Headlightblog.com interviews Dr. Ralf Herrtwich, head of Group Research & Advanced Engineering Driver Assistance and Chassis Systems at Mercedes-Benz, about the future of autonomous cars.
In May 2010 Mercedes-Benz released video and images of its “automated driving” system tests. This test vehicle has a roof-mounted GPS antenna that sends the vehicle’s position data to a control station.
It sounded like something out of a science fiction novel.
In June 2011, Nevada announced that it would become the first state to legalize the operation of autonomous cars for the open road. A year later the state granted the first license for such a vehicle to one Toyota Prius, powered by Google.
Autonomous cars — also known as self-driving cars, robot cars, robo-cars, or driverless cars — are vehicles that can navigate on their own using a range of equipment from light detection and ranging sensors, to radar, video cameras and position estimators. While human intervention is required to set a destination, the self-driving car can handle all of the rest. These vehicles offer several potential societal benefits, such as fewer crashes, greater roadway capacity, transportation for the elderly or handicapped persons, reduced parking fees and more efficient oil consumption. And they bring their share of headaches as well: potential computer crashes, large-scale job losses in the services industry, loss of state revenues from licensing, and the list goes on. Lawyers are already perplexed as to how to handle accidents in this brave new vehicular world.
In 1986 Daimler-Benz launched the PROMETHEUS research project (Programme for European Traffic with Highest Efficiency and Unprecedented Safety), with the goal of opening up new perspectives to road traffic of the future. The project was part of the European EUREKA research initiative and ran for eight years; the first results were presented to the public in October 1994. Many of the findings live on in today’s automobiles, for example in the form of active distance monitoring or the automatic emergency brake.
To loosely paraphrase William Gibson: autonomous cars are already here, they just haven’t been widely distributed. High-end car manufacturers such as Audi, BMW and Volvo have been actively testing autonomous cars for decades now. Mercedes-Benz in particular has had a rich history in this area having developed a vision-guided robot van in the 1980s that achieved a top speed of 60 mph. The mid-90s saw the development of the VaMP and Vita-2 twin “semi-autonomous” cars that successfully drove on a three-lane highway in heavy Parisian traffic. The company then went on to produce a fully self-driving car that made it from Munich to Copenhagen and back via the Autobahn, reaching speeds in excess of 105 mph.
But how close are we to actually adopting this technology in mainstream culture? And what will be the role of the human in this “driverless” future? In order to get a deeper perspective, Headlightblog.com recently caught up with Mercedes-Benz’s Ralf Herrtwich. [Full disclosure: Mercedes-Benz USA is a Razorfish client.]
Headlightblog.com: Thank you for taking the time to speak with us. Could you tell us about yourself, your role at Mercedes-Benz and what your involvement has been in the area of autonomous or “driverless” cars?
Ralf Herrtwich (pictured at right): Mercedes-Benz has probably one of the largest research divisions in the automotive industry and I had the privilege of heading some of its most exciting projects over the course of the past 10 years. My teams were, for example, instrumental in connecting every single Mercedes to the Internet. I have now moved on to head our advanced engineering projects on future driver assistance and chassis systems. We focus on future safety and comfort innovations for Mercedes-Benz – and this involves more and more autonomous vehicle functions.
HLB: Many of our readers might be unaware of this, but Mercedes-Benz has been a pioneer in autonomous vehicles, having developed a robot van in the 1980s and you are also using autonomous vehicles for testing. Could you speak a bit about your history in this area?
RH: Indeed, this technology features prominently in our research history. Its roots can be traced back to a European research project called Prometheus, which examined vehicle automation in a big way in the late 80s and early 90s. Many ideas at that time were a bit premature given the state the technology was in. For example, there were tests on communicating vehicles – but as cellular was not even barely rolled out then, vehicles were supposed to signal their status with their headlamps. But Prometheus pushed some technologies just far enough to warrant their use in production vehicles. Our advanced cruise control, DISTRONIC, resulted from this project. And DISTRONIC is the foundation for many autonomous driving functions in Mercedes vehicles ever since. We gradually extended it, both for safety, now automatically and fully braking the vehicle in case of vehicles, pedestrians or any other obstacles in its path, and for comfort applications such as automatic Stop&Go in traffic jams.
HLB: Your company has said that the future of safety testing will be driverless. But how about driving itself? Looking ahead a few years, what is your sense as to the viability of driverless cars for consumers? How will the role of the driver change in the meantime?
