I am clear that my 6 year old will not learn to drive. By the time she reaches the age when she can, she will either not want to drive or will not be allowed to drive. The car will become a mobile office, or an entertainment center and it will be safer for the car not to be driven by her or by me for that matter! The interesting thing is that this move to autonomous cars is already happening. Automobile manufacturers are increasingly adding autonomous features to their cars and regulators across the world are hurrying to bring in new legislation which enables the move to autonomous driving. The bottom line is that this will have a huge impact on the world we live in.

Currently, most new automobiles use some form of what is called Level 1 autonomy, which are basically warning systems such as parking assistance, with the control of the car in the hands of the driver. However, this is changing and we are now seeing driving assistance systems such as the auto pilot function in the Tesla S which can take control of the steering, braking and acceleration/de-acceleration under certain circumstances. Level 2 autonomy, as it is known, is set to become the norm, especially in the upper end of the car market in the coming years.

The step beyond this is known as Level 3 autonomy, under which automated driving systems perform much of the driving with the driver still sitting behind the wheel and able to take control when he or she wants to and if requested by the car. Liken this to a how modern commercial jet is flown today. These features will be available in the new Audi A8 when it is launched next year, assuming Audi can win a country by country approval for what it calls its Traffic Jam Pilot semi-autonomous hardware. What Audi want is permission for the car to be autonomously driven up to a maximum speed of 60km/h if there are clear lane markings and a divided road with oncoming traffic separated by a physical barrier. Interestingly, Audi is also willing to assume legal responsibility for the car when the Traffic Jam Pilot is engaged.

What is also interesting about Level 3 and some level 2 cars is that they are self-learning. The data they collect from day to day driving is then used to improve the autonomous driving experience of other drivers, with the goal of pushing it forward to the point where no driver is needed.  This is an important point. To make the car fully autonomous, the car needs to replicate how the human eye works and how humans learn and make decisions. Put a baby into car and tell it to drive and of course it can’t. Why? Because it does not have the experience and even if the baby could drive it would not be able to recognize the various objects on the road as it has never seen them before. The same is true with an autonomous car. The car needs to learn from its environment so that it can perfectly understand and react to whatever comes its way. The advantage that the car has is that this data and the learning from it, which by law in most countries takes 17 years (until you can get your driving license), can be aggregated across a whole fleet of connected cars and then shared back to that whole fleet, which means that it can learn faster than the human.

Eventually, such cars will become so good at driving that the driver will not be needed. This is what the car industry calls Level 5 autonomy and when that happens it may be that regulators, under pressure from insurance companies, even ban humans from driving! However, there are still some serious technical challenges such as ‘adversarial perturbations’ that need to be overcome to come to this point.  These are basically mistakes that cameras and the recognition software in the car make due to minor modifications in the camera images, such as changing light conditions. A good example of this was in 2016, where the sensors of a Tesla where blended by the sun and the system failed to recognize a truck which in turn led to a crash. However, in the future ‘end-to-end learning’ approaches, where the car teaches itself how to drive based on huge set of collected driving date, should overcome this issue. But we are still some years away as we need more data from real life situations. But it will happen, and within a decade. And fully autonomous driving will radically change our world.

Autonomous driving will reduce the risk of accidents as well as improve the efficiency of travel and logistics, which in turn will be good for the environment and the customer experience.  Furthermore, low cost autonomous cars will broaden the mobility options for people who cannot drive such as the elderly or the young. In addition, autonomous cars will also accelerate a whole range of business models, including taking car sharing to another level. With no driver needed and the ability to operate 24/7, shared mobility will result in the cost per km of being driven in a car plummeting, which may mean that we won’t feel the need to buy a car again. And if we do choose to buy, it will be because we can use it as a personalized office and entertainment space and avoid the monotony of driving a car!

But there are big social questions from the application of this technology, not least of which is the restructuring of the taxi, public transport and logistics businesses. What do all those professional drivers do going forward? And there is the question around security and how to deal with the risk of cyber-attacks on cars. Finally, there will be those of you that say I like driving. And that’s no problem. I am sure you will be able to do so, exactly as those people who like horse riding can do so today.  But to be clear many of us will not be driving cars in a decade’s time.

  • Tags:
  • adversarial perturbations ,
  • AUdi A8 ,
  • autonomous cars ,
  • commercial jet ,
  • driverless car ,
  • end-to-end learning ,
  • level 1 autonomy ,
  • level 2 autonomy ,
  • level 3 autonomoy ,
  • level 3 autonomy ,
  • level 4 autonomy ,
  • level 5 autonomy ,
  • Tesla S ,
  • Traffic Jam Pilot ,

Leave a Reply