Industry Trends Toward Software Virtualization and Simulation

The automotive industry is shifting from a physical-first testing mindset to a virtual-first engineering approach, driven by the massive rise in software-defined vehicle complexity.

In this episode,, host Suresh Sivavarman sits down with Cedric Armand, Director of Virtualization at Ford Motor Company, to break down the industry's rapid evolution toward software virtualization and simulation. Cedric shares how virtual-first environments enable automotive engineers to run millions of automated regression tests overnight—shifting defect discovery left to prevent costly field failures. Together, they discuss the foundational building blocks of scaling simulation (from virtual ECUs and plant models to cloud-based co-simulation platforms), overcoming the limits of physical HIL benches, managing organizational change for virtual tool adoption, and leveraging AI to accelerate testing efficiency.

Whether you are an automotive software engineer, systems architect, or validation manager, this episode offers a firsthand look at how major OEMs are scaling virtual development to accelerate time-to-market.

In this episode: 
00:00 - Introduction: The Shift from Physical Testing to Virtual-First Engineering
03:00 - Defining Virtualization vs. Simulation in Automotive Systems 
08:15 - The Massive ROI of "Shifting Left" & Running Automated Regression Tests
16:45 - Beyond HIL Benches: Scaling Virtual ECUs and Cloud-Based Co-Simulation
24:30 - Cultural Change & Organizational Buy-In for Virtual Tool Adoption
32:10 - The Future Role of AI in Diagnosing Test Failures and Streamlining Validation

Thanks for listening!

• Email us at: contact.us@etas.com
• Learn more about ETAS on our website 
• Follow us on LinkedIn: @ETAS

Transcript

00:00:02 Voiceover

Welcome to the Empowering Tomorrow's Automotive Software Podcast, brought to you by ETAS, a single source of cutting-edge software and hardware solutions that make automotive embedded systems safe, smart, secure, and sustainable.

00:00:15 Voiceover

Each episode, we'll be joined by ETAS and industry experts to discuss how electrification, automation, and connectivity are impacting the automotive industry.

00:00:25 Voiceover

Now, sit back and enjoy the discussion.

00:00:32  Suresh Sivavarman

Hello, everyone.

00:00:32  Suresh Sivavarman

Welcome to another episode of Empowering Tomorrow's Automotive Software Podcast with ETAS.

00:00:38  Suresh Sivavarman

I'm your host, Suresh.

00:00:39  Suresh Sivavarman

And today I'm joined by Cedric Armand, Director of Virtualization at Ford Motor Company.

00:00:45  Suresh Sivavarman

Cedric brings over 15 years of experience in the automotive domain, from embedded software to controls, calibration, and validation.

00:00:53  Suresh Sivavarman

He's led breakthrough virtualization initiatives at GM and now at Ford.

00:00:58  Suresh Sivavarman

and he sits at Ford's center of gravity for shifting towards virtual-first environments.

00:01:03  Suresh Sivavarman

Cedric, it's great to have you here, and we're looking forward to the conversation.

00:01:08 Cedric Armand

Thank you.

00:01:08 Cedric Armand

Really excited to be here.

00:01:09 Cedric Armand

Thanks for having me on.

00:01:11  Suresh Sivavarman

No worries.

00:01:12  Suresh Sivavarman

So Cedric, I'm going to start you off with a general question, and then we'll kind of go from there.

00:01:17  Suresh Sivavarman

I think based on your background, you've seen the evolution from going from a physical-first testing mindset to now a virtual-first

00:01:26  Suresh Sivavarman

engineering approach.

00:01:29  Suresh Sivavarman

I know for me, this is a challenge because I come from hands-on calibration, I come from hands-on testing.

00:01:35  Suresh Sivavarman

Now to move into a virtual environment, why do you think when you joined Ford, this became so mission critical?

00:01:42  Suresh Sivavarman

What value is it bringing to you?

00:01:44 Cedric Armand

Yeah, great question.

00:01:46 Cedric Armand

It's become mission critical for a number of reasons, right?

00:01:51 Cedric Armand

Number one, software

00:01:54 Cedric Armand

within automotive vehicles is just becoming so important, right?

00:02:00 Cedric Armand

It encompasses like close to 50% of a vehicle's value, right?

00:02:06 Cedric Armand

It's becoming more complex and customers want software and software updates to have more features and to bring it faster, right?

00:02:16 Cedric Armand

So realistically, you know, the demands of where software needs to be,

00:02:22 Cedric Armand

and the quality expectations continue to rise.

00:02:26 Cedric Armand

So we just need to continuously keep bringing more tools, methodologies, and enablement to deliver our software faster, have it have higher quality, and also be able to meet the challenge of how do we actually develop and validate and test something that's growing in more complexity every day.

00:02:48  Suresh Sivavarman

So how do you actually define

00:02:50  Suresh Sivavarman

virtualization or simulation.

00:02:53  Suresh Sivavarman

What is your driving factor behind it?

00:02:56  Suresh Sivavarman

What is the driving differences between what you would do traditionally versus what you're doing now with the modern environment?

00:03:02 Cedric Armand

Yeah, so let me put a couple of little nerdy definitions, right, and then kind of go into it, right?

00:03:09 Cedric Armand

So virtualization simulation, used somewhat synonymously, but virtualization is simply making a virtual version of some

00:03:18 Cedric Armand

physical artifact or some physical system or some physical hardware, right?

00:03:22 Cedric Armand

And then simulation is actually executing those virtual artifacts in a manner such that, they behave the way the real physical system does, right?

00:03:33 Cedric Armand

So the simulation is actually using these virtual artifacts.

00:03:37 Cedric Armand

So virtualization, simulation, right?

00:03:40 Cedric Armand

Huge fields, right?

00:03:41 Cedric Armand

So

00:03:42 Cedric Armand

If somebody's listening or is first coming across, idea of virtualization simulation, right?

00:03:46 Cedric Armand

They might be thinking of like VR headsets, right?

00:03:49 Cedric Armand

Which is virtualization and simulation as well.

00:03:52 Cedric Armand

You know, what we're talking about specifically is virtualization and simulation in for the purpose of software development and testing.

00:04:00 Cedric Armand

You know, so that means you're kind of bringing three major or simple kind of components together.