RH: Automated safety testing is a no-brainer for two major reasons: First, while driver assistance systems are still under development we may not yet have every part of their implementation right. We do not want to put any of our test drivers at risk while we are perfecting our systems. Second, if you test a new version of an implementation of such a system you want to recreate a previous traffic situation exactly as it was before so you can see whether any errors you wanted to correct are actually gone. No human test driver is able to replicate a traffic situation in the same way a robot vehicle can – and we can even do it for a group of robot vehicles. Needless to say, automation of vehicle testing provides us with a few clues on where automation can work for regular drivers, too. While we do not yet aim at automating each and every step of regular driving, there are certain maneuvering situations at lower speeds where automation makes perfect sense. And there are also certain situations in highway driving, pretty much straightforward extensions of today’s DISTRONIC, where automation is viable.
HLB: What problems will driverless cars and autonomous vehicles solve for customers and society in your opinion? What is the business opportunity for Mercedes?
RH: Whenever our drivers do not want to drive and wherever our systems are able to do it, we can presumably think of automation. I think nobody really wants an automated drive on the winding Pacific Coast Highway where driving is sheer pleasure. But not all of our driving is like this. Being stuck in the same traffic jam every day when you go to work, many of us would actually welcome the vehicle to take over. This is where – once the technology is mature enough – we eventually want to offer our customers the choice. They will be relieved from stress and routine, and arrive more relaxed at their destination. But well before we achieve automation of this huge scale, let’s not forget the autonomous features already in our vehicles today. We have autonomous emergency braking down to an art now, covering all sorts of scenarios where an inattentive driver may have overlooked something. Here, autonomy leads to more safety for everybody on the road. And safety is obviously one of our Mercedes-Benz core values, something customers rightfully expect from us.
HLB: Mercedes-Benz recently announced that its 2013 S-Class will feature an autonomous driving system. Could you tell us more about this?
RH: Well, have we indeed? For now, let us only say that the new S-Class will push the envelope in terms of comfort and safety – and yes, also in terms of autonomy – even further. We revealed our F800 research vehicle some time ago and it provides you with a few hints on what to expect. For example, DISTRONIC Stop&Go on that vehicle was extended to Stop&Follow: in a traffic jam your car does not just control the distance to the preceding vehicle, it also follows its steering movements as long as the vehicle stays in lane.
HLB: One of the most significant obstacles to the proliferation of autonomous cars has been that they are illegal on most public roads. Do you expect to see licenses for self-driving cars as we have with Nevada, in the case of Google?
RH: What Nevada regulates, for now, are the conditions for testing autonomous vehicles on public roads. The new thing is that these regulations take the form of universal legislation, not just for Google, but for everyone else as well. Not that we could not perform such tests before, but we always needed to obtain special permissions from the authorities to do it – as we did, for example, for our autonomous test vehicles in Germany. Now Nevada made this a bit easier and it looks as if other states are going to follow. And the legislation that currently holds for test licenses is likely to set the tone for later deployment of production vehicles. But let’s just hope that all states currently coming up with autonomous driving laws think along the same lines; it would be a nightmare if rules for robot vehicles differed substantially by state. One of the challenges for driverless cars seems to be in developing reliable sensors.
HLB: We’ve read that you are developing a 6D stereo vision system, which involves some prediction technology as well. Can you tell us a bit more about your research in that area?
RH: The key to any autonomous function is to give the control software a clear assessment of the vehicle environment. This is what sensors are for. We work a lot with radar and the improvements you observed with DISTRONIC over the past years all have to do with substantial improvements in radar technology. With electronic beam forming and new, broader radar frequencies we expect to get an even better view of the vehicle environment in the future. But we also have very good experience with cameras, only that we feel that 2D pattern recognition will not suffice for the future. This is why we are moving to stereo vision, turning our cameras into 3D sensors. On top of it we developed an image sequence analysis technique which allows us to predict the movement of every single pixel – 3D positions to which we add 3D movements, hence we called this technology 6D Vision. And while I have not given you too many clues on our next S-Class, here is one: 6D Vision is going to be part of it!
What do you think about autonomous cars? Are they viable in the near future? Will the pros outweigh the cons? Would you buy one? Join the discussion!
- Daimler Global Media Site
- Nevada Is the First State to Pass Driverless Car Legislation, Paving the Way for Autonomous Autos; Popular Science, 06.23.11
- Google’s Self-Driving Car Licensed to Hit Nevada Streets; PCWorld, 05.08.12
- Here Come the Autonomous Cars; Fast Company, 10.12.10
- Driverless Cars – the future of automobiles?; Fleetcare, 03.02.11
- A Brief History of Driverless Vehicles; Driverless World, 06.14.11
- Let the Robot Drive: The Autonomous Car of the Future Is Here; Wired Magazine, 01.20.12
- Self-driving Cars by 2013; Drive, 11.15.11
- Where Robot Cars (Robo-cars) Can Really Take Us; Brad Templeton’s website
- Mercedes Uses Robots in Active Driver Assistance System Tests; Motor Trend, 05.11.10