Despite the hype about driverless vehicles, Müller said it would "take 20 to 30 years" before they could co-exist with existing vehicles in cities.
"People driving old cars in the middle of cars that are more intelligent and highly autonomous would be a mess," said Müller, who is Audi's head of driver assistance systems.
However he predicted that technology would take over from drivers in simpler situations such as traffic jams and parking "in this decade."
Müller was speaking to Dezeen during the first public passenger event for Audi's two Concept RS 7 cars, which can drive themselves around a racetrack at speeds of up to 220 kilometres per hour.
The cars use a combination of military-grade GPS, images captured by an array of cameras, which the car compares to a bank of images previously captured on the same stretch of road, and a "path finding" algorithm to help plot the best route for each lap.
Müller said that off-road driving would be "difficult" to achieve, as autonomous vehicle technologies rely on recognition of road lights, markings and signage to determine the car's position on the road.
"You have to make it recognise what is a street, what is not a street, and gravel and mud," said Müller. "This is something that in the technical environment – self-learning machines – is still at the beginning."
This year has seen a flurry of proposals for autonomous vehicles, including a self-driving truck by Mercedes-Benz, a Tesla car that can be "summoned" by its owner, and a concept for office pods that can travel to workers, removing the need to commute.
But legal variations around the world and a lack of trust from customers are still barriers in bringing driverless cars onto the market, according to Müller.
Earlier this year, Google was forced to alter the design for its fleet of autonomous cars after California changed its road rules, requiring all vehicles to have a steering wheel.
"It's an evolutionary approach," said Müller. "It's not going to be that next Monday everything is there and is working, because technology needs to be developed, regulation needs to be developed, infrastructure needs to be developed."
"And the most important thing is the customer. My wife wouldn't get in a car that doesn't have anything. She doesn't trust it. You need to take the customer on that journey with you," he said.
Audi said it currently has no plans to sell driverless cars, but the technology it is developing will be used to offer drivers "piloted" options and to create new safety features.
Müller said these would allow drivers to hand over the boring aspects of their journey to the car, such as traffic jams and parking.
"We're not following a strategy of having – some people call it robotaxi – this car driving empty through the city looking for customers. This is not our strategic goal," said Müller.
"Driving pleasure issues – this is something that we focus on. And on top of that comes driver assistance systems, piloted driving, which are the functionalities where we say we want to help the driver – to avoid accidents."
Design studio Kram/Weisshaar worked with Audi for the first passenger trips of the RS 7 driverless concept cars, designing an in-car film studio to capture passenger reactions. The car brand is also working with architects to identify areas where driverless technologies can have an impact on urban planning.
"The logic in that is to try to understand the urban mobility of tomorrow. Which includes, of course, the streets, but it also includes the parking areas. And to find out where we can generate some value," said Müller.
"If you could just drop your car at the entrance of a parking place and it just gets in and you pack all the cars together without any streets in-between, you would reduce the area, which I think is a great thing."
Read the edited transcript of our interview with Thomas Müller:
Anna Winston: You've been working on this piloted concept for just six months. That's quite fast. How many different types of technology are you applying?
Thomas Müller: We had of course some pre-developing activities working on some of these technologies. But it was very very tough.
Actually you take a normal RS 7. So you use the engine, the steering system, the breaking. Then you put in a localisation system – because the car needs to know where it is – which is basically done by GPS and cameras. And then we took the same algorithm that we had in the car that we drove up the Pikes Peak three years ago autonomously but at low speed. It's a kind of path planner. You drive one lap on the left side of the lane; one on the right side. There are millions of possibilities to comb through and there is some intelligence behind that – assuming what is the best way to go through and at which speed.
And you have a lot of safety technology on board – two power supply systems, two breaking systems, redundancies. We have a radio control system taking care of the car. When we drive driverless we have spotters to switch off the car if we have someone running across the street. So a lot of safety things – to make sure that it works and that nothing happens.
Anna Winston: How long would it take to make a car that could drive autonomously on a real road?
Thomas Müller: If you're talking about cities, this is very complex. Cities will take many many years still to be able to do that.
If it's a city, we call it level four. To just get into your place, put on the navigation system – maybe sit in the rear seat and the thing is driving you to somewhere – this will still take 20 to 30 years.
But if you have simpler use cases like a traffic jam on a highway or parking, I think this could very well happen in this decade.
The biggest barrier in the end is that you have a mixture of non-intelligent and more intelligent cars. If every car was intelligent and every car was talking to each other and every car would be... like airplanes. They're all controlled by a system, and the human being is almost taken out of that system – he's just monitoring the whole thing. That would be much faster. People driving old cars in the middle of cars that are more intelligent and highly autonomous – this would be a mess. If you could manage that situation – I think this is one of the biggest challenges.