00:04:08 Cedric Armand

One is a virtual ECU.

00:04:10 Cedric Armand

which is a virtual version of a particular software stack or the ECU in that software stack.

00:04:19 Cedric Armand

Then the other group of artifacts you'll be bringing are plant models, right?

00:04:24 Cedric Armand

And these are models of, you know, the physical system with which you're trying to control, right?

00:04:30 Cedric Armand

So very simple.

00:04:31 Cedric Armand

It could be like a plant model of an engine, right?

00:04:33 Cedric Armand

So you can have a virtual version of your engine controller, then you can have a plant model of your engine,

00:04:39 Cedric Armand

And then the third component is just some co-simulation platform that allows the control signals from your virtual ECU to talk to your plant models and then feedback signals from your plant models back to those virtual ECUs.

00:04:55 Cedric Armand

So now you have a closed-loop environment, you know, such that it will execute the software, you'll get feedback, and it'll operate similarly, if not, you know, the same as it does in an actual physical vehicle.

00:05:08  Suresh Sivavarman

So what are some of the expectations in respect to tooling to enable some of that stuff?

00:05:13 Cedric Armand

Very good.

00:05:13 Cedric Armand

So there's a whole host of tools to use and to enable this, right?

00:05:18 Cedric Armand

So you need, there are different tools and we could talk about specific ones out there off the shelf in the market, right?

00:05:25 Cedric Armand

But at a simple level, you need a tool that allows that co-simulation to happen, right?

00:05:31 Cedric Armand

That allows the communication of signals back and forth.

00:05:35 Cedric Armand

then you need a tool or tool suites to virtualize your ECUs and virtualize your software stacks, right?

00:05:42 Cedric Armand

And there's different levels and fidelities of virtual ECUs to do that.

00:05:46 Cedric Armand

And then there are tools to make different types of models, right?

00:05:49 Cedric Armand

So it could be plant models, which people are sure are familiar with.

00:05:52 Cedric Armand

You know, there could be something, a Simulink-based model, but a whole host of different tools to make different types of models of different fidelities.

00:06:00 Cedric Armand

But then you also might need other types of models, such as

00:06:03 Cedric Armand

something known as like rest bus models, which is maybe models of other critical signals that happen within the communication network that maybe are not present in your environment, but you need to simulate to some degree, right?

00:06:17 Cedric Armand

So if we want to talk about specific tool chains, we can, or types or examples by all means, but there's a whole host of tools out there for sure.

00:06:28  Suresh Sivavarman

And I think one of the things you mentioned is this is a closed-loop system.

00:06:33  Suresh Sivavarman

So my expectation or my understanding is all these tools have to work together to build this ecosystem, right?

00:06:39  Suresh Sivavarman

So what challenges do you see in this interoperability requirement, right?

00:06:44  Suresh Sivavarman

This is definitely a change from what we do today.

00:06:48 Cedric Armand

Yes, And so the environment, the simulation environment I described, yes, was a closed loop environment, right?

00:06:56 Cedric Armand

You can have a open loop environment, but the only difference is that

00:07:01 Cedric Armand

you would have minimal number of plant models within the loop.

00:07:05 Cedric Armand

So this is usually, those types of environments are used just to, maybe you're checking like diagnostics or communications or other low-levels part of the software where maybe you don't need that feedback loop, right?

00:07:18 Cedric Armand

But even still, your question still applies where the interoperability is very important.

00:07:23 Cedric Armand

And

00:07:25 Cedric Armand

earlier or years ago, interoperability just didn't occur, right?

00:07:30 Cedric Armand

And that was okay, right?

00:07:32 Cedric Armand

Because the solutions were kind of very singular, very pointed, right?

00:07:37 Cedric Armand

Some OEMs, started creating solutions, kind of in-house and from the ground up, companies started creating solutions, right?

00:07:45 Cedric Armand

And they were very, very focused, right?

00:07:46 Cedric Armand

But now, as the ecosystem is growing, right, no one company or tool chain can really truly satisfy

00:07:54 Cedric Armand

all that is needed in a virtualization and simulation, all by itself, right?

00:08:00 Cedric Armand

So, and different companies are going to be better at certain things.

00:08:03 Cedric Armand

So, it is very important to have interoperability between tool chains that make virtual ECUs, that communicate signals, that serve as platforms, that make plant models.

00:08:16 Cedric Armand

So as it becomes more important, right, you just, we need to have certain standards that we kind of work towards, but then also just understanding that, if there's a tool chain here, it needs to work with, these other tool chains out there.

00:08:29 Cedric Armand

And, you need to have open APIs to be able to have that interface.

00:08:32  Suresh Sivavarman

So I'm going to talk about simulation or virtualization.

00:08:39  Suresh Sivavarman

You know, what we used to do on a PC was very simple, right?

00:08:41  Suresh Sivavarman

You reload it on your PC, you run it, and you get some output.

00:08:44  Suresh Sivavarman

When we're talking about this big closed loop infrastructure environment, the compute power seems complex and heavy.

00:08:53  Suresh Sivavarman

So where do you, or where does Ford kind of operate this environment?

00:08:59  Suresh Sivavarman

Is it in a cloud environment or is it in some kind of hybrid blend of the two?

00:09:03  Suresh Sivavarman

Where do you turn the operate it?

00:09:06 Cedric Armand

Yeah, great question.

00:09:07 Cedric Armand

So you're right.

00:09:08 Cedric Armand

When I first started, and even still now, right, I mean, you can do things on your PC,

00:09:13 Cedric Armand

But realistically, at an OEM and even, tier one suppliers, right, virtualization and simulation really needs to be done at scale, right?

00:09:24 Cedric Armand

So let's just think about that for a second, right?

00:09:26 Cedric Armand

At Ford or any OEM, our vehicles average, say, 70 ECUs within it, right?

00:09:34 Cedric Armand

So, think about all these different software stacks, just as a general concept, but then we have different model years, we have different specific vehicles lines, right, and programs that we need to write.

00:09:46 Cedric Armand

So, there's a whole lot of software, a whole lot of different simulations you try to run, and they become intensive, right?

00:09:53 Cedric Armand

So, realistically, running this

00:09:57 Cedric Armand

on PCs is not the way to kind of industrialize or scale that up, right?

