“Autonomous driving will gradually become reality.”
MR HERRTWICH, AUTOMOBILES WILL INCREASINGLY ALLOW AUTONOMOUS DRIVING WITH VARIOUS INTERIM STAGES. ARE MACHINES GOING TO TAKE OVER CONTROL? HOW WILL THE RELATIONSHIP BETWEEN HUMAN DRIVERS AND VEHICLES RUNNING WITH VARYING DEGREES OF AUTONOMY DEVELOP?
No, machines are not going to take over the initiative. The systems in all vehicles of the Mercedes-Benz brand are designed to put drivers back in full control as soon as they want to take over. Our systems a fully dedicated to providing support and relief. The transition from partly to fully autonomous systems is not only a matter of the technical capabilities of the systems, but goes hand in hand with the driver’s growing trust in such automation. Once you personally experience that such a system works, then you trust it in more and more situations.
IS GAINING TRUST IN THE TECHNOLOGY A PROCESS THAT TAKES WEEKS OR MONTHS?
It’s much faster than that. We noticed this in the people testing our prototypes. They monitor the vehicle at the beginning of the drive very closely and keep their hands on the steering wheel all the time. But after an hour they have gained enough trust to turn around in their seat to talk to somebody sitting on the rear bench. The realisation that partially autonomous vehicles are capable operators soon emerges, even for people who are somewhat nervous initially when the vehicle flashes the indicators at 120 km/h and changes lanes.
ARE THE “DIGITAL NATIVES” THE IDEAL AUDIENCE FOR SUCH VEHICLES, OR DO AUTOMAKERS, TECHNOLOGY COMPANIES AND POLITICIANS STILL HAVE A LOT OF EDUCATING TO DO?
A lot of education is necessary in every regard. The “digital natives” count on technology and are happy to trust in technology in many situations. It is also the group that goes on record saying that it actually has better things to do with its time in some traffic situations than driving personally. At the same time it is especially people with an affinity for technology who always need to be reminded that there will be limits to the system and that man has to intervene now and then. A totally different target group is older people who will reach the age of 60 or older within the next decade. These people also are experienced in using technology and the internet. For them, autonomous driving systems will fulfil a different function: These systems will provide support for their own, unfortunately declining sensory capabilities in traffic and thus make safer driving possible.
HOW DO YOU ENVISION THE INTERACTION BETWEEN PARTIALLY AUTONOMOUS OR HIGHLY AUTONOMOUS VEHICLES AND THOSE STILL UNDER FULL HUMAN CONTROL?
For a significant period, if not forever, we will have mixed operating modes, on the basis of the existing mix of older and newer vehicles alone. With regard to autonomous driving this situation presents no problems whatsoever. Our vehicles are designed to travel completely autonomously using their extensive on-board systems and are therefore able to easily handle the most diverse traffic situations with their own intelligence – we do not anticipate a concept for autonomous driving that features a central computer somewhere that controls all vehicles. However, it would be conceivable that autonomous vehicles will temporarily log on to such a system and thus make driving in convoys on the motorway possible, for example. We know from experience that vehicles remain on the market for a long time. Cars are also supposed to retain their value. When all is said and done it is also completely legitimate that there will be people who don’t want to use an autonomous car.
AS A CONSUMER WILL IT BE POSSIBLE TO BUCK THE TREND, IF MORE AND MORE TECHNICAL FEATURES WILL BE REQUIRED BY LAW, FROM THE BLACK BOX TO THE REAR-VIEW CAMERA?
You’ll ‘be hard pressed to find a more recent car without electronics and assistance systems today. When it comes to tyre pressure sensors or event data recorders, there is a public and judicial interest in preventing certain risks. However, at this point in time it would be premature to prescribe automation in any way. There are no indications that only full automation guarantees sound, safe and efficient traffic.
WHAT EFFECTS WILL MANY CONSUMERS’ SOMETIMES EXAGGERATED EXPECTATIONS OF AUTONOMOUS DRIVING HAVE? ARE GRANDIOSE PROMISES SUCH AS BLIND PEOPLE BEING ABLE TO DRIVE A FEW YEARS FROM NOW A HELPFUL INCENTIVE OR MORE OF A HINDRANCE?
On the one hand, it is of course nice to use striking promotional videos to show everything that might be possible. On the other, it would only be fair to admit at the same time that the initial systems will not make such comprehensive visions a reality. Due to technical limitations they will still require the driver in many situations, and from a legal perspective also, we are not yet at a point where the lawmakers are prepared to wave through all the developments in autonomous driving that are already a realistic proposition today. One always has to be careful not to over-hype such innovations, otherwise the first production versions will inevitably lead to disenchantment. Even worse: in that case even what is technically feasible and truly impressive suddenly becomes unsatisfactory.
CONSEQUENTLY, IT IS NECESSARY TO MANAGE CUSTOMERS’ AND CONSUMERS’ EXPECTATIONS …
Indeed. That’s why we keep stressing that autonomous driving will gradually become reality. Initially we will drive autonomously on certain classes of roads, starting with the motorway and maybe only under certain weather or lighting conditions. Somebody tearing through dense fog at night is not part of our vision. In the beginning the driver will also have to monitor the system rather than grabbing a book and tuning out completely.
DO YOU SEE A STEADY DEVELOPMENT PATH FROM DISTRONIC, OTHER DAIMLER INNOVATIONS AND TODAY’S STATE-OF-THE-ART ASSISTANCE SYSTEMS TO TOMORROW’S HIGHLY AUTONOMOUS DRIVING?
