Exploring Virtual ECUs and the Transition to SIL (Software-in-the-Loop)

In this episode, host Suresh Sivavarman sits down with Sangeeta Theru, Virtual Validation Platform Lead at Stellantis. Together, they explore how the automotive industry is evolving from traditional hardware-centric testing frameworks to software-first engineering environments.

As software deployment velocity continues to outpace physical hardware availability, strictly relying on traditional hardware-in-the-loop (HIL) setups is shifting from a core strategy to an operational bottleneck. Sangeeta shares how Stellantis leverages a "shift-left" mindset by scaling software-in-the-loop (SIL) platforms to achieve faster time-to-market, minimize late-stage risks, and drastically lower the overhead costs of physical vehicle prototypes.

Key topics in this episode include:

• The "Shift-Left" Catalyst: Why system-level digital validation is essential for compressed, risk-mitigated development timelines.
• Overcoming the Trust Gap: How engineers can build confidence and establish reliable correlation between virtual behavior and physical target faults.
• What to Test in SIL vs. HIL: Understanding why 70% to 80% of functional and feature validation can bypass physical hardware requirements.
• Demystifying Virtual ECUs (VECUs): A breakdown of different abstraction levels, specifically unlocking the power of Level 3, Level 3 Plus, and Level 4 variants for system-level confidence.
• Global Complexity Management: How virtualization scales to support immense variation across complex global automotive brand portfolios.

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.

00:00:31 Suresh Sivavarman

Hello everyone and welcome to another episode of Empowering Tomorrow's Automotive Software from ETAS.

00:00:38 Suresh Sivavarman

I'm your host Suresh and I'm happy to welcome Sangeeta Theru today, a recognized leader in the virtualization and automotive verification platform from Stellantis.

00:00:48 Suresh Sivavarman

Sangeeta currently leads the virtual validation platform at Stellantis and she's the driving vision and execution of the engineering workbench solutions.

00:00:57 Suresh Sivavarman

Sangeeta comes with many years of hands-on experience in hardware-in-the-loop, software-in-the-loop, and validation testing.

00:01:03 Suresh Sivavarman

And she has played a pivotal role in moving the industry from hardware development testing towards software-first environment testing.

00:01:11 Suresh Sivavarman

In today's episode, we will be diving a little bit into virtual validation and how we can use virtual ECUs to shift left and reduce our reliance on hardware-based hardware-in-the-loop benches and vehicle-based testing.

00:01:25 Suresh Sivavarman

Sangeeta, welcome to the podcast.

00:01:26 Suresh Sivavarman

It's great to have you here.

00:01:28  Sangeeta Theru

Thanks, Suresh.

00:01:29  Sangeeta Theru

It's great being here.

00:01:30  Sangeeta Theru

And I think this would be a great opportunity to deep dive into the XIL topics and virtual validation at scale.

00:01:38 Suresh Sivavarman

Yeah, today I'd like to talk a little bit about how software velocity is outpacing hardware, right?

00:01:45 Suresh Sivavarman

Meaning how can we move testing and validation further up the validation chain so we can get to the market faster without relying on having to get hardware available to us.

00:01:55 Suresh Sivavarman

And the other thing is I want to focus a little bit about how hardware in the loop can be considered a bottleneck, not a strategy.

00:02:03 Suresh Sivavarman

How do we bypass that?

00:02:05 Suresh Sivavarman

How do we move forward?

00:02:06 Suresh Sivavarman

And then positioning virtual validation as an inevitable means of developing software rather than experimental means.

00:02:15 Suresh Sivavarman

So what I want to focus on first, Sangeeta, is a little bit about why does Stellantis and the industry in general want to shift left

00:02:25 Suresh Sivavarman

What is the driving factor that you see that's making you make this change?

00:02:31  Sangeeta Theru

Absolutely, Suresh.

00:02:32  Sangeeta Theru

I think the first thing, again, we have to understand here is why are companies, let alone only automotive, right, pushing digital validation so hard?

00:02:42  Sangeeta Theru

Actually, right now, a lot of focus has been there on how to shift left, digital validation.

00:02:49  Sangeeta Theru

This world of simulation is not new to automotive industry, right?

00:02:52  Sangeeta Theru

We've been using simulation methods quite extensively in a lot of our development activities right from the beginning.

00:03:01  Sangeeta Theru

If you talk about like CAE analysis for vehicle dynamics, performance, structural crash performance, so on and so forth, all of these concepts have been used in early development processes, especially in the automotive industry, very early on.

00:03:19  Sangeeta Theru

But what is different right now is how do we scale this?

00:03:23  Sangeeta Theru

How do we move this virtual need and virtual development into a complete system-based integration and testing, right?

00:03:32  Sangeeta Theru

With software as a focus, because again, software is becoming big right now.