00:10:03 Cedric Armand

And so you need to have a cloud-based platform with which users, internal users who are trying to use these environments to develop and test their software, can be able to access these environments and these virtual artifacts for their testing, as well as, you know, whatever testing pipelines, and you know, that they might emphasize, just so that, and this serves 2 purposes.

00:10:25 Cedric Armand

One is that now you're more easily able to share environments between 2 users, right?

00:10:31 Cedric Armand

So let's say you and I, Suresh, we're working on a gateway ECU, right?

00:10:37 Cedric Armand

And we want to develop the software and we're developing it together.

00:10:41 Cedric Armand

We want to be able to share that environment and kind of use the same thing, right?

00:10:44 Cedric Armand

And so it becomes very

00:10:46 Cedric Armand

easy, right?

00:10:47 Cedric Armand

People don't realize it if they haven't gone through it.

00:10:49 Cedric Armand

But if I have a bunch of files on my computer and I say, hey, Suresh, I want you to look at the same thing.

00:10:53 Cedric Armand

I got to zip all these files or post it or whatever and hope you put all the same configurations exactly right.

00:10:59 Cedric Armand

Chances of that working out are slim to none.

00:11:01 Cedric Armand

But if you have it hosted on a cloud somewhere and somebody access it, boom, okay, you pull down, you make like a duplicate copy, I have to do the same thing.

00:11:09 Cedric Armand

Now we're working the same thing.

00:11:11 Cedric Armand

That's one part.

00:11:12 Cedric Armand

The other part is virtualization and simulation.

00:11:16 Cedric Armand

have an immensely high return on investment, right?

00:11:20 Cedric Armand

But they also have a high initial investment to get it to get it to work.

00:11:27 Cedric Armand

So you increase that return the more you scale, right?

00:11:31 Cedric Armand

And that means you want to be doing more and more testing, millions and millions of testing.

00:11:35 Cedric Armand

And frankly, in automotive industry, right, we do not test our software

00:11:40 Cedric Armand

Nearly as much as we need to, right?

00:11:42 Cedric Armand

There are different thoughts, within the industry about what it is, but a rule of thumb that you would hear is you want to spend 50% of your effort.

00:11:53 Cedric Armand

developing the software and 50% of your effort testing it, right?

00:11:56 Cedric Armand

And so you're not realistically, companies are just not going to hire as many, let's say, software testers as they are software developers, right?

00:12:06 Cedric Armand

But you can reach that ratio by being on a cloud-based platform and being able to run, you know, millions and millions of tests overnight, regression, kick those off, right?

00:12:17 Cedric Armand

And just leverage, you know, having, you know, big server farms.

00:12:21  Suresh Sivavarman

So maybe I want to touch on the ROI topic again, right?

00:12:26  Suresh Sivavarman

You mentioned that the return on investment long-term is huge.

00:12:30 Cedric Armand

Yeah.

00:12:31  Suresh Sivavarman

What are some success factors that you believe need to occur to reach that ROI?

00:12:37  Suresh Sivavarman

And what are some of those investments you need to do up front in order to get this type of environment functional?

00:12:44  Suresh Sivavarman

For example, I know navigating plant models and integrating them into

00:12:49  Suresh Sivavarman

software in the loop environments, there's a challenge there, right?

00:12:52  Suresh Sivavarman

But is it mostly human effort or is it infrastructure effort that takes more time to get set up?

00:13:01 Cedric Armand

This is what I'll say.

00:13:01 Cedric Armand

Having gone through this, having benchmarked other OEMs who have done it, right?

00:13:07 Cedric Armand

Generally, an approach that works is creating a cloud-based platform

00:13:14 Cedric Armand

where users can access and have different workflows, right?

00:13:18 Cedric Armand

And then you need to have an infrastructure to virtualize your ECUs, right, to make virtual ECUs to use.

00:13:25 Cedric Armand

And then, yeah, you need to have a way of developing the necessary plant models.

00:13:30 Cedric Armand

And then you also need the effort to integrate them to make these environments, right?

00:13:34 Cedric Armand

So

00:13:35 Cedric Armand

Technology-wise, to, let's say, get the platform and get the infrastructure and things like that, right?

00:13:40 Cedric Armand

You can get that up and running.

00:13:42 Cedric Armand

that's been done.

00:13:43 Cedric Armand

There's ways to optimize it.

00:13:45 Cedric Armand

There's companies, there's partners out there who've done this a couple of times.

00:13:51 Cedric Armand

Within organizations, yeah, you need the human power and the organization structure where you have

00:13:58 Cedric Armand

Engineers who know how to develop virtual ECUs, engineers who know how to make plant models, engineers who know how to integrate them into environments, right, to set those up for users.

00:14:07 Cedric Armand

And then the users need to get brought up to speed on how to use it.

00:14:11 Cedric Armand

And then there's the culture aspect, right, where with that I'd say, you know, this is a change, right?

00:14:18 Cedric Armand

So many, many years ago when hardware in the loop or Hill Benches first came out, right, nobody wanted to use those because everybody wanted to do things in vehicle.

00:14:26 Cedric Armand

You know, we fast forward.

00:14:29 Cedric Armand

Right.

00:14:29 Cedric Armand

Everybody, everybody wants to just use Hill benches, which they're good, right?

00:14:34 Cedric Armand

And maybe it was like 10, 15 years ago when software in the loop started coming out, right?

00:14:39 Cedric Armand

Nobody wanted to do that, started to gain a little steam.

00:14:42 Cedric Armand

And now there's like this next evolution of software in the loop where it's really becoming a bigger, grander scale.

00:14:48 Cedric Armand

You're working with different fidelity levels and abstraction levels of virtual ECUs.

00:14:54 Cedric Armand

And so, you know, that's a bit of a cultural change as well.

00:14:57 Cedric Armand

And also,

00:14:58 Cedric Armand

Part of that is not only using different tools, but hey, now you got to start developing your software earlier, right?

00:15:04 Cedric Armand

Now you got to test more of those things and have integrated workflows.

00:15:08 Cedric Armand

So, you know, that part of it has to come along as well.

00:15:12 Cedric Armand

So all of it needs to come together for it to be a success.