Anna Winston: What about off-road driving?
Thomas Müller: It's very difficult because one thing that we need to drive autonomously is lights. On a gravel road I don't reckon that you'll have lights. You have to make it recognise what is a street, what is not a street, and gravel and mud. This is actually something that human beings can do very well because you learn it. It's not because you're born with that knowledge.
You learn by driving what is a gravel road, where the borderline of a road is, and this is something that in the technical environment – self-learning machines – is still at the beginning. If you have algorithms that can learn as they go – robots or cars or whatever – then you could get a lot of what human beings have and then eventually, some day, gravel roads will also be possible.
Anna Winston: What is the point of projects like the Concept RS 7?
Thomas Müller: We focus on two things. First of all we have our pre-development activities that we want to take to serious production. One is, for instance, if you have to do an evasive steering manoeuver on the highway. You need to change lane very fast. Most people when they get into that situation – in front of a crash – they are so focused on the car still in front that they don't react. Most of them don't even push the brake, as they're shocked and they know they're probably about to have a really hard accident. This is something you need to do at the limit. So this is something that we’re working [on] also here. Planning, analysing, positioning – where am I? Do you have a clear lane?
These are use cases that we can take out from projects like this. The other one is marketing – showing competence, showing that we are ahead of piloted driving, and that we can manage these cars and that we're having fun with you.
Anna Winston: So there isn't a plan to go fully autonomous as a business strategy?
Thomas Müller: No. Not at all. We're not following a strategy of having – some people call it robotaxi – this car driving empty through the city looking for customers. This is not our strategic goal.
There are two kinds of cars. We have this kind of car that just takes you from A to B – which is unemotional. We want to have our cars be something very emotional. Se we need to have a very, very nice design.
People buy the cars because they look good. That's the first reason someone gets into a dealership. Then he makes his test drive. He needs to be integrated in that machine and he needs to feel it. He needs to have pleasure and say 'Yeah, that's me, this is part of my body'. Driving pleasure issues – this is something that we focus on. And on top of that comes driver assistance systems, piloted driving, which are the functionalities where we say we want to help the driver – to avoid accidents. We want to support him in situations where he doesn't have this pleasure of driving. Like traffic jams as I said, or parking in a parking spot in the morning. So if he wants to drive he will drive. If he wants to have the assistant helping him he will have the assistant helping him. But he's in command.
Anna Winston: In terms of legislation, how far away are we from being able to use a lot of this functionality on public roads?
Thomas Müller: A lot of activity is happening in Europe, happening in the US, some starting in China. In Japan this is also starting now. But it's still very regional, which makes it more difficult. Today you can test, but you cannot sell this technology. Even if it's just a traffic jam pilot or something like that, you cannot put it in a car because someone cannot buy it.
It's difficult to say when it's going to happen. I think it's feasible that it will happen in this decade. But one thing I think is clear: the US is ahead. And they didn't sign the Vienna Convention on Road Traffic from 1968, so they don’t have to wait for this to be changed. They can just start working on their laws, and they're doing that already. The US could be the first country where you could have the first piloted functionalities on the road. And I think the US has a huge potential – think about the highways. All of them doing 45-60 miles an hour.
Anna Winston: California is changing its regulations to insist that all vehicles have steering wheels and brake pedals.
Thomas Müller: They wanted to put a fleet of Google cars on the road without anything, and the [DMV] said 'no forget it'. It’s an evolutionary approach. It's not going to be that next Monday everything is there and is working, because technology needs to be developed, regulation needs to be developed, infrastructure needs to be developed. And the most important thing is the customer. My wife wouldn't get in a car that doesn't have anything. She doesn't trust it. You need to take the customer on that journey with you.
Anna Winston: And you’re working with architects and designers outside of Audi. What's the logic in that?
Thomas Müller: The logic in that is try to understand the urban mobility of tomorrow. Which includes, of course, the streets, but it also includes the parking areas. And to find out where we can generate some value.
Where do we have to focus on if we’re building this technology soon? Having the whole system of cars in this mixed scenario of intelligent and less intelligent cars working to increase the efficiency of the traffic – this would take some years.
But the area that you use today for parking – this, I think, is actually the biggest problem. Cities are reducing the parking areas and so we need to find a way to pack more together. If you could just drop your car at the entrance of a parking place and it just gets in and you pack all the cars together without any streets in-between, you would reduce the area which I think is a great thing.