00:03:40  Sangeeta Theru

We are basically moving into a world where mobility, mechatronics is all merging together, right?

00:03:47  Sangeeta Theru

And right now the focus is more on how to begin early virtual validation at a system level.

00:03:57  Sangeeta Theru

So one of the major things is obviously speed.

00:04:01  Sangeeta Theru

We don't want systems to be developed and validated at a very late phase in the process, in the vehicle development process.

00:04:11  Sangeeta Theru

We want to condense that shorter and so that we can integrate earlier, we can test earlier,

00:04:17  Sangeeta Theru

so that we can prevent risks.

00:04:20  Sangeeta Theru

What if our system fails?

00:04:22  Sangeeta Theru

Can we find those issues earlier on in the process so that we can fix it early to prevent late detection of issues, which adds on to cost, warranty, all of the challenges that we have to face from a root cause analysis and also fixing at a very late phase in the process.

00:04:46  Sangeeta Theru

Again, I alluded a little bit from the development cost standpoint, but also in the early development processes when the hardware doesn't exist, right?

00:04:56  Sangeeta Theru

And typically we end up with such kind of situations, right?

00:05:00  Sangeeta Theru

ECUs may not be available.

00:05:01  Sangeeta Theru

The chipset is probably under development, right?

00:05:04  Sangeeta Theru

And in such kind of scenarios,

00:05:07  Sangeeta Theru

does software development need to stop and wait for all of this hardware development to complete before we can start validation or start development?

00:05:17  Sangeeta Theru

My answer to that is no.

00:05:20  Sangeeta Theru

We should start development even when hardware doesn't exist because it gives us the leverage to be ahead of in the cycle.

00:05:29  Sangeeta Theru

And so with that comes, okay, what is the benefit of having virtual validation when hardware does exist, right?

00:05:37  Sangeeta Theru

And obviously, cost of physical prototypes is way higher.

00:05:42  Sangeeta Theru

If you want to start development and testing in an integrated environment, building such kind of system HILs or building system or vehicles, building vehicles is going to be costlier.

00:05:55  Sangeeta Theru

And it's definitely better to do such kind of development work and validation work using software in the loop platforms early on.

00:06:07  Sangeeta Theru

The other one is the scalability, right?

00:06:08  Sangeeta Theru

All of your test scenarios and corner cases that you want to kind of like validate in a vehicle environment, can we even do that?

00:06:19  Sangeeta Theru

Can we even exercise the bandwidth of the test cases that we want to test in a vehicle in an actual vehicle environment?

00:06:28  Sangeeta Theru

ADAS is an example, right?

00:06:30  Sangeeta Theru

All the autonomous features that we want to test in a vehicle, can we actually do that in a complete vehicle environment?

00:06:37  Sangeeta Theru

I would say it's quite impossible.

00:06:40  Sangeeta Theru

We have to move some of that testing into a more virtual environment so that we can exercise the system again very vastly.

00:06:51  Sangeeta Theru

And then comes, again, that talks a little bit about the scalability.

00:06:55  Sangeeta Theru

Also talking about the variant management.

00:06:58  Sangeeta Theru

It's huge.

00:06:59  Sangeeta Theru

We are a company with 14 brands spread out globally, right?

00:07:06  Sangeeta Theru

So and we have a lot of variations between vehicle to vehicle, brand to brand.

00:07:14  Sangeeta Theru

Can we basically test one variant and call it a feature set that will be applicable to other variants?

00:07:24  Sangeeta Theru

We cannot.

00:07:25  Sangeeta Theru

And taking the variants into a development process, can we build vehicles?

00:07:32  Sangeeta Theru

Can we build physical prototypes for all the variants that we are building, right?

00:07:36  Sangeeta Theru

And so this is where virtual validation really, really becomes a game changer.

00:07:42  Sangeeta Theru

So with that, I think we can go into a little bit of a detail.

00:07:46  Sangeeta Theru

I think your question was more about HIL versus SIL, right?

00:07:50 Suresh Sivavarman

Yeah, maybe I have a follow-up question there.

00:07:53 Suresh Sivavarman

I think you mentioned you want to be able to find issues before hardware even exists.

00:08:00 Suresh Sivavarman

Someone like myself came from

00:08:02 Suresh Sivavarman

very much a hardware testing environment.

00:08:04 Suresh Sivavarman

We validated everything at a system level, on a car or on a physical ECU, to convince people like me or others like me, what kind of issues can you find in a software in a loop virtual environment that can basically ease the burden of having to retest it or validate it on a physical hardware?

00:08:27 Suresh Sivavarman

And what is the confidence level of doing that?

00:08:32  Sangeeta Theru

That's a great question, right?