00:15:15  Suresh Sivavarman

So when we're talking about the business side of it, right, we're asking management for huge investments.

00:15:21  Suresh Sivavarman

We're going to promise you this ROI.

00:15:24  Suresh Sivavarman

But the first question they're going to ask you is, all right, how do I track my investment?

00:15:27  Suresh Sivavarman

What are the KPIs that I would see as success factors, right?

00:15:30  Suresh Sivavarman

So in your role, what do you see as things that you would track in order to say, this virtual, virtualizing this platform into this ECU and these set of vehicles or whatever it may be, this is how I know that I'm doing better than I was previously.

00:15:47  Suresh Sivavarman

What are some of those success factors?

00:15:49 Cedric Armand

Yeah, absolutely.

00:15:50 Cedric Armand

So, you know, the

00:15:53 Cedric Armand

main benefits of virtualization and simulation at scale are you, and this is, there are industry proof points, right?

00:16:04 Cedric Armand

But one is you're going to shorten your software development time by 10 to 12 months.

00:16:13 Cedric Armand

You can also shift it to the left by 10 to 12 months.

00:16:17 Cedric Armand

And then the third, I think I said two, but it's really, it's really three, two of the 1st 2 are related.

00:16:22 Cedric Armand

And then the,

00:16:23 Cedric Armand

The last is you're going to shift your defect discovery or your software bug discovery to the left, right?

00:16:30 Cedric Armand

And I want to get into that a little bit.

00:16:32 Cedric Armand

Anybody who works on software, developed software knows that you're going to have defects, you're going to have bugs.

00:16:37 Cedric Armand

It's just the nature of developing it, right?

00:16:39 Cedric Armand

And when you're doing that, it's just about are you going to discover them in time and early enough when it's cheaper, right?

00:16:46 Cedric Armand

So if I write a piece of software right now, but you know, I find that bug

00:16:53 Cedric Armand

tomorrow, fairly cheap fix, right?

00:16:57 Cedric Armand

But if I write a piece of software today and then two years, I don't find that bug until that vehicle is launched and it causes a warranty problem, right?

00:17:06 Cedric Armand

Very expensive to fix, right?

00:17:07 Cedric Armand

So the idea is you, know, it's cheaper the earlier you find these defects.

00:17:11 Cedric Armand

And it's, again, an industry proof point where virtualization at scale kind of shifts

00:17:18 Cedric Armand

the bugs that you find, let's say, at a system integration phase and a production beyond phase are like 65% of those bugs are found earlier, right?

00:17:29 Cedric Armand

So if we think about that, right, and then the KPIs we want to track, right?

00:17:34 Cedric Armand

So one is easy.

00:17:36 Cedric Armand

You want to track your bug and your defect discoveries, right?

00:17:40 Cedric Armand

And when they are happening and how many are being found

00:17:44 Cedric Armand

through specifically virtualization and simulation-based testing, right?

00:17:48 Cedric Armand

That's an easy KPI.

00:17:50 Cedric Armand

The other that you want to track is, you know, how quickly or what your time is from, let's say, software feature, you know, starter development to validation, right?

00:18:04 Cedric Armand

So, not that a company would do this, but let's just say, you know, if you had, you know, two different software organizations within your company, right?

00:18:14 Cedric Armand

You track one that's not using virtualization and say, you track how long it's taking for it to develop a software feature and validate it on average.

00:18:21 Cedric Armand

And then you have your parallel company, your parallel organization that is using virtualization, and you track how fast that is, right?

00:18:29 Cedric Armand

You're going to see that difference in time.

00:18:30 Cedric Armand

And so, tracking that as a software organization and companies is critical, and that's where you'll see that benefit.

00:18:38  Suresh Sivavarman

So I think I understood, we could probably move things to the left.

00:18:43  Suresh Sivavarman

Maybe regression testing can be done in a software environment.

00:18:48  Suresh Sivavarman

Diagnostics, fault injection, these are probably things you can do.

00:18:53  Suresh Sivavarman

Having different variants, variations, you can do explosion testing there, right?

00:18:57  Suresh Sivavarman

To fix any matching.

00:18:58 Cedric Armand

Yeah.

00:18:59  Suresh Sivavarman

So there are things I believe that you can do in a virtual environment, and there's things that you still cannot, right?

00:19:05  Suresh Sivavarman

I believe there's things that are still time dependent, safety critical, things like that, still require physical validation.

00:19:11  Suresh Sivavarman

So what percentage of development have you realistically been able to move virtual today?

00:19:19  Suresh Sivavarman

What are you expecting to move in the next years?

00:19:22  Suresh Sivavarman

And how much will still require physical validation?

00:19:26  Suresh Sivavarman

So how much are we still requiring HIL and vehicle validation for?

00:19:31  Suresh Sivavarman

How much can we actually move up front into a software development environment?

00:19:36 Cedric Armand

This is a really, really good question.

00:19:38 Cedric Armand

And I

00:19:40 Cedric Armand

have seen it because over the course of my career, I did software development and I was at a point where software in the loop really kind of came in and I saw, what things we can do.

00:19:52 Cedric Armand

So, when I was doing software development, I would kind of do some basic steps where I would first, develop my algorithm and then say, test it just like a little Simulink model.

00:20:03 Cedric Armand

Then I would compile it into, a virtual ECU and into a software in the loop environment, test it there.

00:20:10 Cedric Armand

Then I would take it to a physical hill bench.

00:20:13 Cedric Armand

Then finally, I would take it to a physical vehicle, right?

00:20:18 Cedric Armand

So that was like my workflow.

00:20:20 Cedric Armand

And that's, I think that's like a good way to do it, right?

00:20:22 Cedric Armand

Because you're going to find different potential problems in gating at the different types of testing or different

00:20:30 Cedric Armand

ways of testing your software.

00:20:33 Cedric Armand

But that was like my experience and what I've seen.

00:20:34 Cedric Armand

And so I've continuously kind of benchmark and look at like research and see what other OEMs and tier ones are finding and what they're kind of shifting, right?

00:20:45 Cedric Armand

So the most optimistic of people say, you know, you can get, you can do like 90 to 95% of your testing, I'll say between

00:20:59 Cedric Armand

software in the loop and physical hill benches, right?

