Thought Leadership

A Closed-loop Development Process for Autonomous Vehicles – The Future Car Podcast – Transcript

By Conor Peick

The global pandemic has disrupted many aspects of society, including the processes of many major industries. The automotive industry is no exception, and over the last couple of years automakers have encountered plenty of challenges, both new and familiar. Given the sudden and disruptive nature of the pandemic, it would be no surprise if companies had to shift their focus to the near-term to secure their footing, putting longer-term plans on hold. And yet, automotive companies proved to be very resilient to these ups-and-downs, continuing to pursue their long-term plans for more electrified vehicle lineups, product innovation, and more efficient processes. But what about autonomous vehicle? Much has been made of the ongoing development of this potentially revolutionary technology, but has the pandemic changed these plans? What is the outlook for AVs today, what are the challenges, and how are companies moving forward?

In a recent episode of the Future Car podcast, we attempted to answer these questions and more. I was joined by Nand Kochhar, VP of Automotive and Transportation Industries at Siemens Digital Industries Software, and Matthieu Worm, VP of Business Development and Partnerships and Co-founder of Simulytic, a new startup venture from Siemens AG. You can listen to the episode here, or read the transcript below.

Conor Pieck
Conor Pieck, Writer – Global Marketing

[00:00] Conor Peick: The last 18 months have been a challenge for many industries, and this was no exception for the overall automotive and transportation ecosystem. But the industry has proven to be very resilient in the past few months by continuing to pursue long-term goals for product and process innovation. We see a strong shift towards vehicle electrification as nearly every OEM transfers their portfolio, and as legislation pushes new bans for combustion engine cars. As the electric vehicle leaves the early adopter phase, it will be welcomed by a broad audience of very excited customers. But what about that other massive trend? What about autonomous vehicle development? With EV plans coming to fruition, are companies refocusing their efforts on AVs? And what is the outlook for these AV programs? What are the challenges and how do they overcome them? Today, we will seek to answer these questions and more as we chat with a pair of seasoned industry experts: Nand Kochhar, VP of Automotive and Transportation Industries, and Matthieu Worm, Director of Autonomous Vehicles, both of Siemens Digital Industries Software. Nand, let me start off with you.

Nand Kochhar – VP of Automotive and Transportation Industries

[01:15] Nand Kochhar: Thank you, Conor. Pandemic, of course, in the initial phase was a little bit of a disruption as people had to figure out how are we going to handle work, working remotely, away from their offices, connectivity of when you start working remotely, things of that nature. So, I will say those were a generic the entire automotive and transportation industry had to quickly relearn wasn’t specific to AV but AV development was a part of that process of the pandemic. However, few good things happened. Number one, in my mind, pandemic helped companies realize that the power of the digital way of doing things. So, the companies which were on digital transformation journey, they already had the infrastructure in place, they had the software solutions for developing their vehicles in place, they were able to adapt quickly. And that also did one great thing, which was the reliance on simulation part of the autonomous vehicle development. And so I think that very positive things came out of something which was not all that positive – pandemic – that the reliance on simulation, trust on simulation increased. Being practical, however, it did impact the physical testing environment as people had to move away from the labs or proving grounds, initially, till there was an understanding of the CDC guidelines, or in every country, the country-specific guidelines of social distancing and things of that nature. So, it did have an impact on the physical testing and development part of it for a short period of time.

[03:04] Conor Peick: Excellent. And you hit on this already but did you see any changes in how we can make AVs a reality? Any changes to the timeline?

[03:12] Nand Kochhar: Yeah, of course, making AVs a reality, there are several things happening or continue to happen along the way. First of all, is technology. Technologies continue to mature. As you know, AV, you can divvy up into three key parts: sense, think, and act. The senses are the perception what our eyes do when there is a human as a driver. So, as you know, the technology in terms of cameras, LIDAR, radars, that continues to mature to a great level that companies are relying either just on cameras and the perception algorithms continue to improve. So, that’s a big enabler. Same thing in the thinking process. There’s a whole host of technologies for the second piece, which is processing of the information and a decision making of that information. So, that one is technologies like 5G technologies, the computing power, connectivity, everything happening through cloud, automobiles running working in a shadow mode. So, they are in a continuous learning environment through artificial intelligence and machine learning. So, all those technologies which on their own might not mean anything, but when you pull them all together, are the huge enablers for autonomous vehicle development. And then the same thing carries into the “act” piece of it. The cars or the automobiles, any of the transportation systems are not just pure mechanical anymore, there’s a huge electromechanical piece, electrical electronics, also powered by the software. So, that continues to improve allows us to do the AV development a reality.

