purists” out there who will not want to be in a self-driving car. If we look at all the pieces in this puzzle it does make a couple of things clear. Cameras and sensors are more advanced and can react quicker in emergency braking situations. The AI being built into car systems are quickly becoming more mature. Self-driving cars are not prone to being distracted. A computer will be not drink and drive. AI will not suffer from fatigue.
The full change over to autonomous vehicles will not happen overnight. This is not just a simple switch of one or two carmakers pushing for this, it will really require a whole industry change. That will take time. For the next several years, we can expect that we will continue to have a mixture of self-driving cars at different levels of autonomy for many years to come. In the future we may start to see that we will have lanes relegated to just self-drive vehicles and other lanes for car without self-driving automation.
Is our society ready for this change? Are other countries going to be lead in this or are they going to lag behind? The next 10 to 20 years will surely be very interesting. I’m eager to find out what the next chapter in this story will be. Maybe it will be flying cars?