00:21:02 Cedric Armand

But a lot of that being with software in the loop.

00:21:05 Cedric Armand

But I think it kind of breaks down where, development and unit testing, by unit I mean like you're testing like the actual software just running on its software stack, right?

00:21:15 Cedric Armand

So kind of code verification, algorithm development stuff, right?

00:21:19 Cedric Armand

Like 99% of that can just be done with software in the loop, like virtualization, simulation.

00:21:24 Cedric Armand

Integration

00:21:26 Cedric Armand

and function testing, right?

00:21:27 Cedric Armand

So how it integrates into maybe a bigger system and how some features are actually functioning.

00:21:33 Cedric Armand

you're probably at like 70% of that can be done, with software in the loop.

00:21:38 Cedric Armand

Now, to your, when you mentioned, safety critical, so validation and safety testing, that's probably 50-50, right?

00:21:45 Cedric Armand

Where, you know, you need physical assets to do it, 50% you could do, you know, with virtualization and simulation.

00:21:51 Cedric Armand

But no matter what, you always want to do, you know, final vehicle verification testing.

00:21:58 Cedric Armand

right?

00:21:58 Cedric Armand

Or at the very least, final physical hill bench testing, right?

00:22:03 Cedric Armand

You always want to do a final physical test, right?

00:22:06 Cedric Armand

And that will always remain because you always want that last step before that software goes to the consumer.

00:22:13  Suresh Sivavarman

So we talked a little bit about your experiences, right?

00:22:16  Suresh Sivavarman

What you've been able to experience in your past and what you're doing today.

00:22:20 Cedric Armand

Yeah.

00:22:21  Suresh Sivavarman

So

00:22:22  Suresh Sivavarman

first off, what kind of surprised you, maybe from a virtualization perspective, right?

00:22:27  Suresh Sivavarman

What did you think maybe was more difficult to do that you were able to do?

00:22:31  Suresh Sivavarman

On the other side, what did you think that you could do that maybe became more difficult to do, right?

00:22:38  Suresh Sivavarman

Because I think we can all speak to saying, oh, virtualization looks like a great vision, and it probably is, but there are still limitations today, right, that prevent us from going completely, 100% virtual.

00:22:53 Cedric Armand

The probably the most surprising thing, so this is the first is not technical, it's actually a return.

00:23:00 Cedric Armand

So when I actually looked and started to research what the actual true return on investment is, I was surprised by how much it is.

00:23:07 Cedric Armand

And it makes sense once you do the math, but when you kind of think about like the value of being able to deliver software, say like 12 months earlier, you know, that first time the market value is immense.

00:23:19 Cedric Armand

So just think about

00:23:21 Cedric Armand

let's say you have three different OEMs, right?

00:23:24 Cedric Armand

And each of them today says, man, I figured out how to make a flying car, right?

00:23:31 Cedric Armand

whichever company is able to get that to market first, you know, is going to win, right?

00:23:37 Cedric Armand

It's just straight up, right?

00:23:39 Cedric Armand

So, you know, that value is huge.

00:23:41 Cedric Armand

But on the technical side, what was surprising to me was I always thought of like virtualization simulation as

00:23:49 Cedric Armand

Hey, those results need to match the hardware, but what I found kind of maybe fairly quickly is, no, they don't, because hardware is not repeatable, right?

00:24:01 Cedric Armand

So, when you have a physical ECU...

00:24:04 Cedric Armand

that is running, right?

00:24:06 Cedric Armand

It's going to have, capacitance and resistance and be affected by temperature, and whatever setup, that is on a physical hill bench, right?

00:24:15 Cedric Armand

If I work on a physical hill bench today, okay, great.

00:24:18 Cedric Armand

But then, you know, Suresh, when you come and work on it the next day, you're working with whatever I left over and whatever mess I made on that hill bench, right?

00:24:25 Cedric Armand

And so it's not repeatable, right?

00:24:27 Cedric Armand

So you're gonna find issues, right?

00:24:29 Cedric Armand

But

00:24:30 Cedric Armand

There are some issues that you will only reliably find in virtualization simulation, because that virtualization simulation is just software running software, and software is repeatable, right?

00:24:41 Cedric Armand

It runs the same thing again and again and again, you know, once you take the physical hardware aspect out of it.

00:24:47 Cedric Armand

So that was one of the surprising things I found, and I was like, oh wow, you know, it does have, you know, benefits where it's like you're just going to find some things that you're not going to find in physical hardware.

00:24:58 Cedric Armand

And this becomes important

00:25:00 Cedric Armand

because so many times, and even till today, right, there will be issues that come, back in the field or even in late stage development testing that they're not repeatable.

00:25:11 Cedric Armand

Like something happens, something weird, a customer says something, and then it's like, we can't repeat it, or a problem not found when it's at the dealer, right?

00:25:19 Cedric Armand

And there have been enough instances when I started using virtualization simulation where we found issues,

00:25:25 Cedric Armand

that it was like, we didn't find it.

00:25:26 Cedric Armand

We didn't see it on the hardware, but when we dig in, it's like, it is truly an issue.

00:25:30 Cedric Armand

And it's like, man, that's one of those issues that was going to bite us in a year and we weren't going to know.

00:25:35 Cedric Armand

It was going to be a pain to figure out.

00:25:37 Cedric Armand

So that's one part of your question.

00:25:39  Suresh Sivavarman

So where's the limitation, right?

00:25:44  Suresh Sivavarman

Where do we believe that we, no matter what, still need to go back to hardware, right?

00:25:51  Suresh Sivavarman

Where is the interaction with hardware that's still needed?

00:25:53 Cedric Armand

Yeah.

00:25:54 Cedric Armand

I mean, there's, at the end of the day, it's just like a general limitation, where I think people kind of understand inherently where, virtualization simulation, you're still just replicating the hardware and you're never going to get that exactly right.

00:26:11 Cedric Armand

And not only that, when you have virtual models and simulated environments, these are, they're always like works in progress, right?

00:26:20 Cedric Armand

And so really,

00:26:22 Cedric Armand

the way to use those effectively is, first, let's say you have a virtual ECU, right, which is obviously mimicking or simulating your software stack.

