As the safety ratings of organisations like Euro NCAP and the US Insurance Institute for Highway Safety (IIHS) become harder to achieve, the role of electronics and advanced driver assistance systems (ADAS) will become increasingly important. One of the outcomes of this will be to take the industry ever closer to semi-autonomous and fully autonomous cars. Indeed, safety suppliers like TRW see semi-autonomous and ultimately fully autonomous driving as the logical outcome of safety technology developments.
Google unveiled its self-driving car prototypes in May 2014; Volvo said in June it is preparing to test its autonomous car prototypes in Gothenburg; and at the 2013 Frankfurt Motor Show, Daimler’s Chief Executive, Dieter Zetsche, was driven on stage at his press conference by the Mercedes-Benz S 500 Intelligent Drive, the same self-driving S-Class that famously followed the route taken by Bertha Benz in 1888 when she made the first ever long-distance car journey.
Even BMW, for so long the champion of Freude am Fahren (“the joy of driving”), is developing its own autonomous car technology to relieve the driver of that very task.
Proponents of autonomous cars usually underline their usefulness in allowing drivers to relinquish control of the vehicle to technology designed to work in tedious, repetitive situations like slow-moving traffic jams.
BMW is taking a different approach to reach the same ultimate goal by developing the technology to work at motorway or expressway speeds. “In our opinion, highly automated driving involves the driver being able to press a button when the car is travelling at 130kph motorway speed, and for the car to take over the driving task,” explains BMW’s Dr Werner Huber, Project Manager Driver, Assistance and Environmental Perception at BMW Group Research and Technology. “On the one hand, this involves high complexity due to the vehicle’s speed. On the other hand, there’s reduced complexity due to the type of traffic – there are generally good road markings and road signs, there is nothing coming in the opposite direction, and there is nothing crossing in front of the car.”
Drivers who are used to regular slow-moving, bumper-to-bumper city traffic might consider travelling at speeds of 130kph (81mph) to be enjoyable, rather than tedious. For Huber, “The fun of driving is on the weekend on a rural road, in the mountains. Standing in a traffic jam or driving a few hundred kilometres along a boring highway is not fun. You can use the time for other things. And a long boring trip is not a safe trip. You get tired, you get inattentive.”
Here, BMW’s autonomous car technology can take over what Huber refers to as the relatively simple job of driving – keeping in lane, and not colliding.
Driver taking back control?
Much of the discussion about the role of the “driver” in an autonomous car has centred on whether that driver is able to instantly or even quickly take back control of the vehicle if required. However, even after a few minutes of autonomous drive travel, a driver may be completely unaware of their surroundings. “We cannot rely on the driver in a critical situation; we must rely on a very good car,” says Huber. For this reason, BMW’s vision of highly automated driving involves taking the driver out of the driving loop. “Give the driver the chance to do other things. Then it’s a real benefit. It’s not a benefit to sit behind the steering wheel and just observe what the car is doing. And if the car does 99% of the job very well and in 1% it fails, and as a driver you’re expected to intervene, well, that’s not fair.”
What drivers do with their time whilst being driven autonomously must also be carefully considered. Someone eating or reading a large newspaper, for example, has little opportunity to quickly take over the control of the vehicle.
Achieving highly automated driving is therefore not only a technical issue, says Huber – it’s also about identifying what people can do with their time.
BMW has shown that its highly automated cars are capable of overtaking buses, slalom driving, and even drifting. However, this was to illustrate technical capability, and to illustrate how the vehicle can respond in critical situations. “In reality, highly automated driving is not about drifting,” grins Huber.
“Normally the car will be driving at highway speed, the driver will be in position, and we have to offer them some form of entertainment, or change the interior displays so that he can work in the car. But you have to also offer the ability to quickly regain control of the driving task. That means the controls and steering wheel must be there. The displays must function in a way that they can be used during normal driving, but they shouldn’t disturb the driver when the car is in self-drive mode. Yet the driver should be able to quickly return to full driving mode.”
Much of the technology is already in place
Just as Daimler highlighted that its self-driving Mercedes-Benz S-Class used essentially existing technology, so too is Huber keen to emphasise that BMW’s highly automated vehicle technology is a logical extension of existing technology. “We have automatic gear shift; we have automatic throttle, we have electric power steering and we can talk to each brake,” says Huber. “So the basic technology is already available. We’re also able to drive into a garage by controlling the car remotely with a key. We already have everything onboard every BMW.”
There is, however, a key difference between BMW’s normal cars and the self-driving prototypes: a high-precision differential GPS (DGPS). “GPS is precise to between 1 and 5 metres, depending on the satellite constellation. In this car, if we want to drive say, through a path of cones, it needs to be millimetre accurate,” explains Huber. “That’s why we implemented the DGPS. In a later highly automated car, we will have a very precise map to localise the car. That means that unlike driver assistance systems, we need to install additional and more precise technology to steer the car. But other than the environmental sensors, we don’t need too much more for an autonomous car.”
To a layman, then, it appears to be ‘simply’ a case of making the systems talk to each other, but Huber adds a strong sense of reality to such thoughts: “We don’t rely on that. We have a back-end connection for car-to-X, where the X is called the back-end centre. Through crowd sourcing, cars help to improve the centre’s understanding of what’s going on outside. And the centre is necessary to supply the car with a high definition digital map, which we cannot buy in. For the motorways we are driving at present, we have a specific HD map, but we would need it for the whole network.”
And that needs to also be a highly precise topographic map. Huber says BMW has the algorithms to develop such a map in-house, and in the future, it will be combined with data from other vehicles to constantly improve the map. “This is why we need this connection to a back-end, to improve the quality of the environmental model, and amend the situational interpretation of what is going on. If a car further up the traffic queue brakes hard, then we don’t have to wait until the traffic pulse moves down the line of vehicles to our car. We know it electronically – a hard brake and everything can stop at the same time.”
Not before 2020
So the big question is: when will this become a reality? “Not before 2020,” says Huber, echoing timescales suggested by other companies that have committed to launching self-driving cars, including Nissan. Aside from the technological development that is still needed, the legal framework crucially needs to be established.
2020 seems close, but in consumer electronic terms, that’s a long way off. “Yes, but in automotive terms, autonomous driving means completely changing the architecture of a car,” says Huber. “We need more redundancy. A data network in a car is designed for the requirements it has to fulfil, and there is no requirement for redundancy. If anything fails, then it fails. We just have to ensure that the car is still safe. In an autonomous mode you have the same requirement, but there is no driver to intervene if the steering or brakes fail. If such things happen in an autonomous car, it must be able to self-diagnose and it must be able to survive ten or 15 seconds until it’s safe. So we have to find approaches for redundancy. And we are working on those concepts. The data networks of cars will change in the future, that is certain, but I’m confident that we can handle it. It’s a question of cost too. But these are problems we can solve.”
Google has opted for a self-driving car that operates at low speeds in urban environments. Huber suggests that BMW sees inner city applications as being further down the line. For now, BMW’s focus is on highly automated motorway and expressway driving.
“We know the sensors we need for the motorway scenario. Highly automated inner city driving is still far away. At BMW we are talking about the highly automated motorway scenario, perhaps travelling like that for a few hundred kilometres. That’s how we see the first step in handing control over to the car.” Freude am Fahren? It might be time to consider reworking that marketing slogan.