00:08:33  Sangeeta Theru

I think it's a very common problem statement, at least in this, when we talk about HIL and SIL, right?

00:08:42  Sangeeta Theru

So far, what we have identified is the biggest hurdle is not the tool capability, right?

00:08:48  Sangeeta Theru

We already have SIL platforms running at scale.

00:08:54  Sangeeta Theru

It is the confidence and correlation between what runs virtually versus

00:09:00  Sangeeta Theru

how does that compare to an actual fault, actual issue that I figured out in the hardware system, correct?

00:09:10  Sangeeta Theru

So again, engineers like yourself, like you mentioned, are going to be trusting a virtual ECU only to an extent, or even a SIL platform, you know, only to an extent that it is proven by back-to-back testing, repeatable, right?

00:09:28  Sangeeta Theru

You run one test,

00:09:30  Sangeeta Theru

Does it provide the same results the second time?

00:09:34  Sangeeta Theru

And also can provide the similar results like a physical ECU or physical system under same conditions, right?

00:09:46  Sangeeta Theru

Getting to the point, I think it's got several enablers, several reasons.

00:09:52  Sangeeta Theru

And in general, I don't think it's the exact resemblance between physical hardware and

00:09:58  Sangeeta Theru

software in the loop that we have to really go behind as a validation point, right?

00:10:02  Sangeeta Theru

We'll talk about it a little bit.

00:10:04  Sangeeta Theru

But answering your question about what is fundamentally, what could we test in an SIL platform, any testing that does not need timing determinism, right?

00:10:18  Sangeeta Theru

Anything that does not need that cycle accurate realism between hardware and software.

00:10:27  Sangeeta Theru

I think that's something where we move the testing into a hardware environment, a test bench and HIL, so on and so forth.

00:10:34  Sangeeta Theru

Because if you think about it, physical ECUs do have, you know, jitter.

00:10:39  Sangeeta Theru

You have the instructions that, you know, service routines that kind of like, you know, have to interact correctly in the system so that you can execute a function feature accurately in the right order also, right?

00:10:55  Sangeeta Theru

They may also behave very differently under thermal stress or even in a stress condition, voltage, thermal stress.

00:11:03  Sangeeta Theru

All of these testing need to happen with the physical ECU, CPU bandwidth.

00:11:10  Sangeeta Theru

That's another great example, right?

00:11:12  Sangeeta Theru

I don't think you can move such kind of tests into a complete virtual validation environment.

00:11:18  Sangeeta Theru

But that being said, that's typically about 20 to 30% of the testing.

00:11:23  Sangeeta Theru

What can we do for the remainder of the 70 to 80% of the test, right?

00:11:29  Sangeeta Theru

We have to use still platforms.

00:11:32  Sangeeta Theru

Why?

00:11:33  Sangeeta Theru

Because we don't need instruction accurate, cycle aware test execution in order to test the remainder of the 70 to 80%.

00:11:43  Sangeeta Theru

I'm talking more about the functional validation, feature validation that you typically do.

00:11:50  Sangeeta Theru

at the vehicle level or even at the ECU level, be it even component, right?

00:11:55  Sangeeta Theru

But the bigger bang for the buck comes when you start integrating more and more ECUs at an earlier phase of development and you start early integration tests in an integrated environment with

00:12:08  Sangeeta Theru

virtual ECUs, plant models, bus communication, and also putting in, throwing in the test scenarios, the road models, environment models, everything to test specific feature functionality, be it autonomous, be it infotainment-based, body domain, so on and so forth.

00:12:26  Sangeeta Theru

I think that's the biggest advantage of using virtual validation, especially in a SIL environment.

00:12:33 Suresh Sivavarman

So I think this leads to a perfect next question.

00:12:36 Suresh Sivavarman

I think we talked a lot about virtual ECUs and virtual validation.

00:12:40 Suresh Sivavarman

How do you define a virtual ECU?

00:12:44 Suresh Sivavarman

We've heard of level 1, level 2, level 3, level 4.

00:12:48 Suresh Sivavarman

My understanding is level 1, level 2, you can do application later type of activity.

00:12:53 Suresh Sivavarman

But we're talking about software in the loop with this level of detail.

00:12:57 Suresh Sivavarman

What value do you see in type 3 and type 4

00:13:02 Suresh Sivavarman

Virtual ECUs on level 3 and level 4 virtual ECUs that help close this gap between what an HIL or hardware in the loop system would do and what a software in the loop system can do.

00:13:13 Suresh Sivavarman

So what are we really unlocking in terms of system level confidence with like this type 3, let's say type 3 virtual ECU?

00:13:21  Sangeeta Theru

I would expand this into even type 3 plus, which is another topic of discussion here, right?