00:26:32 Cedric Armand

And so let's say you get your first iteration of your virtual ECU.

00:26:36 Cedric Armand

So you take that and then you start developing your software, right?

00:26:40 Cedric Armand

Now, that software that you're running on your virtual ECU or as your virtual ECU, it is also like validating your virtual ECU model, right?

00:26:50 Cedric Armand

And then, so it really ends up becoming this iterative cycle, or kind of one hand washing the other, where your virtual ECU is being used to develop your software, your software is being used to develop and validate your virtual ECU, and you kind of go around and around this iterative loop each time, you know, iron sharpening iron, you know, each model and each software iteration getting better.

00:27:12 Cedric Armand

So just understanding that it's never going to be exact, and it's like always like kind of a work in progress.

00:27:17 Cedric Armand

We understand that

00:27:19 Cedric Armand

you always want to do at least one final physical verification to see that you covered all your bases.

00:27:27  Suresh Sivavarman

That's really good insights.

00:27:29  Suresh Sivavarman

So, you know, I'm at ETAS and ETAS, we're still growing into this virtualization journey, right?

00:27:38  Suresh Sivavarman

And I think a lot of people that are going to be listening to this podcast are going to be interested in understanding what virtualization means.

00:27:44 Cedric Armand

Yeah.

00:27:46  Suresh Sivavarman

There's also a portion of this, which is

00:27:49  Suresh Sivavarman

cultural and organizational.

00:27:52  Suresh Sivavarman

There's naturally people who are going to have a lack of trust in the results that come out of a virtualization environment.

00:27:59  Suresh Sivavarman

Based on your experience and based on how you've managed to grow your organizations into this world, what are some key messages that you would give to those that are, I would say, not believers, but not full on, you know, they're not fully

00:28:18  Suresh Sivavarman

absorbing it yet.

00:28:19  Suresh Sivavarman

They're not into it, right?

00:28:21  Suresh Sivavarman

What motivational words would you say that this is why we need to do virtualization?

00:28:25  Suresh Sivavarman

And here's kind of the evidence to say this is trustworthy.

00:28:30 Cedric Armand

Yeah.

00:28:31 Cedric Armand

So I'll kind of, I'll say like generally, I forgot who exactly said it.

00:28:38 Cedric Armand

Maybe it was

00:28:39 Cedric Armand

I think it was maybe the CEO of Rivian, right?

00:28:43 Cedric Armand

And he was talking about software-defined vehicles, right?

00:28:47 Cedric Armand

The latest and greatest in vehicle software and EE architecture.

00:28:54 Cedric Armand

And he said something to the effect of, in, I don't know, 5, 10 years, he does not imagine how a automotive company is going to be able to successfully sell vehicles

00:29:08 Cedric Armand

and be in the marketplace without having an SDV architecture, right?

00:29:11 Cedric Armand

Just because it's going to be very difficult to meet the demands and keep up with all the OEMs who do, right?

00:29:20 Cedric Armand

And I say the same thing about virtualization simulation, right?

00:29:25 Cedric Armand

I don't see how, right, automotive companies, right, or software companies in general, are going to be able to keep up

00:29:37 Cedric Armand

if you do not employ virtualization, right?

00:29:39 Cedric Armand

If Ford, right, is, and obviously, this is a competitive market, but I'm just saying, if any OEM company, right, if Ford is, employing virtualization at scale, right, and developing their software, 12 months faster than OEMB, right, how are you going to keep up, right?

00:29:59 Cedric Armand

If, you know, if one company is delivering their software features, you know, earlier, if they're finding, you know, their bugs and they're saving money,

00:30:06 Cedric Armand

in quality and able to take that money they're saving and investing it in other places.

00:30:10 Cedric Armand

How are you going to keep up, right?

00:30:12 Cedric Armand

And so the limitations are absolutely true, right?

00:30:17 Cedric Armand

But you, it's just, as long as you understand them and mitigate them, right, and then use virtualization simulation in the way it's intended, right, and do the, you know, the right kind of tool or simulation for the right type of testing, right, then you find the benefits.

00:30:32 Cedric Armand

And

00:30:33 Cedric Armand

And so anybody who is skeptical, I mean, you can ask the deep questions like, what is the return on investment?

00:30:38 Cedric Armand

What is, all those things, right?

00:30:40 Cedric Armand

And you can go into those deep discussions, but I just would just ask the person in general, all right, if you're not going to do this, how are you going to deliver your software, you know, a year faster, right?

00:30:49 Cedric Armand

How are you going to get your bugs and defects down, you know, to that rate?

00:30:54 Cedric Armand

And so.

00:30:55  Suresh Sivavarman

It's really, it needs to be both a bottom-up and top-down approach, right?

00:30:59  Suresh Sivavarman

There needs to be some factors pushing from above, but at the same time, there needs to be buy-in from the bottom, the actual users to say, this is what I need to do to be successful.

00:31:10 Cedric Armand

Yeah, and from the top down, right, I think leaders need to be saying, hey, we're going to improve our software quality by X amount, right?

00:31:22 Cedric Armand

And we're going to deliver our software faster by X amount, right?

00:31:26 Cedric Armand

Put that pressure on your organization.

00:31:29 Cedric Armand

And then, but then also enable them, right?

00:31:32 Cedric Armand

And then the, software developers and, kind of middle managers, right?

00:31:37 Cedric Armand

they'll say, okay, this is the pressure, right?

00:31:39 Cedric Armand

And then they need to say, okay, I need, this is the solutions I need to meet those demands, right?

00:31:45 Cedric Armand

And, one of them would be virtualization and simulation.

00:31:48 Cedric Armand

And then the leadership has to say, okay, I will give you the money to invest.

00:31:52 Cedric Armand

If this is what you say you need to meet what I'm asking, then here's the money, you know, go and use it, et cetera, et cetera.

00:31:58  Suresh Sivavarman

So,

00:31:59  Suresh Sivavarman

Virtualization is an exciting space, right?

00:32:02  Suresh Sivavarman

I think it's a lot of things going on there.

00:32:04  Suresh Sivavarman

But there's a buzzword running around right now with AI.

