Mercedes-Benz's Michael Hafner talks hands-off driving and semi-autonomy
Tim Stevens is the Editor and Chief.
I'm Brian Cooley, Editor at Large.
And joining us now is Doctor Michael Hoftner who's the director of driver assistance systems and active safety at Mercedes.
That covers quite a bunch of interesting stuff.
Absolutely, a lot of interesting stuff.
Now we've been talking about the e class a lot as one of the major cars here unveiled at the show.
A lot of interesting technology in that car.
Can you walk us through the autonomous functions in that car?
You know in what situations does it work, how fast can you go, and how long can you take your hands off the wheel for?
Yeah, we have introduced the driver system [INAUDIBLE] last in this vehicle and brought the whole automobile to the next level.
It's still semi automated.
This is due to regulations, and also to some technology that needs to be further developed.
We have introduced the drive pilot, which is capable of helping to keep the distance, to steering, to detect speed limits, and automatically take them in your speed and cruise control.
So you get a lot of comfort and support in your driving.
So the car will automatically just [UNKNOWN] based on speed limits, based on distance from other cars, and stay within the lane, right?
Exactly, so if you turn the system on, the car brakes and accelerates on its own.
It keeps the speed, it helps steering, we even were able to introduce a lane-change functionality that supports you in changing lanes on specific roads.
Now, there have been cars already that have had, including many of yours, that have had active lane-keep assist.
But if you take your hands off the wheel they quickly warn you to get back on the wheel.
What's the behavioral programming Of this new system.
Is it going to allow me to keep my hands off the wheel for some duration?
For some duration is correct wording.
We are not allowed, regulatory-wise that the customer has to put the hands off the wheel for a long time.
So we need to remind The driver from time to time to keep his hands back on the steering wheel and the eyes on the road.
But we have an intelligence [UNKNOWN] detection that kind of looks from what the road is.
If it's very straight, we allow a little longer.
And that keeps you up to 30 second or a minute of time until you are still reminded and really have to go back on driving otherwise you would have an emergency [INAUDIBLE]
Is that something that you can disable?
I know Tesla has a little check box that you can actually turn that off entirely if you want to.
No, I'm totally not allowed to disable the [UNKNOWN] even if many [UNKNOWN] would love it but there's something behind this and that is send [UNKNOWN] in to drive up [UNKNOWN] control.
[UNKNOWN] can run in 99
[UNKNOWN] something percent of the driving very accurately.
But at the very moment that there is something, I think the driver needs to be able to take over control.
And there is where the biggest discussion is going on that I hear around autonomous driving, Doctor Hoftner, is the relationship over the next what, over ten years of semi autonomous when the car will at some times need to throw it back to us.
Assume that we're engaged and ready to take that role, that is a treacherous area, going forward.
How are you gonna work that out that you're gonna give me fair notice that I have to get involved, because I assume, it's going to be at a time when there's something urgent about to happen, not something routine.
And what we need to do in order to do the next step is that we really assure that the car is able to drive by it's own for a time, and we probably will start to have the highly automated functions on specific roads only.
So where we really are sure that we know everything that the maps are accurate, that the sensors see enough Or even [INAUDIBLE] connected with the backend server.
So we will know for a time, that we can load the driver to go out of the loop for some time.
And that he is also allowed some time, a few seconds, before he really needs to go back.
We will not ask, [INAUDIBLE] driving, but we will introduce later the driver to Be responsive again within a millisecond.
And will the car rely on any sensors to detect the awareness of the driver, whether it be heart rate or simply to see if they're looking at the road?
We are planning to have some monitoring in a way that the driver is not falling asleep or something.
So yes, because in a couple of seconds We want the driver to be able to take over if length merge, or if there is something if he wants to leave a specific road we [UNKNOWN] do such things from highway only.
So, if you wanna leave the highway we would remind a while before leaving the highway that the drivers would take over control.
So, That is done by the system on it, mm-hm.
Let's turn now from the soft driving tech, we're fascinated by which is your new Car-to-X communication.
The first series production car that has the ability talk to other entities in terms of a digital discussion of it's actual logistics and movement.
What is the core of that, what have you built into the car?
We thought it was about time to stop the [INAUDIBLE] problem and somebody now had to start really establishing the communication.
What we do in this car is that the [INAUDIBLE] communication like over a telephone and smart phone.
