Continental CTO compares move to software-defined cars to climbing Everest

BERLIN -- Continental Chief Technology Officer Gilles Mabire has a unique way of explaining the difficulty of the industry's move to software-defined vehicles. "We are not climbing the Alps anymore. We are in the Himalayas," he said, likening the challenge to summiting the famous mountain range's high peaks, with Mount Everest being the highest. That means the German supplier has to develop its products in a completely different way, he said. Under Mabire's guidance, Continental is also helping automakers maximize the benefits of over-the-air updates and tap into the power of artificial intelligence to slash the time it takes to solve complex tasks. He discussed this and more with Automotive News Europe Managing Editor Douglas A. Bolduc and Correspondent Nathan Eddy.

What are the main challenges in the move toward software-defined vehicles?

We need to understand the magnitude of the transformation. We are not climbing the Alps anymore. We are in the Himalayas. We need different equipment, and we need to train in a different way. In the hardware-defined vehicle approach, we developed the hardware based on the specifications written more or less by our customers. We were developing and producing products in a very top-down approach. In a software-defined approach, we start by asking ourselves: "How do we distribute the software functions in the car to enable the mobility of the future?" This future includes frequent updates and the addition of new functions that come in after the initial sale of the car. These new functions will impact how and what people will use, purchase and consume in the future. Therefore, we will have to develop our products in a completely different way than before. You need to have a deep understanding of the vehicle architecture and, of course, how to develop your software. We have to think about this even if the hardware does not yet exist, which we can do through virtualization and twinning methods.

Will you work with partners?

We have a partnership with Amazon Web Services to develop what we call our Continental H framework, which is an environment to develop software openly without the hardware and progressively make the hardware in the loop in the second stage, saving costs and breaking down the complexity. For us, the software-defined vehicle means anticipating a change, proving to our customers that we understand it, and then we can support them by bringing the right capabilities to the table to make this change possible.

How will over-the-air software updates evolve?

For a complex development, an OTA update is like life insurance because you can develop your car and still make sure that it can be upgraded with new functions afterward. The first element is to standardize and democratize the connectivity, which we are doing by rolling out 5G in the market. This will ensure that the pipe to feed software into the vehicle will continue to be improved. The second aspect, which aligns with the move toward the software-defined vehicle, is that you need to define the architecture upfront to ensure that you upgrade your car and its functions in a holistic way. The performance of the hardware is improving quickly thanks to SoC (system on chip) components and more memory. This will allow you to offer more frequent OTA updates that will include partial upgrades.

What is the main benefit?

You don't have to fully upgrade the vehicle's software system. You just upgrade certain features. This is all possible from a technical point of view, you just need to create the right architecture upfront.

What will be the focus of Continental's new AI lab in Berlin?

One focus is to develop competencies on the technology around AI to help our projects going forward in assisted driving, autonomous mobility, computer vision, and supporting industrial projects we have within the company. The second focus is to address social topics and problems that go beyond the usual business of Continental.

Such as?

We believe our DNA can be an enabler to help solve some problems around sustainability, CO2 reduction, mobility opportunities and the transportation of goods. This includes last-mile delivery with robots and other things that are not in our core business today.

How will Continental integrate AI-powered technology into the business?

It's about making the right decisions. One focus is on AI-empowered employees, which involves economies of scale and boosting efficiency and productivity from the design perspective down to manufacturing. We are looking for the biggest pain points.

Could you give an example?

One would be the growing number of requirements for vehicles. It's not just the components. This includes high-performance computers and infotainment system, which are extremely complex. More than 80,000 or 100,000 different requirements must be written down. You need some incredible capabilities to sort them out and introduce them to your development team. It takes months. By training AI to learn how to interpret requirements, breaking them down and organizing them, this process could take weeks if not just days in the future. This would speed up the development process, significantly reducing cost and boosting efficiency, so there is more time to focus on the key challenges that need to be addressed in the project.

Do you have any other examples?

When it comes to manufacturing or predictive maintenance, AI can anticipate -- across the complete supply chain -- that a problem is going to create a quality issue. It can do this just by monitoring key parameters in the machine to determine what could create a big issue later. By feeding this data into your factory, you can address and improve the situation. Another area where AI can help is with component shortages, which have impacted everyone during the last two years. We are putting in place some predictive models using AI to see how the market is evolving. We can determine how to maximize the output while also optimizing the production schedule to align with the customer's need. In short, this makes the complete supply chain machinery more controllable and more predictable. That is now powered by an AI algorithm developed by our team.

How are you integrating AI into product development?

We use AI to improve our ADAS technologies, from camera fusion to computer model labeling. To recognize an object on the street with a probability of a 90 percent, you need to train your algorithm. Therefore, you need to collect a lot of data and manually label what is a person, what is a bicycle, what is an animal and what type of animal it is, and so on. With automatic labeling powered by AI, we can now simulate real-use cases without driving. You can reproduce many corner cases that you probably will never see or are very unlikely to experience in the real life. But you can train your algorithm millions of times, making development shorter but safer. We can run millions of kilometers in a couple of days with complete simulation. We can elaborate the test plan accordingly, which gives us a depth in terms of testability far higher than what is possible with a human being. All these elements improve the quality of the product we put on the market.