[05:04] Conor Peick: And Matthieu, do you have anything to add there?

[05:06] Matthieu Worm: Yeah, I think the second part of your question was then the impact on the timeline. I tried to find evidence of a slowdown in the development but actually did not yet find any proof points. So, definitely, there have been some initial articles in summer last year that investments in autonomous vehicle technology would be reduced, while investments in electric vehicles would be increased. But in practice, I’ve seen still very active markets, still new partnerships being announced, still new investments being made, and also new technological breakthroughs being presented. So, honestly, I did not see yet a true proof point that the timeline has been infected. 

[05:52] Conor Peick: Interesting. So then how about consumer demand? Has consumer interest changed or faded at all?

[05:59] Nand Kochhar: On the consumer front, I think there’s still a strong interest. And that’s primarily coming from acceptance of technologies in general. As you know, a certain level of autonomy is already on the roads, and there’s more and more acceptance. And the benefits of having autonomous vehicles, they haven’t faded away. So, I think the consumer demand remains strong. And when the next generation of people come in buying automobiles are experiencing the journey on the automobiles, I think continue to have an interest in that. Obviously, there’s not really one answer for the whole world. It also depends on regions; people in Asia Pacific versus European region versus America’s might have a different environment from policies, from legal requirements, and the consumer interest go with that, better risk-taking ability, or adaption of technologies, I should say, people have different level of acceptance. So, those factors play a big role, as well, in terms of accepting of these technologies.

[07:15] Conor Peick: It’s not necessarily an interest as much as just getting people used to the idea of a self-driving car. Is that true?

[07:23] Nand Kochhar: Yeah, I mean, people have to feel comfortable, and people need to be assured that it’s going to be a safe journey. Safety, as you know, is a big factor in that. So, it is an acceptance of technologies. And the culture does play a role, and that’s why gave you a different view on the different regions. Because in general, certain parts of the world, or even in fact, some specific countries, people are more open to adapting to the new technologies, adapting to an autonomous vehicle, adapting to electrification as an example. So, not everyone in the world is on the same level of acceptance. Obviously, some are leaders and the other quick followers, and then slowly everyone else adapts those technologies.

[08:14] Matthieu Worm: And I think another indicator is that right now self-driving capabilities in highway, pilots are typically very expensive options at cars that have multi-thousand dollar or euro options, but people are willing to pay for that. And so there is a willingness to pay for self-driving technology, even on privately sold vehicles.

[08:36] Conor Peick: Could you speak to the different levels of autonomy and where we’re at currently maybe for listeners who aren’t aware of the spectrum of autonomy? 

[08:44] Matthieu Worm: In general, people use the SAE levels of automation as defined by the SAE organization, ranging from 0 to 5, where 0 is pure manual driving, and Level 5 is automated driving under all conditions and to every destination that you can think of. And then there are these intermediate steps 1, 2, 3, and 4. Today, we are roughly at Level 2, especially at premium vehicles. And Level 2 means that there is some automation like autonomous emergency braking systems that will activate the brakes in case of a pedestrian that appears in front of the vehicle without any action from the driver. So, that is kind of pure autonomy but it’s under very limited conditions. And the driver is always responsible under all conditions and needs to have his eyes on the road at all times. Whereas, if we move to Level 3, that starts to shift, and under specific conditions like a highway pilot, it might be possible to close your eyes or read a book and don’t pay attention to traffic. But at the same time, under all other conditions, then that specific highway scenario, you are still the one to drive the vehicle. At Level 4, vehicles can bring you from A to Be but in a limited operational design domain as they call it. And then finally, Level 5 is where that operational design domain, in theory, gets rid of all its boundaries, which I think is a fairly, potentially impossible automation level.