To our backend server.
So information out of the car is sent via the communication module in our backend server, and the backend server aggregates and interprets the information, and give it back to other vehicles that need the information, like for example, if we detect icy roads, we send the information up with our radios And then other vehicles that would come to that location get a pre-warning that there are icy roads ahead.
So this is not a formal implementation of the eight two dilemmas car to car standard that's being developed right now.
It's not a direct car to car at the moment, that is going to be another step in the future
With the current system there is a little latency but for many, many functionalities like icy roads or accidents, construction area, it makes a lot tense and you can do a lot good functionalities with the little latency.
All right, so this 2017 E class car that were embedded radio, that is gathering road and condition information that they can determined
Saying that up to a centralized Mercedes server that it share with other 2017 E class at this point.
Exactly that's the way we start but we are open to include like other OEMs or even more vehicle sending up information.
The more vehicles send up information
The more you know about the roads, and the more vehicles can share and use that information.
What do you want to see happen in terms of the regulatory fabric, whether it's for car to car communications, whether it's for rules about autonomy?
I know car makers are pretty disturbed that in the United States we're going state by state with laws and regulations.
I assume you'd rather see a single blanket Federal policy.
And is there, is it feasible to even look for a global set of regulations, or is that asking too much?
It's still a big risk for all [UNKNOWN] to have worldwide regulations.
But I know we have to live with what's happening, and internally we see a lot of support.
From regulatory activities to really make autonomous driving possible.
So it's not that we have to argue too much, it's a common thing.
So there's no point to be unhappy about this It would of course be better if we we had something nationwide or worldwide, always been better because you don't have to develop different technologies.
But for the moment we can live with the progress that's been done.
And one area that we've seen different legislation internationally versus here has been in the ability to park your car without being in your car.
BMW has that in the 7 series internationally Where you can get out of the car, take your remote control on your key fob, and steer the car into your parking garage effectively.
The E-Class has this as well, but on BMW they had to disable it here in the United States.
But, it sounds like the E-Class, you will be able to do that here in the US.
Is there some questionable legality there, or is there any reason that you guys can do it but BMW can't do it?
I don't know the exact reasons why [UNKNOWN] BMW did not introduce the system here.
It's at the end all about your safety concept, probably.
I mean, we are introducing our system with a smartphone communication, and make the driver always being active.
So you don't have just to push something, but you need to do some specific extras in order to keep the car running, and have to observe what's happening.
So probably it's about the overall concept that you have and whether or not that makes it regulatory feasible.
I am [UNKNOWN] for the regulation that would not allow to do that.
What do you tell people that are skeptical about self-driving?
There are many responses we get from users.
They tell tend to fall in three buckets.
One, I'm a good driver, we all think 2. I like to drive, and 3. I don't trust computers.
Those are three pushbacks we get from a lot of consumers.
Especially mainstream people who are not technology-forward.
How do you think your company will address those three areas of, I guess, fear and doubt?
I think all the three opinions are fair to have.
Our answer is You know, it's just a proposal, an offer that we make with automated driving.
All the cars that will offer you automated driving can manually be steered, too.
So there is no need that you have to use it.
You can always drive by yourself if you like.
And want to do so.
And going forward, what's the balance going to be, when we're talking about improving the safety of a Mercedes Benz, what's the balance between active safety and ultimately trying to, you know, providing more air bags, providing safety systems within the car versus more autonomous functionality to avoid accidents, which is gonna have the biggest gains when it comes to actually reducing fatalities and reducing injuries on the road?
The safety functions that we've introduced, they're always active in the background.
They continue to have the biggest impact as driving gets more and more safe.
And we'll continue to push, and move forward with those active safety systems.
And they are also our basis to run cars autonomously.
There is no question, nobody wants to have an accident with autonomous driving.
And we put a lot of effort that this can be achieved also.
It seems as though with the passive systems like air bags, and, or with the anti lock brakes and stability control, we seemed to have hit a floor, and have pushed accidents, at least in the United States, we pushed accidents down about as far as we can, in terms of annual rate.
We're in the low 30,000's fatality per year, and have been there for a few years.
Is the only breakthrough left going to be semi-autonomy and advanced driver assists?
At least the one with the most effect to gain in the next couple of years.
We have introduced our automatic emergency braking systems in the serious in the US and Europe.
And we see a big gain if the car automatically brakes.