00:13:28  Sangeeta Theru

So again, going back to the definitions of what level 0 to level 5 ECUs are, level 0 is basically controller models, a subset of application models that you will run typically in a MIL environment or an early SIL environment.

00:13:48  Sangeeta Theru

Level 1 will include some application code, right, for test coverage of

00:13:56  Sangeeta Theru

application software, ASW.

00:13:59  Sangeeta Theru

Level 2 may contain some simulated base software included in the VCU.

00:14:07  Sangeeta Theru

Level 3 comes where there is production based software integrated with the production application software as an integrated package, basically wrapped up, built for

00:14:22  Sangeeta Theru

x86 target, right?

00:14:23  Sangeeta Theru

Where typically we're talking about classic AUTOSAR where the MCU layer is emulated and you basically have any layer above the MCU or the MCAL layers real, right?

00:14:39  Sangeeta Theru

All of the software code that goes into the target ECU.

00:14:44  Sangeeta Theru

So that's level 3.

00:14:46  Sangeeta Theru

And then level 4 VECU is where you have typically like hardware models, be it SOC, an MCU, or even the peripheral models, right?

00:14:58  Sangeeta Theru

All of that simulated on a computer environment on an x86 or an ARM Graviton instance.

00:15:06  Sangeeta Theru

And you basically have the code flashed directly as an executable onto these models.

00:15:13  Sangeeta Theru

Now, something in between level 3 and level 4 is level 3 plus.

00:15:17  Sangeeta Theru

And what does that mean?

00:15:19  Sangeeta Theru

It's very similar to level 3 production-based software, but more fidelity drawn into the lower layers of the hardware, not generally the actual SOC or the MCU model with peripherals running, but more so maybe an ARM SOC actually modeled on cloud.

00:15:42  Sangeeta Theru

It may not be the actual plan of record SOC or MCU, but it could be closed on target, especially for simulating or emulating the infotainment systems or camera-based systems, autonomous systems, which brings in a little bit more fidelity into the software layers that we want to build and deploy and test on cloud, right?

00:16:11  Sangeeta Theru

and also bringing in the hypervisor layers directly onto cloud versus using the same hypervisor on hardware targets.

00:16:25  Sangeeta Theru

So that's what we define as level 3 plus, and I know industry uses various definitions of this.

00:16:31  Sangeeta Theru

So going into where are we seeing our biggest benefit and what's the typical use case of level 3, 3 plus as compared to level 4, right?

00:16:40  Sangeeta Theru

And level 5, again, it's really CU with target hardware.

00:16:45  Sangeeta Theru

Level 3, 3 plus is where we have found the sweet spot, where we can get the cross compilation of our actual software, actual source code,

00:16:58  Sangeeta Theru

against an x86 or an ARM core target.

00:17:04  Sangeeta Theru

And the evaluation between speed of how the simulation runs versus the performance, right, or the testing that we have to do at the subsystem or the system level, we find the benefit when we build level 3 and level 3 plus VECUs use.

00:17:29  Sangeeta Theru

I think that's a learning from Stellantis site, right?

00:17:32  Sangeeta Theru

Because when we get into level 4, since we are actually building the target SOCs, target MCUs, and the peripheral as models, they tend to run slower and they tend to, you know, not, especially when you start integrating these ECUs in a system environment, it makes the simulation run very, very slow.

00:17:57  Sangeeta Theru

And especially in a CICD environment or even for a developer use case, this becomes a kind of like a user experience that is not really very wanted, right?

00:18:14  Sangeeta Theru

So level 3 and level 3 plus run much better and strike the balance between speed and performance is what we have learned.

00:18:25  Sangeeta Theru

running at scale and at the same time running at a target speed that is very useful for a developer or a testing engineer to get results faster.

00:18:41  Sangeeta Theru

And even introducing these in the CI/CD workflows where we can run nightly builds, nightly test cases, I would say that's the sweet spot.

00:18:50 Suresh Sivavarman

So I think I'd like to explore that a little bit, right?

00:18:53 Suresh Sivavarman

You mentioned

00:18:55 Suresh Sivavarman

virtual ECUs running on x86 targets or ARM Cortex targets, whatever it may be, Graviton.

00:19:02 Suresh Sivavarman

But my assumption is that these are not desktop use cases anymore.

00:19:06 Suresh Sivavarman

We're not running them on a PC.

00:19:08 Suresh Sivavarman

We're scaling them to some kind of cloud capability.

00:19:13 Suresh Sivavarman

In this software-in-the-loop environment, how many tests are you able to run in parallel?

00:19:19 Suresh Sivavarman

What does that scalability look like and how does it integrate into a CI/CD pipeline?

00:19:25 Suresh Sivavarman

Are these tests done all headlessly?

00:19:29 Suresh Sivavarman

Are they scripted-based CLI tests or are you actually, is there someone actually able to use a GUI, see what's going on, or is it done after the fact where you get the results and then you visualize it?