[10:17] Conor Peick: I think that’s a really interesting thing to think about in the future is whether that Level 5 is ever truly possible.

[10:25] Matthieu Worm: Yeah, absolutely. And as I said, I honestly believe that it won’t be, and at least it will take decades before we reach that. And up to that point, we will achieve, for sure, very high levels of automation. And making the drive from one city to another city within a certain country is definitely feasible. But to travel between countries across borders into different sets of regulations, into different types of infrastructure, that will be very tough, I think.

[10:56] Conor Peick: Currently, it sounds like most OEMs are right in the Level 2 capability range. Would you agree with that?

[11:03] Matthieu Worm: Yeah, I would agree to that. And Audi has – last year – presented a vehicle with Level 3 highway pilot capabilities. But due to regulation, it was not possible to really bring that to market. I think it was in March this year that Honda has introduced on the Japanese markets, a Level 3 system. So, that would then, I believe, be the first production vehicle with Level 3 automation.

[11:26] Conor Peick: And so then I’m curious, what do we know about what it takes to safely increase these levels of automation in vehicles?

[11:35] Nand Kochhar: As you know, when you increase the level of autonomy, as Matthieu mentioned, it’s not about just the automobile at that time, because now your car is becoming connected as it needs to be connected to the infrastructure that needs to be connected with the other cars. So, the whole host of things need to happen in order to have a safe journey. That’s an important point, which will help us move from level whether it’s from 2 to 3, and then from 3 to 4, that looking at the entire infrastructure and the ecosystem of the automobile.

[12:11] Conor Peick: Matthieu, do you have any other thoughts there on how we can safely increase these levels of automation?

[12:17] Matthieu Worm: I think one prerequisite for increasing automation levels is a new development process at the industry side. What I typically use as an example is an Adaptive Cruise Control system, which is a fairly advanced system. It maintains your speed at the highway, it maintains a certain distance to other vehicles, it reacts to new circumstances, new systems can help you change lanes automatically without any interference of yourself. These systems are verified on a proving ground with roughly 150 tests. So, if you successfully pass these 150 test scenarios on the proving ground, the system is ready to be released and can go into production. That’s a Level 2 automation, and that whole setup is built on the fact that there is always a driver as a backup solution for all other scenarios that might happen that the car is not capable of handling. If you move to Level 3, you can no longer rely on the driver taking care of all these unexpected scenarios, so that means that you testing, your verification process needs to be massively extended, either by driving endless numbers of miles and test kilometers with a large fleet of test cars or through simulation. So, that’s step one. 

[13:36] Matthieu Worm: The prerequisite for higher levels of automation is a development process that allows you to continuously test under a massive variety of conditions. The second part that needs to be in place to go to higher levels of automation is a closed-loop development process, meaning that if you have released a specific software release and an accident happens or series of accidents happen, you’ll need to have the possibility to improve the software and do [14:05 inaudible] of all the vehicles on the road to make all systems perform better. And as long as the automotive industry, in general, does not have that capability in place, and that engine, that closed-loop or as it is called continuous integration, continuous deployment development process, higher levels of automation will not become a reality.

[14:26] Conor Peick: That’s really fascinating. So, you mentioned a new development process, I was wondering if you could elaborate on what that might look like.

[14:34] Matthieu Worm: The newness is mainly in the fact that you need to be, on one hand, support continuous integration and continuous deployment, meaning Over-The-Air software updates and even hardware updates could be planned in such a context. And the second thing that I believe is essential is what we call in Siemens a Chip-to-City perspective, that there is no longer a relationship between an OEM, a Tier 1 supplier, and a Tier 2 supplier; where it is one-way traffic, where the OEM puts his requirements on the table and where the suppliers have to deliver for the lowest price possible. It is a much more integrated machine. It’s a computer on wheels, where the chip-maker and the traditional tier suppliers and the automotive OEM work on a centralized vehicle architecture that requires centralized software stack, that requires cooperation across the boundaries of all the players in that supply chain. It’s that closed-loop element, and it is the Chip-to-City elements that I believe are critical ingredients of a new development process.

