Mary Gustanski, who leads engineering at Delphi, sees a need for a new vehicle IT architecture.


Delphi’s Gustanski sees more upgradable infotainment systems coming

Delphi wants to be at the center of the auto industry’s transformation and the global automotive supplier has been moving away from traditional powertrain technologies to connectivity, automation and electrical vehicle architectures. The core strategy revolves around the company’s aim to build “safe, green and connected” products and systems. “We’re taking our whole hardware portfolio and are adding much more software capability,” said Mary Gustanski, Delphi’s vice president engineering & program management. Gustanski recently spoke to automotiveIT at Delphi’s offices in Troy, Michigan.

Arjen Bongard: Please give us your assessment of where we stand with the big trends of connected cars, automated driving and electrification.

Mary Gustanski: To enable all the new features that everybody wants, you not only have to have more software, but also need electrification. For years, electrification was mainly seen as the way to meet emission standards, but today it’s also aimed at getting more power to the vehicle. With downsized engines the norm, there is little extra power available to add the latest infotainment or active-safety features.

Which is why you need 48 volt in the car.

From a vehicle architecture standpoint, it costs about 150 dollars to enable dual voltage. A lot of the standard electronic components run perfectly well on 12 volt, but you use 48 volt to get a small electric motor into the car to provide the power that lets you implement some of the new technologies. In the US, for example, it will enable stop-start systems that aren’t very popular because of the lag experienced when you start driving again. Forty-eight volt gives you instantaneous power.

You need more electrical power and you also need more computing power and connectivity.

The auto industry has been lagging in this area, but automotive requirements are now starting to accelerate sharply. That’s partly because of the automated driving trend and advanced driver assistance. These systems are complex and require connectivity inside and outside the car. This year, we are launching systems with data speeds of 1.5 gigabits a second. We’re already designing for data speeds of 12-to-15 gigabits/second by 2025. Diagnostics is a big driver of this.

With complexity growing, are cars becoming multi-computer moveable platforms?

Our strategy revolves around centralized computing, which reduces the number of individual computers in a car. If you do nothing, cars will have 100 controllers. With centralized computing, you will maybe need 65 and four or five supercomputers to start out with for infotainment, active safety, powertrain, chassis and convenience features. They back each other up.

Do you need a completely new electrical and data architecture for a highly automated vehicle?

The industry is taking an incremental approach. It would be easier if the architecture were brand-new, but no-one can afford that. So as you go forward, you add new capabilities. Adding sensors and small domain controllers is a way to add complexity into existing architecture without having to disrupt the architecture you have.

So that’s the advantage of Tesla? It has no legacy architecture.

The incremental approach gives you an architecture that’s not as flexible, but I wouldn’t say it’s not as good as Tesla’s. Tesla’s approach does have the advantage that it allows automated vehicle assembly. It’s easier to use robots if you don’t have to string wires through the car’s seating. But you don’t save money, as the cost of components and wiring is actually higher. Ribbon cables are more expensive than traditional round wire.

And because it has no legacy, Tesla became the first brand to offer over-the-air updates.

Tesla designed a software-based vehicle and started OTA. After some initial hesitation, automakers and government authorities - are now getting used to the idea. Tesla has made everybody move faster in the direction of OTA. It’s one of the reasons we acquired (OTA specialist) Movimento. Delphi is pushing to make OTA happen. You need to keep up with the Joneses, so you will have to offer some functionality upgrades over the life of the vehicle.

But automakers remain reluctant to open up the car for OTA updates, yet everyone agrees this capacity is essential to have tomorrow’s vehicles meet customer expectations for the life of the car.

Delphi wants to be a bridge to this technology. It’s not really feasible for a traditional automaker to introduce OTA overnight. But if you adopt centralized computing for active safety, you can put in OTA capability. You can also have software that learns, and as you become comfortable with, say, a level 3 safety feature, you can just turn it on. Automakers can do that without starting over from scratch. And the added complexity can be handled by the traditional architecture of the car. This will evolve, but it will take time.

How much time?

In the next three to five years there will be a greater degree of OTA. Carmakers are already building infotainment systems that are reconfigurable. You will see more features that are going to be upgradeable. Particularly in the infotainment space, where the gaming industry has led the way, automakers realize they’re going to have to offer this. They’ll be more conservative in areas such as powertrains.

Let’s talk about self-driving cars. Are automotive systems ready for this trend?

There are two approaches to automated driving. Nvidia says we’ll let the vehicle learn as it drives and then it will be able to drive. That’s called machine learning. The issue we see with this approach is that you cannot validate the technology. You cannot open the controller and see the lines of code. You can test how it performed yesterday, but you don’t know how it will perform tomorrow. It might well be flawless, but you cannot prove it. Delphi takes a different approach by building a rules-based foundation. Then you use artificial intelligence to improve and enhance the system.

Like all automotive companies, you too need lots of software engineers. How have you been doing on the hiring front?

We’ve done pretty well and currently have 6,300 software engineers out of 20,000. They work mostly in electronics and safety-related areas and there are about 1,000 in powertrain. We have a three-step approach in our hiring. The first is that we recruit computer-science graduates fresh out of school. Second, we retrain engineers who are already working for us, especially those who understand system integration. And third, we use a lot of third-party software providers and we acquire companies or align ourselves with them. Examples are our acquisition of Movimento, which has very strong OTA capabilities. We don’t want to reinvent the wheel and with the help of other companies we can achieve our goals very quickly.

And this has helped change the corporate culture at Delphi?

Absolutely. When you bring in new talent, they see things with new eyes and have a different way of looking at things and doing things. They have a different way of innovating than someone who’s been here for 30 years. We’re also getting people to see and talk about convergence and connecting the dots, which lets you create some pretty exciting systems. Even as recently as three years ago, there was much more divisional thinking.

What kind of role does the CIO and his IT department play in this transformation?

The CIO is the enabler. He tells us how different development processes should work, how we will use the cloud, process data. He also makes sure we have sufficient bandwidth for connected systems and we adopt cyber security measures companywide.