00:19:41 Suresh Sivavarman

What does the deployment look like going from a traditional PC-based testing to a scalable

00:19:48 Suresh Sivavarman

of cloud integrated testing?

00:19:50 Suresh Sivavarman

What does that environment look like?

00:19:54  Sangeeta Theru

To answer your question, I think it's more distributed, right?

00:19:57  Sangeeta Theru

Again, there are different user communities on how the users use the platforms.

00:20:04  Sangeeta Theru

I would say the development community who have hands-on code, who are building code every second, they tend to, and want to test the code right away.

00:20:18  Sangeeta Theru

They have or they like the manual interaction between the SIL environment and their test that they want to run on this environment.

00:20:28  Sangeeta Theru

Make a change in the code.

00:20:30  Sangeeta Theru

build the code, integrate the platform, and run a manual test, understand if the change that they made actually resolved or resolved an issue that they were trying to resolve or, basically going back into whether it met the feature requirements, right?

00:20:50  Sangeeta Theru

But that portion of user community is, I would say, getting less and less as we move into more

00:20:59  Sangeeta Theru

automation and CICD DevOps practices using DevOps practices in automotive software development.

00:21:08  Sangeeta Theru

Here, I think SIL platforms at scale become really a huge enabler because what we have actually proved out is a subset of like 4,500 test cases run at scale on the cloud every night, built and integrated as a platform and run on

00:21:28  Sangeeta Theru

Run on cloud nightly, and this is completely headless, right?

00:21:33  Sangeeta Theru

You're starting off your workflow with just pushing the code into the changed code into a workflow where...

00:21:41  Sangeeta Theru

pipelines and automation kind of like takes it from building of the software to the platform building to running, deploying like the platform so that you can grab the test cases that you want to run on this platform all the way up to like publishing test results and the engineer basically comes back the following day morning to analyze the pass and the fail tests.

00:22:04  Sangeeta Theru

so that they can bring in the new software, or rerun some of the tests that they found failed for root causing and other purposes, right?

00:22:14  Sangeeta Theru

So what we have figured out is with Phil, we are able to parallelize this testing, pull in more instances, right?

00:22:23  Sangeeta Theru

Run these instances more seamlessly for this end-to-end workflow that we are talking about.

00:22:30  Sangeeta Theru

It's definitely a cheapest solution, a cheapest solution at least 2.5 times less than running these tests parallely on hardware in the loop systems.

00:22:43  Sangeeta Theru

This is the comparison we have done.

00:22:44  Sangeeta Theru

At least the deployment, testing, basically providing the results, runs completely headless.

00:22:53  Sangeeta Theru

And it has proven to be a much more usable use case within Stellantis than actually the manual testing.

00:23:00  Sangeeta Theru

But of course, again, it's a split.

00:23:03  Sangeeta Theru

For a developer use case who wants to see results for the change that they have done, they use manual testing.

00:23:10  Sangeeta Theru

But a majority of that is also moving into automated test cases, run overnight, so on and so forth.

00:23:17 Suresh Sivavarman

So even hardware in the loop can have automated testing, right?

00:23:22 Suresh Sivavarman

So is there a possibility to have a hybrid approach where you do an X number of tests on a software in the loop system and then immediately it can trigger a set of sequences or a set of tests in the hardware in the loop?

00:23:35 Suresh Sivavarman

Can this all also be automated and kind of housed in a cloud-based environment where not only are you getting the virtual target, but you also get some results out of the physical target?

00:23:47  Sangeeta Theru

Absolutely.

00:23:48  Sangeeta Theru

I think this is where the split between

00:23:50  Sangeeta Theru

70 to 80% on virtual versus 20 to 30% on the HIL.

00:23:56  Sangeeta Theru

This study, I mean, this result that came out of the study, I think it's very important because exactly to your point, we can run parallel agents, SIL agents, to cover 70 to 80% of our testing.

00:24:14  Sangeeta Theru

And simultaneously, we can

00:24:18  Sangeeta Theru

filter out any of the hardware-based tests that you have to perform and scale this on the hardware-in-the-loop environment parallelly again.

00:24:27  Sangeeta Theru

But you will use less number of HILs as compared to the SIL agents that you would use.

00:24:35 Suresh Sivavarman

So I think this kind of goes into my next section of questions, right?

00:24:38 Suresh Sivavarman

You can almost use the hardware-in-the-loop or HIL kind of as a verification for some engineers to feel comfortable that the results are

00:24:47 Suresh Sivavarman

validated from the SIL.

00:24:49 Suresh Sivavarman

But in my opinion, I think talking to many people around the industry, also within our own organization, the hardest blocker I see is getting people to believe that this is the right way to work, that this is a trustworthy and validated version of results that can be used to make decisions.