[15:46] Nand Kochhar: The approach is how you do this is through what’s called Model-Based Systems Engineering. And since we touched on autonomous vehicles or any of the modern vehicles are no longer just pure mechanical or hydraulic – functionally upgraded, they’re electromechanical. What’s even more important is the software piece. All kinds of software, whether it’s embedded software or the application software, all that happening within the vehicle. So, our answer from a Model-Based Systems Engineering is with a heavy emphasis on software, and we call that a software and systems engineering approach to solve all these complex issues. As you can imagine, moving from a traditional vehicle to an autonomous vehicle, the degree of complexity has increased exponentially. And so you need a new way of design, and development, and verification, and validation. So, that’s why the systems engineering approach is absolutely critical and the key enabler

[16:50] Conor Peick: Nand, what do you see as the technical limitations to what is currently possible, you know, where are the boundaries? And that could be in vehicles themselves or in these development approaches.

[17:01] Nand Kochhar: Again, you can divvy up technical boundaries or limitations of the growth needed in several categories. We could start with the most important piece as having an EE architecture, Electrical and Electronics architecture, in the vehicles. So, as you can imagine, the traditional vehicles were starting with a certain number of ECUs to do the decision-making, and that number was growing exponentially. And now, with the autonomous vehicle, there is a new way of looking at this having a new architecture, where you could have either a single processing unit or multiple processing units by a region within the car, a section within the car so that all calls for electronic and electrical architecture. Also, what we call Software-Oriented Architecture, Matthieu touched on, autonomous vehicles are possible so that you have Over-The-Air updates, both for the development phase as well as the maintenance or ongoing basis. So, for example, you need a continuous learning for the development of autonomous vehicle. So, yourun if you have a software architecture which will allow you to pick up all the data from the sensors while vehicles are being used, what’s called the shadow mode, and then you can enhance any of those scenarios which we did not think about or the users did not think about but the vehicle ended up experiencing, so you will have a record of that and you will continue to enhance or the algorithms will continue to enhance based on their learning to. So, of course, to achieve that you have a lot of artificial intelligence and machine learning going on to deliver that. 

[18:46] Nand Kochhar: Then you look at the computing power. Now, we’re talking of massive data. So, now you need a partition, how do you process all that information in the vehicle, on the cloud? And it becomes another challenge that you can have terabytes of data all the time. So, which data is valid? How do you pick a minimum set of data and still do the same amount of learning? So, the whole thing about data analytics, computing power becomes a big thing. I think, we talked about things like Over-The-Air update, easy to say, but you need the entire mobility infrastructure in the autonomous vehicle as well, as the vehicles need to talk to the infrastructure, the lighting, traffic system, and the other vehicles, etc. So, I think all those are the areas as we grow the technologies will help us deliver and continue to expand those boundaries to make AVs a reality.

[19:45] Conor Peick: Matthieu, how do you see this all happening for any company that’s active in autonomous vehicle development? Where do they start with this closed-loop development process we’ve talked about?

[19:56] Matthieu Worm: Trends from the past, say, five years is that companies move away from having a simulation department that you call when you need a simulation. And particularly in a computational fluid dynamics domain, there is a strong heritage in having a team that you send your request to, to do thermal or aerodynamics analysis, and you get that done. And the simulation is done by the simulation specialist. Whereas, over the past five years, you see that simulation has become part of every engineer’s job, rather than that you need a simulation specialist to run a simulation. Every engineer that develops systems – whether it’s mechanical, electrical, or software – uses modeling in one way or another. That’s very important in that transition that I mentioned earlier. If you want to have higher levels of automation, you need scenario-based testing. And a big part of what Nand has introduced as Model-Based Systems Engineering is that you run continuous verification based on scenarios for the autonomous vehicle domain. That’s really important. So, having simulation embedded in the standard processes is step one. And step two is a transition towards automated testing of large numbers of scenarios. And I think the majority of the players in the automotive industry are currently working on that. So, I think that’s where we are in the timeline of that transition.