00:32:08  Suresh Sivavarman

Where do you see AI kind of not just improving the efficiency, but maybe taking some of the burden from some of the early investments and making it simpler to allow AI basically enabling you to be able to use virtualization?

00:32:28  Suresh Sivavarman

faster.

00:32:28  Suresh Sivavarman

Where else do you see AI playing a role in this space?

00:32:32  Suresh Sivavarman

Like maybe with test generation or fault analysis, or where do you see that?

00:32:38  Suresh Sivavarman

Where do you see it really growing?

00:32:40 Cedric Armand

Absolutely.

00:32:41 Cedric Armand

So I'll get like very, well, I'll say generally, right?

00:32:45 Cedric Armand

I think everybody.

00:32:47 Cedric Armand

is familiar with AI in different ways, right?

00:32:49 Cedric Armand

And that it can just be an efficiency gain in what you're doing, right?

00:32:52 Cedric Armand

So if your job is to write test scripts, right, it can write test scripts faster or code it for you faster, right, do things, right?

00:32:58 Cedric Armand

It could just be an enabler in general.

00:33:01 Cedric Armand

Very specifically, there are two use cases, one that kind of realized as more of an adoption tool, but another was as a use case as an efficiency gain.

00:33:14 Cedric Armand

So the first, the efficiency gain I'll say is

00:33:17 Cedric Armand

When a software developer or tester is using a virtual environment to develop their software, right?

00:33:27 Cedric Armand

And they develop some software and they run a test, and let's say that test fails or some unexpected results, right?

00:33:35 Cedric Armand

When that happens, to be frank, you do not know if it was a true software bug.

00:33:42 Cedric Armand

if there was a problem with the virtual ECU, if it was a problem with the model or some models that were within that environment, or if it was a problem with the test script that ran the test, right?

00:33:55 Cedric Armand

So there's a number of things that you have to do, what I call is like a first level triaging to figure out, hey, where is this issue coming from, right?

00:34:03 Cedric Armand

And that's, and that's the honest, you know, thing that occurs type of testing, right?

00:34:08 Cedric Armand

And, you know, AI, and I,

00:34:12 Cedric Armand

seen this demonstrated, that is a very good tool to use to do that first level triaging, right?

00:34:19 Cedric Armand

Because you can train a model to know some things initially, but then every time you run a test, and especially when it fails, right, you can take that information, right, and continue to learn it so that, when it happens again or something similar, right, it starts to learn so that, yes, when a test fails, you know, I'm making it very much more simple, but at first it pops out and says,

00:34:40 Cedric Armand

your AI agent comes out and says, Hey, I think it's a test script issue, you know, go look at this, right?

00:34:45 Cedric Armand

So, at least it points you and helps you kind of guide you a little bit faster.

00:34:48 Cedric Armand

That's one aspect.

00:34:50 Cedric Armand

The other is, which I guess is simple, but just kind of AI agents where, you know, at scale, when you're when you have a virtual platform and you have virtual environments and

00:35:07 Cedric Armand

test suites and tools and everything, all that stuff, right?

00:35:10 Cedric Armand

It's a lot for a user to, come on board and figure out how to use it, right?

00:35:15 Cedric Armand

And, back in the day, we used to have, physical manuals, right?

00:35:19 Cedric Armand

Then we started to have, help functions and tutorials that are online.

00:35:23 Cedric Armand

And now it's like, there's no way I can read through or understand, right?

00:35:27 Cedric Armand

And so, just to like learn or figure out how to do something, whatever, right?

00:35:31 Cedric Armand

You just ask your AI agent that's attached to, say, your virtual platform,

00:35:36 Cedric Armand

And that's going to enable users to adapt or adopt the methodology and the tools that you're bringing to them much easier and faster.

00:35:46  Suresh Sivavarman

Nope.

00:35:47  Suresh Sivavarman

So we fast forward five years, right?

00:35:51  Suresh Sivavarman

You've been in this role at Ford.

00:35:53  Suresh Sivavarman

What does virtual first or virtual Ford look like five years from now?

00:35:59  Suresh Sivavarman

What would you consider will be a success factor for you?

00:36:04 Cedric Armand

Yeah.

00:36:04 Cedric Armand

So in five years, this is assuming I don't win the lottery, right?

00:36:09 Cedric Armand

Yeah, exactly.

00:36:11  Suresh Sivavarman

Or maybe you're so passionate even after you win the lottery.

00:36:14 Cedric Armand

So passionate, I win the lottery, but I'm still in it for the love of the game.

00:36:17 Cedric Armand

All right, let's go.

00:36:18 Cedric Armand

All right.

00:36:20 Cedric Armand

So in five years, right, what I would expect is we have like a fully mature

00:36:29 Cedric Armand

virtual platforms, so meaning you, www..forvirtualization.com, where our internal users, as well as our tier one partners, are able to go in and access all the virtual artifacts they need, as well as the tools they need, environments, et cetera, tests to do their software development and their software testing, right?

00:36:56 Cedric Armand

In five years, you know, I want to have

00:36:59 Cedric Armand

at least half of our software developers within Ford, using the platform, actively, right?

00:37:06 Cedric Armand

So having it be used at large scale.

00:37:08 Cedric Armand

In five years, I want, our cloud provider bill, to be high, meaning that we're just, we're running millions and millions and millions of tests, right?

00:37:21 Cedric Armand

Valuable tests, obviously, but the point is that, we're investing and we're using this at scale.

00:37:26 Cedric Armand

And then

00:37:28 Cedric Armand

As we're, as we develop, new software stacks, new vehicle architectures, right?

00:37:34 Cedric Armand

I want virtualization to be at the forefront, to be used at the very outset so that we make the right decisions early enough, right?

00:37:45 Cedric Armand

We learn faster, right?

00:37:46 Cedric Armand

We don't have to wait on hardware, you know, to kind of figure out some things.

00:37:51 Cedric Armand

So that's where I hope for it is in five years.

00:37:56  Suresh Sivavarman

Can you do it alone or do you need partners?

00:37:58 Cedric Armand

You need partners for sure.

00:38:00 Cedric Armand

So, I'd say to develop like a platform, so kind of like for, the platform that uses to kind of hop on, there are partners who know how to do these things.

00:38:14 Cedric Armand

There are some off-the-shelf solutions from like engineering service providers, right?

