New regulations, questions about safety and cybersecurity, massive upheaval in the auto industry – the participants at the TQ round table listed many challenges of autonomous vehicles that must be overcome before the widespread use is possible. Nonetheless, they are certain that the time will come.
Why not?”, says Prof. Amos Albert, CEO of Deepfield Robotics, when asked whether he would have permitted a self-driving car to bring him to this meeting of experts. Prof. Eric Sax, Director of the FZI Research Centre for Information Technology in Karlsruhe, is a little more cautious: “I wouldn’t like to drive in a car without having a view of the road. The systems are not yet sufficiently sophisticated for that.” On that point, all the round-table participants are essentially in agreement. “At the moment, there is no vehicle on the market that has more than level 3,” according to Jens Kahrweg, General Manager EMEA at Savari. In his opinion, this is also because the legislation has not yet been adapted and liability issues must still be clarified. Prof. Sax also believes that the greatest difficulty is not in the technology, which is already available by and large, but more in making the autonomous systems safe: “With traditional methods, you look to see if the system is executing the action that was previously defined. But the plethora of situations that arise in daily road traffic can no longer be mapped with this method.” Albert agrees that it is simply not enough to calculate the failure probability of hardware and software in order to determine the safety of a self-driving vehicle. “When the ambient conditions are changing too quickly, the safety of the system cannot be calculated. We need other methods.” According to the CEO of Deepfield Robotics, one option is the proof test, for example allowing vehicles to drive as many kilometres as possible with the technology activated, and monitoring them to verify the reliability of that technology. Another is to define safe states and then ensure that the vehicle can return to such a state in the event of a problem. “This would obviate the need to calculate so many potential situations”, according to Albert.
“If autonomous driving is sold as a service, this will also reduce the entry barriers for users.”
Prof. Amos Albert, CEO, Bosch Deepfield Robotics
AI and V2X will be available in a few short years
However, the variety of possible real-world situations to which a self-driving car must respond is not just a massive challenge for safety and reliability. Prof. Sax explains: “As it is impossible to predict every event, a self-driving car needs artificial intelligence.” In this way, it can gain experience independently and learn how to respond correctly. The technology is still in its infancy, however. “But in just a few short years, this will no longer be an obstacle to progress,” asserts Prof. Amos Albert. “The computing power will then be available thanks to Cloud-based systems and deep learning applications.”
In the future, the knowledge gained would then be shared between the self-driving vehicles themselves. According to Jens Kahrweg of Savari, however, the vehicle networking required to do this still needs to be created. Savari was founded in 2008 and is in the process of developing the necessary V2X solutions. “Networking like this would act like another sensor and offer additional redundancy, in order to ensure the functionality of a vehicle.” While the future mobile communications standard 5G does not exist yet, this should not be a problem in five years’ time, according to Kahrweg. “How widely this is then rolled out and whether it can actually do everything it promises is something we are working on with the industry. But it should then be possible to network vehicles.” On the other hand, the modified Wi-Fi standard IEEE 802.11p has already been introduced. “The exciting question is whether both technologies will exist in parallel in the future – that’s what we’re hoping for at Savari. Because, among other benefits, having two systems increases the functional safety of self-driving vehicles.”
“Autonomous driving will -revolutionise entire business models.”
Thomas Staudinger, Vice President Marketing, EBV Elektronik
One of most important challenges of autonomous vehicles is cybersecurity
While networking and comprehensive communication do increase safety, they also simultaneously represent a risk, as pointed out by Thomas Staudinger, Vice President Marketing at EBV Elektronik: “Every point in such a network is a possible target for hackers. In a successful attack, not only could an individual vehicle be manipulated: an entire fleet could.” FZI Director Sax considers the over-the-air updates of future vehicles to be a particular risk: “Software is loaded onto the vehicle – and it will be a huge challenge to identify whether or not it is valid.” With hardware security building blocks, authentication and cryptography, a first firewall can be built. “We are also working on anomaly detection methods,” explains Prof. Sax. This is essentially based on the concept that the usual signals exchanged within the vehicle by the various components are known and can be checked for their plausibility. “The moment an unknown pattern emerges, the anomaly detection springs into action.” In the extreme case, the vehicle can then be brought into a safe state. “Ultimately the cybersecurity of the vehicle is a question of system architecture,” says Staudinger. “To create a solution here, however, the most varied of players must come together – by that I don’t just mean the car manufacturers or companies like Google and Facebook, but also all the very different stakeholders in such a network.”
“The networking of vehicles with one another has already begun; the missing connection, the networking with the infrastructure, will be made possible on a comprehensive scale by future mobile technologies.”
Jens Kahrweg, General Manager Savari EMEA
Upheaval for car manufacturers
Jens Kahrweg sees advantages here for car manufacturers who are just coming into the market: “For a manufacturer who has been producing millions of cars for decades, it is much more difficult and costly to completely overhaul their entire vehicle architecture to suit this new reality. Someone starting from scratch, however, can achieve this innovation leap much more easily.” The established car manufacturers will have to deal with significant upheaval as a result of self-driving cars – because the points of focus will change and software will play a very different role. Prof. Sax adds: “This is why the companies from Silicon Valley are walking all over us. But there is a limit to how much they can do that, because the traditional car manufacturers have other know-how – for example, in the fields of mass and variant production. However, the German car industry in particular will have to make some effort to tackle this learning curve. In future, it will no longer be the gap size that determines whether a car is sold.” Prof. Amos Albert does not believe that the demise of the traditional car manufacturers is imminent either: “Naturally, other companies are very strong in selected software technologies at the moment. But everyone is working at full throttle on new algorithms and at some point, this will even out and the old strengths will play a role again.”
On one thing, all the round-table participants are agreed: the success of car-makers will depend more on whether they can adapt their business models. “In the future, we will not be selling a car, we will be selling mobility,” says Thomas Staudinger of this development away from a product and towards a service. “Perhaps the manufacturers of the future will no longer sell their cars to private customers, but to fleet operators instead,” opines Jens Kahrweg, pointing to transport service operators like Uber or car-sharing operators like Car2go. “When a car-maker is manufacturing only for large customers, the OEM can quickly become a Tier1,” according to Staudinger. Kahrweg continues the line of thought: “And when the OEM develops into a mobility service provider, the suppliers can move up from below and take over the corresponding added value.”
“Electromobility with electrified, decentral actuators and auxiliary equipment will be a door-opener for autonomous driving functions.”
Prof. Eric Sax, Director, FZI – Research Centre for Information Technology
From niche to mainstream
And yet these huge changes will only be relevant if the autonomous car actually establishes itself on the market. The best way of doing that is for it to offer a financial advantage. “Of course, it is wonderful to have a goal of completely eradicating road deaths by 2050,” says Thomas Staudinger. “But ultimately someone must be paid to make autonomous driving a reality. Everything moves more smoothly when the bottom line shows a profit.” This is also why Prof. Sax is firmly convinced that autonomous driving will first be realised in the commercial vehicle sector. “According to estimates, a haulier or municipal service provider can save up to 50 per cent on their costs by introducing self-driving vehicles.” FZI has, for example, run the figures for the Stuttgart public transport service and calculated that it could save well in excess of 100,000 euros on personnel costs per year by automating the trips to the depot. “Autonomous driving will develop from precisely these niches, from applications with manageable scenarios. And that can be done right now,” says Sax.