00:25:14 Suresh Sivavarman

I see that there's a gap.

00:25:15 Suresh Sivavarman

There's a gap in trusting the virtual results.

00:25:18 Suresh Sivavarman

An engineer typically comes from ownership, right?

00:25:21 Suresh Sivavarman

They own a feature, they own a function, they've tested it, they validated it.

00:25:25 Suresh Sivavarman

In the modern world, you may code it, but you push it to the cloud, someone else is validating it, or some entity is validating it, and it's being consumed and pushed forward.

00:25:38 Suresh Sivavarman

In your role, how have you been able to

00:25:42 Suresh Sivavarman

successfully or even unsuccessfully convince people that this is the right mindset to shift left?

00:25:50 Suresh Sivavarman

How do you encourage people to try this and believe in it?

00:25:54 Suresh Sivavarman

Because as you stated, this is a vital way for us to get to market faster.

00:25:59 Suresh Sivavarman

This is the way we are going to reduce our development time.

00:26:02 Suresh Sivavarman

But there's always going to be a bottleneck in the people and the personalities.

00:26:06 Suresh Sivavarman

What have you seen in your experience that has kind of

00:26:11 Suresh Sivavarman

help mitigate that gap or lessons learned for people to understand moving forward.

00:26:16  Sangeeta Theru

That's a great question, Suresh.

00:26:17  Sangeeta Theru

I think it's the psychological trust versus technical trust, right?

00:26:22  Sangeeta Theru

It's one of the biggest, I would say, challenges that we face as an industry, right?

00:26:32  Sangeeta Theru

Coming from a very traditional automotive, OEM,

00:26:38  Sangeeta Theru

I myself have I started my career testing and developing in vehicles and I grew my interest to virtual validation and development through experience and through really understanding what benefits does simulation

00:27:01  Sangeeta Theru

actually can play in doing things earlier in a more easier way than waiting for vehicles, waiting for prototypes to be built, or even changing vehicles for rapid prototyping purposes, where typically I will probably not be able to complete my testing that I want to do in a particular day because of a number of challenges that could occur.

00:27:26  Sangeeta Theru

So it's always the balance between what do I want to accomplish

00:27:31  Sangeeta Theru

versus what can I accomplish?

00:27:36  Sangeeta Theru

So as a senior engineer, right, we all learn to debug in a very hardware-based environment, probing pins, oscilloscopes, right, hardware in the loop systems, basically to understand hardware modes, failure modes very intricately, and things like that.

00:27:55  Sangeeta Theru

So

00:27:56  Sangeeta Theru

For those people, obviously this whole topic of virtual ECU is going to feel more abstract.

00:28:04  Sangeeta Theru

I mean, it's a software problem, right?

00:28:06  Sangeeta Theru

It's not a hardware problem, right?

00:28:10  Sangeeta Theru

So how can we mitigate this?

00:28:11  Sangeeta Theru

And again, I'm sure this is a common challenge, right?

00:28:16  Sangeeta Theru

And how do we mitigate this?

00:28:19  Sangeeta Theru

Obviously, this needs to be very...

00:28:24  Sangeeta Theru

closely driven by both a team that is delivering these platforms versus the team that is actually on the receiving end trying to work through their daily work, whatever they do in the hardware, and slowly and steadily moving those test cases or testing into the environment that is now completely virtual.

00:28:51  Sangeeta Theru

right?

00:28:52  Sangeeta Theru

So the responsibility is distributed here.

00:28:55  Sangeeta Theru

The push and the pull, right?

00:28:58  Sangeeta Theru

The back-to-back testing, trying to prove out that the virtual environment actually resembles a hardware-based environment.

00:29:05  Sangeeta Theru

I think that responsibility is definitely upon a platform team, right?

00:29:12  Sangeeta Theru

For example, my team will be responsible for making sure that we are able to prove that, yeah, hardware behaves

00:29:18  Sangeeta Theru

Behavior and the software behavior is very similar to what we are seeing on the hardware in the loop, right?

00:29:27  Sangeeta Theru

But even more so, making the workflows more seamless for our customers, more easier, similar workflows, like, for example, providing debug capabilities.

00:29:41  Sangeeta Theru

breakpoint capabilities, trying to make sure that they are able to find false root cause very effectively in the software in the loop environment.

00:29:51  Sangeeta Theru

So this goes back to the tool availability, right?

00:29:54  Sangeeta Theru

So tools that they are typically used to using in the hardware in the loop environment, how can we provide a similar capability in the SIL environment?

00:30:05  Sangeeta Theru

I think focusing on those things,

00:30:07  Sangeeta Theru

Another example could possibly be like in an HIL, the test engineer, since the hardware is all connected, the electrical connections already exist, the test engineer is responsible for flashing the software on the ECU, just like how they would do in a vehicle too, right?