00:38:18 Cedric Armand

But

00:38:21 Cedric Armand

You want to leverage their expertise, right?

00:38:23 Cedric Armand

You can do these things in-house and whatever, but it's just not going to be as good, right?

00:38:27 Cedric Armand

You're not going to have the lessons learned from other OEMs, right, and the general industry knowledge, right?

00:38:33 Cedric Armand

So you need that.

00:38:34 Cedric Armand

You need cloud provider partners, right, who are invested in automotive solutions because they will have frameworks and have done this before, right, to kind of enable them and go faster and have some lessons learned.

00:38:50 Cedric Armand

You need like tool chain providers, right?

00:38:52 Cedric Armand

Like others who, work in the co-simulation and simulation to provide their tools and bring their tools that you naturally use on hardware to have versions that can be used in virtual environments and are, you know, cloud native.

00:39:07 Cedric Armand

We need partners to create virtual ECUs, right?

00:39:11 Cedric Armand

We didn't get into like the different levels of virtual ECUs and that's okay, right?

00:39:14 Cedric Armand

But there are different partners who have expertise in having

00:39:19 Cedric Armand

hardware emulation, levels of virtual ECUs and that expertise, right?

00:39:25 Cedric Armand

So you definitely need a lot of partners to make it work.

00:39:29 Cedric Armand

And the other benefit of not only getting it to work is that if you do it all by yourself or more by yourself than you probably should, it's very hard to stay on the forefront and the cutting edge, right?

00:39:43 Cedric Armand

You mentioned, you know, virtualization is exciting.

00:39:45 Cedric Armand

It is exciting.

00:39:46 Cedric Armand

And it is.

00:39:47 Cedric Armand

because it is always evolving and new things are being added, right?

00:39:52 Cedric Armand

And so being aligned with partners helps you kind of stay on that cutting edge and continue to bring innovation to it, right?

00:40:00 Cedric Armand

So today, I might know, hey, these are all the best things in virtualization and simulation.

00:40:04 Cedric Armand

That's great, right?

00:40:06 Cedric Armand

And even if I win the lotto and I'm still doing it for the love of the game, I can, yeah, sure, stay on top of it as best I can.

00:40:12 Cedric Armand

But

00:40:13 Cedric Armand

It's too big.

00:40:14 Cedric Armand

It's too big a space, right?

00:40:14 Cedric Armand

There's no way, right?

00:40:16 Cedric Armand

And also, you know, I'm not going to come up with all the best ideas in the world, right?

00:40:20 Cedric Armand

So partnerships are not only vital, they're just absolutely necessary, right?

00:40:25 Cedric Armand

You can't do it.

00:40:26  Suresh Sivavarman

I fully agree with you.

00:40:27  Suresh Sivavarman

I think we're in the same mindset here at ETAS, right?

00:40:30  Suresh Sivavarman

There's things that we know what we're capable of, but there's also others in the industry who know what they're good at.

00:40:37  Suresh Sivavarman

And if we want to get to market faster and we want to be able to supply companies like Ford with the right tooling,

00:40:43  Suresh Sivavarman

working together with others to build a common platform or a common ecosystem, I think enables everyone to be more successful.

00:40:52 Cedric Armand

Yeah.

00:40:54 Cedric Armand

When I first got into, I'll say like automotive and controls and software development, right?

00:40:59 Cedric Armand

I got introduced to ETOS and Inka, right?

00:41:03 Cedric Armand

Everybody knows Inka, right?

00:41:04 Cedric Armand

Use it for calibration and things like that.

00:41:06 Cedric Armand

And now, at the point where it's like, okay, we want to do calibrations and do it

00:41:13 Cedric Armand

virtually, right?

00:41:14 Cedric Armand

It's like, obviously, we partnered with, our companies would partner with ETOS, for the calibration solutions.

00:41:21 Cedric Armand

And you can, now you're saying, I want to go to ETOS and be like, hey, I want to make sure I'm able to do this in virtual environments as well, right?

00:41:29 Cedric Armand

So, some of these partnerships are like kind of continued and you just, you're kind of bringing maybe companies or partners who

00:41:36 Cedric Armand

we're really successful in the hardware space and you say, hey, this is where we're going.

00:41:39 Cedric Armand

We're going to go there, right?

00:41:40 Cedric Armand

And some are who kind of natively just kind of focus in on virtual solutions and things like that.

00:41:48  Suresh Sivavarman

No, I think that's a great insight.

00:41:52  Suresh Sivavarman

Well, Cedric, I think that was excellent discussion.

00:41:55  Suresh Sivavarman

I really enjoyed the conversation.

00:41:56  Suresh Sivavarman

I think we got a lot of insights into what your vision is and where you guys are heading towards.

00:42:02  Suresh Sivavarman

Really appreciate you jumping on here.

00:42:04  Suresh Sivavarman

and having this conversation and kind of bringing all this information really to all our audiences.

00:42:10  Suresh Sivavarman

So I appreciate your time and definitely your willingness to.

00:42:15 Cedric Armand

Join.

00:42:15 Cedric Armand

Absolutely.

00:42:16 Cedric Armand

Really appreciate the invite and having me on.

00:42:19 Cedric Armand

It was a lot of fun.

00:42:21 Cedric Armand

I do geek out on this stuff.

00:42:23 Cedric Armand

I do do it for the love of the game.

00:42:25 Cedric Armand

And so, you know, hopefully, you know, you'll have me back at some point.

00:42:30  Suresh Sivavarman

We definitely will.

00:42:31  Suresh Sivavarman

I'd like to explore more of this journey as you go through it, right?

00:42:35 Cedric Armand

Yeah, fantastic.

00:42:36  Suresh Sivavarman

All right, well, thank you.

00:42:37  Suresh Sivavarman

Thanks everyone for joining.

00:42:41 Voiceover

Thank you for joining this episode of the Empowering Tomorrow's Automotive Software Podcast.

00:42:45 Voiceover

Please leave a comment or review with your feedback or what you'd like to hear in future episodes.

00:42:51 Voiceover

To learn more about automotive embedded systems and ETAS's capabilities, visit our website at ETAS.

00:42:56 Voiceover

That's E-T-A-S.com.