The Next Level of Driver Assistance Will Be Based on AI/ML

Harry Kroeger, Board member and President of Automotive Business at

I have been working in the automotive business for the better part of my career and it’s clear that we are now in the decade of its greatest change. What we’re experiencing is truly like going from the horse and buggy to the evolution of the first car. The big transformation is happening on two fronts: One is around propulsion – the rapid movement going from combustion towards electric powertrains. The other is the amount of intelligence embedded in vehicles to provide increasing levels of driver assistance and passenger safety.

The next level is when vehicles will take you where you want to go without you having to take the wheel. To get there, it will require greater development of artificial intelligence (AI) and machine learning (ML) and much better power efficiency of the technologies used for that.

That’s why I’m excited to join after years of working on early autonomous vehicle developments at Daimler and Bosch. Technology has always been my passion, but I wasn’t interested in doing the same thing with another big auto industry brand. Time is precious, and I want mine spent with the people and work making an impactful difference. After an awesome meeting with Krishna in Stuttgart last year, I was convinced that had the team, vision and technology to enable the industry to take a major leap forward.

The work is doing provides the brain for Advanced Driver Assistance Systems (ADAS), and compared to everything else in the market this will be very disruptive. It’s a thrill leading a team responsible for making autonomous driving on all levels so much
better than anything we’ve seen today. The advancements we are leading will help make cars much smarter and significantly safer.

It has been a long road to get to the current state of driver assistance, but the future can get here faster with more disruption in the market. Innovative developments in AI and ML will contribute to improvements similar to how AI pushed computer chess programs well past the level of those that were taught by grandmasters.

AI and ML Will Take Vehicles to the Next Level

The lower levels of driver assistance are well-established. Level 1 provides functions like help in parking your car. Level 2, which provides partial automation like the standard adaptive cruise control, has been around since 1998.

The very first Level 2 cruise control was based on radar, with developers trying to make an antenna small enough for a car. The chip antenna changed that, introducing a relatively low level of computing power. Automakers continued to add to that power, with the addition of some more computing power, with more sensors, like cameras. However, there were still a lot of false positives and false negatives limiting development.

We’ve seen that the industry has been somewhat stuck based on the prevailing technology. A car company asking suppliers for a new electronic control unit (ECU), for instance, wanted to have the chips for that ECU on the table six years before the car was launched. This car was to be produced for seven years, with after-sales stretching 10 years longer. In a world where chip companies produce new chips every few months, the technology being built into cars got dramatically outdated during the lifespan of a specific car. You could see it in the first navigation systems in cars, which were impressive at first, though soon looked like daguerreotypes from the 1850s compared with the screen of a $100 smartphone in terms of smoothness of rescaling the map, selecting menus, and so on.

I have always been a proponent of getting to a refined Level 3 and affording drivers to not pay full attention when traveling while still present to intervene when necessary. It’s a great customer value, enabling people to ease the boredom of long road trips by reading a book, checking email, or watching something on Netflix. This level of automation can’t be achieved with classic computer vision (CV) technology. I think it’s fair to say that AI and ML will open the door to advancing the market to a sophisticated and accepted Level 3.

We can’t get to the performance jump we need by the standard methods. For Level 3, we’re likely to have systems with two paths. One will use AI/ML for the advanced performance of functions, and another will employ straight deterministic algorithms to make sure you don’t run into a white truck stopped across the highway, blending with the sky – like it happened in recent years with some systems not well designed.

Automotive Is at an Inflection Point offers an alternative path built on innovation that will soon make high-quality Level 3 autonomy a reality. Today, our software centric Machine Learning System-on-Chip (MLSoC) Platform and Effortless ML is empowering auto suppliers with 10x better performance, the lowest power and cost, and the easiest “push-button” software experience this industry has ever seen.

Eventually, we’ll propel the industry forward to Level 4, where the driver can get some solid sleep because the car knows everything it has to do, and Level 5, where there’s no driver at all and people can just sit in the back. The path to those levels depends on perfecting Level 3, and that is where innovations and disruptive work in AI and ML come into play.

It’s the perfect point in time to give automotive a new direction. This is a role I’m honored to take on, as the team at delivers the change this industry has been waiting for.