So there's so still more accidents that we can avoid.
And the more safety functionality you put into your cars, like for example, we have introduced now an innovative steering system that supports your maneuver if you wanna avoid an accident with some pedestrian.
That you don't over steer or under steer.
And there are still a few ideas left, or many ideas left, of every [UNKNOWN] that can optimize safety functionalities in vehicles.
And on that basis, we can then also start automated driving.
And presumably that then means that things like insurance rates are gonna go down as well.
Can you give me your thought on Questions about liability and what the future holds there when we're talking about an autonomous car crashing into another autonomous car or a non-autonomous car crashing into an autonomous car?
Well you have to design your system in a way that of course you have redundancy in your system.
The system must be absolutely reliable if it run autonomously, no question about it.
So that's why we will also have Of a sensor equipped in the car.
Currently we do have stereo cameras and radars for automatic sighting.
We will add probably LIDARs too.
We will have detect localization algorithms that we exactly know where the car is.
We have accurate maps, so we know a lot of what's happening around the car.
And the sensor controller, needs to calculate, redundant, [INAUDIBLE].
We have redundant sensors, can make decisions to [INAUDIBLE] or something, if one sensor isn't seeing accurately at, due to weather conditions or something.
And also the algorithms run redundantly.
So we really are sure that there is no error in the system.
I was intrigued by remarks at CES by a counterpart of yours in the industry, Gill Pratt over at Toyota.
He believes the biggest work that'll be done to convince buyers, regulators and insurers that autonomous or highly assistive car systems will work, is to expose how the cars make their decision Not just to show great results but to go to the black box in the middle that got there and expose that so everyone can see how they're doing what they do.
What's your take on that?
How useful that will be in this social journey we're taking?
I would support him in saying that we need transparency in what the car does in the various specific situations.
That will help, that we get a common and a social consensus on how we implement algorithms.
So that's a fair way to go.
Another way to gain confidence from the customers of course is that they start trying the systems.
For example in traffic jams at low speeds.
And they gain confidence by using it.
The more you use it, the more confidence you gain.
And I think that's helpful in order to Knowing, and getting to know the system, and then use it.
Yeah, I don't know why the first time you tried adaptive cruise control-
[INAUDIBLE] nerve racking.
Didn't wanna trust it the first time.
Mm-hm, especially when there's a car ahead of you slowing down-
[CROSSTALK] Okay, I've got my foot right on the brake, ready to go, just in case.
But all of a sudden it's like, wow, it really works.
And then all of a sudden you trust it.
It's very, very nice.
What would you say would be the next step beyond where we are now?
So we've seen speed limits increasing for these systems, you can now go faster on the highway in a semi-autonomous mode.
What'll be the next logical step?
Because it won't be an overnight jump from here to a [UNKNOWN] car.
How do you think the system will take the next step from here?
Yeah, the big next step that we wanna make is that we use more map data because we see that map data's also getting more and more accurate.
Today we don't judge it to be good enough for [UNKNOWN].
If there's a roundabout in the middle, but there isn't one, or vice versa, the car would do stupid things then.
[UNKNOWN] a bad thing?
With [INAUDIBLE] technology [INAUDIBLE] can send up information, also, when they are running on [INAUDIBLE] so we really get knowledge and live data where the [INAUDIBLE] and if there is any changes in the streets.
So we will gain accurate maps in the near future, that will allow us to take them as another input into our algorithms and have them even more sophisticated.
And that's powered of course by the here consortium that recently finished the deal to collaborate on strong really excellent map data
That was one reason, yeah.
And as you add LIDAR to these cars and add more accurate sensors, that means these cars can effectively be building and rebuilding those maps as they're driving around.
Or they can be scanning the terrain and beaming that data up to make sure that the map is updated instantly if anything changes.
Before you go, we always like to ask a personal question.
What are you driving from the Mercedes lineup these days?
Well I'm driving currently a fuelless shooting brake.
A very nice car.
See, we don't get that.
No, it's one of those things you guys keep for yourself over in Europe.
A shooting brake.
That would be nice to have.
That would be nice.
Let them know, we want that in the US.
All right, [LAUGH] very good.
All right, Dr. Michael Hafner, he's in charge of autonomous driving and driver-assist, advanced technologies at Mercedes Benz.
Thank you for joining us.
You're very welcome.