[21:28] Conor Peick: Do you see partnerships becoming an important part of this closed-loop process?

[21:33] Matthieu Worm: Yeah, absolutely. First, independent of that closed-loop process, so the amount of investments that are needed to develop full self-driving technology are even bigger than people thought them to be 10 years ago. So, the headeris between 10 and 15 players worldwide that are dedicatedly developing a Level 4 self-driving stack. These companies spend a couple of billions a year. So, massive amounts of money are needed to develop these true Level 4 stacks. And having between 10 and 15 companies means that that might still be too many. So, we can actually expect some further consolidation in the domain to develop that technology, meaning that in the end, you have maybe three players in every region to develop a full self-driving car. So, that consolidation takes place right now, took place over the past five years. You need to check the media on it, but there is every month big announcements like multi-billion dollar joint ventures or mergers or acquisitions. And below that, even more, just partnerships and cooperation contracts. The big driver for partnerships, first, I think is the complexity of the technology and the amount of money that is needed to develop a full self-driving stick. 

[22:58] Matthieu Worm: And then relating it back to that closed-loop development process, if the car becomes a computer on wheels, and you want to treat it as a computer on wheels, you need cooperation across the borders of the individual players in the supply chain, and you need to build a much more integrated product. And we know Tesla has developed their own full self-driving chip, as they call it. Where the full self-driving chip, by the way, still runs a Level 2 ADAS (Advanced Driver-Assistance System). So, it’s definitely not a full self-driving vehicle yet, that’s really important to realize. VW also announced that they started designing their own chip. So, they also consider the core chip to be the heart of the car of the future and no longer the powertrain or the chassis system, as we’ve seen in the past 50 years. Whereas Volvo, Mercedes, has put their eggs in the basket of Nvidia to deliver that core compute platform and the capability to run these Over-The-Air software updates, that a CI/CD development process that I just mentioned. So, I think the industry is really searching for how to implement that closed-loop engineering process and interaction with the end customers and the end-users using these vehicles. And they truly rely on partnerships there. 

[24:24] Conor Peick: Nand, how do you see these partnerships evolving? Are they strategic exclusive partnerships with a specific aim to co-develop throughout the upcoming number of years?

[24:35] Nand Kochhar: Yeah, I think I would say most of the partnerships I see are strategic when the announcements come out. Just the latest stuff this week, Magna buying Veoneer, one of the safety ones just last couple of days. So, as Matthieu said, every day you hear new partnerships. And they are rarely strategic in nature because what’s happening is the boundaries are blurring between technology companies and so-called traditional OEMs. They’re also blurring between Tier 1, Tier 2 suppliers to the car manufacturers. At the same time, you also have startups. Startups are writing their own rules because they have no legacy information or legacy processes to worry about, and they don’t have to worry about the current production. So, you can see is a whole host of things going on. Here’s a very traditional 100-plus-year-old product development manufacturing organizations. Their problems, what they need to solve, are a little bit different than a startup than a Tier 1 is trying to grow, become totally integrated. And that’s why I think some of the things, partnerships are absolutely critical. Because in my mind, no single person, doesn’t matter how large of an organization it is, is not able to solve this level of complex problems on their own. 

[26:03] Nand Kochhar: So, I think they all are extending hands very strategically; what they think their gap is; what they can do on their own; and what they need to partner with someone and grow strategically to deliver the end objective of delivering the autonomous vehicle for the end consumer or developing a new business model. It might not even be a regular consumer, the way we have consumer today, it could be all in fleet operator, and that gives a new business model. That’s how I see partnerships maturing, and continuing to mature as we go forward.