00:30:26  Sangeeta Theru

The test engineer is basically going to do that.

00:30:28  Sangeeta Theru

They're going to run a certain level of tests.

00:30:32  Sangeeta Theru

And then they're going to troubleshoot if the issue is on the HIL or is it a product issue, right?

00:30:42  Sangeeta Theru

But how do we make sure that we are able to differentiate that and basically make sure that the workflow enables the same test engineer to flash

00:30:58  Sangeeta Theru

not a target software, but a VECU, but the workflow is the same.

00:31:05  Sangeeta Theru

That is very critical.

00:31:08  Sangeeta Theru

building an environment and giving the capability for a test engineer to, or providing the tools for the test engineer to flash the software, just like how they would do on an advert hardware on the loop.

00:31:19  Sangeeta Theru

Those are the kind of details that make the workflows much seamless and automated for a customer community.

00:31:27  Sangeeta Theru

So those are all the things that we focus on as a platform team.

00:31:31  Sangeeta Theru

But what is the responsibility of the test engineer, right?

00:31:35  Sangeeta Theru

Or what is the responsibility of an engineer who is typically being used to the hardware in the loop?

00:31:41  Sangeeta Theru

Again, taking the tests that they typically perform in the hardware, seeing what those test cases are going to be applicable to a SIL environment,

00:31:54  Sangeeta Theru

and slowly and steadily making that transition into the SIL environment, building that trust slowly and steadily.

00:32:05  Sangeeta Theru

And with automation, CI/CD workflows, I understand the challenges of interfacing software into a hardware system, right?

00:32:16  Sangeeta Theru

Automotive systems are mechatronic systems, so it's hard.

00:32:20  Sangeeta Theru

It's hard, and especially when you talk about, I have real hardware and I'm writing software for that actuator or reading a sensor.

00:32:29  Sangeeta Theru

But we can bring in the CI/CD workflows, DevOps practices into this environment too.

00:32:35  Sangeeta Theru

But it's the trust that the testing teams or development teams have to build slowly and steadily from their end.

00:32:44 Suresh Sivavarman

Okay, so comfort zone is always a challenge, right?

00:32:50 Suresh Sivavarman

people, unless you force them, people tend to stay where they are most comfortable.

00:32:56 Suresh Sivavarman

If I feel I can test on my vehicle or on an HIL and I have my tools already, I'll continue to do so.

00:33:04 Suresh Sivavarman

What organizational levers can a company push to make this change or force the change, like reducing the number of hardware or reducing the number of vehicles?

00:33:15 Suresh Sivavarman

What kind of measures do you see that are ways to encourage

00:33:19 Suresh Sivavarman

forcefully encourage some people to make a change, so at least they start using it.

00:33:25  Sangeeta Theru

I would say, instead of using the word forcefully, it's how do we get the user community to embrace this new technology and actually start using it in their workflows, right?

00:33:42  Sangeeta Theru

So again, this is also twofold.

00:33:45  Sangeeta Theru

There is the top to bottom.

00:33:49  Sangeeta Theru

and bottom to top.

00:33:52  Sangeeta Theru

It has to come from both directions.

00:33:57  Sangeeta Theru

The top leadership has to embrace this and just like you said, have the right questions to the engineers, to the leaders, to the managers about

00:34:16  Sangeeta Theru

are you using the right methods for development and testing?

00:34:20  Sangeeta Theru

I think that's the responsibility of a leader, right?

00:34:23  Sangeeta Theru

When you're talking about top down.

00:34:25  Sangeeta Theru

And when you talk about the bottom up, it's basically how can I do my job better?

00:34:34  Sangeeta Theru

Because again, I may not end up having an HIL freely available or a vehicle freely available for doing my job all the time.

00:34:44  Sangeeta Theru

right?

00:34:45  Sangeeta Theru

How can I be more efficient in doing my job?

00:34:51  Sangeeta Theru

And that constant curiosity and thirst of finding these new technologies, new tools that can help their workflows to be more easier, more seamless.

00:35:06  Sangeeta Theru

So it needs to come from both ends.

00:35:09  Sangeeta Theru

is what I think, right?

00:35:11  Sangeeta Theru

Some push from leadership by measures of reducing the number of vehicles, reducing number of hardware in the loop systems that can be deployed for every programs.

00:35:25  Sangeeta Theru

And I think, again, I don't think we can sustain this, especially thinking about the enormous number of the variants that we have, correct?

00:35:36  Sangeeta Theru

the cost versus how much we have to test, especially keeping quality in mind.

00:35:43  Sangeeta Theru

I don't think we can sustain that with all hardware testing.