Various control systems represent significant historic milestones. The Electronic Stability Program (ESP), for example, was an absolute turning point that allowed the general public to see all at once what automatically intervening systems are capable of. The introduction of DISTRONIC marked another breakthrough, because for the first time it offered a convenience function – automatically keeping your distance from the vehicle ahead. I guess the first autopilot systems will be a similar turning point. Whether they will already be of assistance in stop-and-go traffic or only in driving on the open road is hard to say. These things always only become clear in hindsight.
WHAT CHALLENGES DO AUTONOMOUS VEHICLES PRESENT FOR TOMORROW’S CARTOGRAPHY? DO WE NEED TO COMPLETELY RESCAN THE NAVIGABLE SECTIONS OF THE WORLD IN DETAIL OR IS THAT PROCESS ALREADY ONGOING?
There is no need to rescan the entire navigable planet. But everywhere we want to drive autonomously we must perform new measurements. The question of whether that needs to be done before the introduction of the systems or only when autonomous vehicles will travel a route is still unanswered. I can definitely imagine a vehicle learning a route after having been driven on it once or twice and can then handling it autonomously the next time around.
HOW IS SURVEYING THE GLOBE TO PROCEED, WHERE IT IS NECESSARY AND APPROPRIATE?
A mobile network operator wouldn’t start building the network in the boondocks, but in major cities where the most customers can be found. For autonomous driving this would primarily involve recording the precise courses and lanes of motorways and main roads, because these are heavily travelled transport routes. The map producers constantly scan these roads and their surroundings anyway. If that happens with increasingly better sensors, the precision will improve of its own accord, and details can be incorporated into the existing scanning process of the map providers. Of course it would be even better if we succeeded in applying a self-improving system with a networking effect, whereby the vehicles draw their own maps. That too is no longer a dream of the future.
CAN AN AUTONOMOUS VEHICLE DRIVE WELL AND SAFELY WITH THE DATA RECORDS AND MAPS AVAILABLE TODAY?
Today’s map data used by a navigation system, while definitely highly accurate, is not always accurate enough for autonomous driving and can lead to certain error messages or situations in which the system flounders. For example, the end of a lane may not be mapped accurately, as a result of which the computer cannot reconcile the map and the measurements of on-board sensors. These are situations in which the driver would have to take over. Improved maps will reduce the incidence of such occurrences.
HIGH-PRICE SYSTEMS SUCH AS THOSE ON THE VEHICLES OF THE DARPA CHALLENGE MAKE HEADLINES IN THE AREAS OF SENSOR SYSTEMS, BUT THEY ARE STILL VERY EXPENSIVE FOR SERIES-PRODUCTION USE. CAN INTELLIGENTLY ASSEMBLED ARRAYS OF AFFORDABLE SENSOR TECHNOLOGY AND SOFTWARE CLOSE THE GAP? OR ARE THEY TEMPORARY SOLUTIONS WHILE WE WAIT FOR THE EQUIVALENT OF MOORE’S LAW, UNTIL LASERS BECOME AFFORDABLE?
A rotating laser scanner on the vehicle roof can see all around, covering large distances at excellent resolutions. This is ideal for scanning the surroundings, but from today’s point of view this technology is too expensive for series-production use and due to its design does not offer the prospect of a radical drop in price in the foreseeable future.
WHAT ALTERNATIVES PRESENT THEMSELVES?
We are testing the combination of radar systems together with optical cameras. Smaller laser scanners installed in the vehicle corners could be used in addition. Those are all options for getting the technology ready for market. A crucial aspect is the vehicle’s complete 360-degree vision. Regardless of what technology is being used: ultimately it must not be more expensive than today’s driving assistance systems – a couple of thousand euros, therefore.
TECHNOLOGISTS OFTEN TALK OF “GOOD ENOUGH” WHEN IT COMES TO INNOVATIONS – THAT SOMETHING IS USED WHEN IT’S JUST GOOD ENOUGH. CAN THIS CONCEPT ALSO APPLY TO AUTONOMOUS DRIVING OR DO WE NEED TO DEMAND UNERRING UTMOST PRECISION AND RELIABILITY?
“Good enough“ is simply not good enough; we do need to work a little harder than that. But neither does such a system have to be able to totally master all situations from the start: the partially autonomous stage is a great path down the middle. The driver will for the time being retain ultimate control. At the same time, we as development engineers keep learning with each development stage. That was also the case with other driving assistance systems. For instance, before we designed a fully autonomous emergency braking system, we introduced stages in which Brake Assist only acted when the driver applied the brakes slightly. As soon as a system works to the satisfaction of the customer, it is possible to take the next step and progressively remove the driver from the equation.
HOW LONG WILL IT TAKE BEFORE THE DRIVER CAN BE REMOVED FROM THE EQUATION IN AUTONOMOUS DRIVING?
That depends on the traffic situation. At low speeds, in stop-and-go traffic or when parking it is a matter of years. At high speeds and in complex situations the driver will be involved at least for the next ten years, I guess.
WHAT IS THE IDEAL SITUATION IN WHICH YOU PERSONALLY BELIEVE AUTONOMOUS DRIVING MAKES THE MOST SENSE?
I can think of two situations off the top of my head. On the one hand, I find it immensely nerve-racking when I am in a traffic jam and constantly have to do the same thing with the accelerator pedal, steering wheel and brakes over and over again, just to cover a small distance in a long period of time. And on the other hand, I think it would be great if autonomous driving would make routine trips easier for me – for example, the 15 kilometres I drive every day to and from work. In both cases I’d greatly appreciate the extra freedom or relaxation.