00:35:48  Sangeeta Theru

So it's basically the push from leadership to say, hey, you know what?

00:35:54  Sangeeta Theru

We're going to use only certain number of vehicles or hardware in the loop systems, subsystems, vehicle hills, whatever the systems are, right?

00:36:02  Sangeeta Theru

Plywood box in the

00:36:05  Sangeeta Theru

validation and development process.

00:36:07  Sangeeta Theru

And at the same time, the engineering community, the engineers to start thinking about how efficiently they can use these methods and platforms in their day-to-day work.

00:36:21 Suresh Sivavarman

That makes sense.

00:36:22 Suresh Sivavarman

So I guess the last question or last information to share here is what does the future look like?

00:36:29 Suresh Sivavarman

We've talked a lot about

00:36:31 Suresh Sivavarman

where we're heading, what you're doing today, some of the challenges we faced along the way, assuming we can scale and assuming that this grows, what does the acceleration look like?

00:36:41 Suresh Sivavarman

What does the end product look like?

00:36:43 Suresh Sivavarman

What is your vision of what this could be?

00:36:47  Sangeeta Theru

I think virtual validation is already happening, okay?

00:36:54  Sangeeta Theru

So it's in a pilot phase is what I would say at this point of time, right?

00:36:59  Sangeeta Theru

The next step is to understand how we can integrate more and more ECUs, more and more systems that cover more of the testing that you would typically do in a vehicle, right?

00:37:17  Sangeeta Theru

Cover multiple domains, ADAS, body, infotainment, chassis.

00:37:23  Sangeeta Theru

How can we bring all of this testing

00:37:26  Sangeeta Theru

into a more virtual environment, right, in a system environment.

00:37:30  Sangeeta Theru

We talked about how simulation has already existed, but now we're talking about truly doing testing and development in a system environment, right?

00:37:40  Sangeeta Theru

So I would say we're moving towards this goal of a virtual vehicle.

00:37:46  Sangeeta Theru

And again, we can talk about what does a virtual vehicle mean?

00:37:50  Sangeeta Theru

Do I really need to build an entire vehicle on cloud?

00:37:54  Sangeeta Theru

I don't think that's what I'm stating.

00:37:57  Sangeeta Theru

It is what do you need to test for what purpose?

00:38:02  Sangeeta Theru

And again, this differs between a team member, a calibration team member, a development software developer, a systems engineer, an architect, so on and so forth.

00:38:15  Sangeeta Theru

So building a virtual vehicle that is trusted enough by the engineering community so that we can make this

00:38:24  Sangeeta Theru

engineering and business decisions with no physical prototypes needed for development, for learning, right?

00:38:31  Sangeeta Theru

But then you still need to go into a physical environment for confirmation, for certification, so on and so forth.

00:38:42  Sangeeta Theru

I mean, do I need a physical system for architecture freeze, for software freeze, right?

00:38:49  Sangeeta Theru

Hardware sizing.

00:38:52  Sangeeta Theru

safety concept.

00:38:54  Sangeeta Theru

I don't need a physical environment for this, all of this.

00:38:58  Sangeeta Theru

We can move all of this into a more virtual development first, get to a point where you have to move into a physical environment, and close that correlation loop by running tests in a physical environment.

00:39:16  Sangeeta Theru

and not doing that only in single ECU environments, but just scaling that into subsystems and beyond to cover a majority of the vehicle validation requirements.

00:39:29  Sangeeta Theru

I think this is what I see as the future and building the tools that lay underneath that can make this more seamless, right, even from more efficient, optimized from a platform development standpoint, you know, bringing in

00:39:46  Sangeeta Theru

what I call as a system OS concept, bringing that into production and scalable environment.

00:39:57 Suresh Sivavarman

Perfect.

00:39:58 Suresh Sivavarman

That's a great vision, I think, for everyone to kind of understand and see how they can grow into that environment.

00:40:06 Suresh Sivavarman

With that, I'd like to thank you, Sangeeta.

00:40:08 Suresh Sivavarman

It was a great discussion.

00:40:10 Suresh Sivavarman

We got a lot of insights today on

00:40:12 Suresh Sivavarman

how virtual validation is being scaled as Stellantis and also what it could mean for the future of automotive software development.

00:40:21 Suresh Sivavarman

Again, thank you for participating and hope to do it again soon.

00:40:26  Sangeeta Theru

Thanks, Suresh.

00:40:26  Sangeeta Theru

Thanks, ETAS, for this opportunity.

00:40:29  Sangeeta Theru

Great discussion.

00:40:34 Voiceover

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

00:40:38 Voiceover

Please leave a comment or review with your feedback

00:40:41 Voiceover

or what you'd like to hear in future episodes.

00:40:44 Voiceover

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

00:40:49 Voiceover

That's ETAS.com.