[26:39] Matthieu Worm: And by the way, I think there is an analogy with our own company. So, Siemens has been very strategically acquiring companies, building this new digital industry software company. And I think Siemens is the only one right now that can digitize from — If I focus on the automotive domain, we can digitize the chip, we can digitize sensors, environmental models, controllers, E and ER architectures, individually is used. We can digitize then at the level of the vehicle, the chassis systems, the vehicle dynamics behavior, the power train behavior. We can put that vehicle in a digital environment. And not only do we have the digital thread, but we also add to that the test components, so the physical testing and data acquisition systems that are needed to collect the sensor data and to really close that loop that we refer to before. I think it’s tough to claim that Siemens could be an indicator of the global trends, but maybe it is. Siemens is definitely one of those. We’re number 11, I think, in most innovative companies in the world last month in the BCG analysis. It’s not completely out of whack to look at ourselves and see ourselves as an indicator of the strength where mechatronic system development can no longer be handled by an EDA company, a CAD company, and a simulation company. What you need is an integrated set of tools and development processes to tackle the challenges of today.

[28:13] Conor Peick: Do you think these integrated tools as you were just talking about, do you think that can help support these partnerships we’ve been talking about?

[28:20] Matthieu Worm: Yeah, I think it’s essential. And I can refer to a customer discussion that we had over the past six months, where there is a new partnership announced between a traditional Tier 1 and a traditional chipmaker. But the chipmaker has established development processes, verification processes for virtualization of chips, coverage-based testing approaches to ensure that you fully test all the almost an infinite number of combinations of transistors that you can think of in that chip. And on the other end, the automotive industry that has a verification process going through modeling in the loop, hardware in the loop, software in the loop, vehicle in the loop, and proving ground testing, and full vehicle testing on the road, the ISO 26262 Functional Safety Standards, the Automotive SPICE development processes. 

[29:15] Matthieu Worm: So, it’s a very different way of looking at verification and validation. It’s a very different way of translating requirements into test cases into conclusions. And certainly, these two worlds need to work together and they need to build an integrated product. And the nice thing of them being Siemens is that we can bring two people into the call that understand both the chip perspective as well as the vehicle development and system development perspective. And that we start to merge both worlds. And that we, for example, now can couple a tip emulator with a virtual vehicle in a virtual environment. And that we can develop load cases for that full self-driving chip, as Tesla calls it, or the next-generation chip that VW will develop an autonomous-vehicle-specific chip design that needs to handle autonomous-vehicle-specific use cases, and mean that we can help the chip industry with an environment to test the chip in the right context. Whereas we can then help the automotive company reusing the same test suite and the same test cases to test the full vehicle performance afterward. So, it brings a lot of continuity throughout the supply chain, and a lot of knowledge-sharing and understanding of the different perspectives of the different players in these partnerships.

[30:38] Conor Peick: One of the other things that that we need to address is that there are currently pretty limited regulations in place surrounding autonomous vehicles. So, Nand, how do you see that evolving in future years? And how can the automotive industry play a role in the development of these new regulations?

[30:54] Nand Kochhar: Again, this is new technologies, new areas. You always start from a regulatory perspective, that’s where the word “limited regulations” exists. It’s a Catch 22. Industry works with the regulators, for example, in North America, National Highway Traffic Safety Administration, or in Europe, you could work with the agencies like Euro NCAP, which gives the safety ratings of today’s vehicle for passive safety. So, one of the things these regulatory authorities have been doing is already starting to give credit for a level of autonomy in their cars or the features which are being offered in the cars to count towards that star rating system. So, for example, the autonomous emergency braking, if your vehicles have that, it gives a certain amount of credit or contributions towards the overall rating calculations. So, I think in that way of looking at things, the technical organizations like Society of Automotive Engineers, as a body across different OEMs or the automotive industry, whether it’s a Tier 1 supplier or the OEMs themselves, have an opportunity to work with the agencies, show them the way what they’ve been developing, testing, what their views are, and getting the authorities to buy into that and have that iterative process. 

[32:29] Nand Kochhar: That’s how, in my experience, the past, let’s say the safety regulations in every country, that’s how they get developed; that you work hand in hand with the regulatory authorities. And then what’s the art of possible? Where do we need to take the technology for the sake of larger good, for example, safety in general? What do we need to do? And that will continue to evolve. So, industry plays an important role. And those regulations will continue to mature and get to a point that it is just like today’s passive safety regulations we have on the automotive industry.

[33:08] Conor Peick: Matthieu, you already talked a little bit about the verification validation of autonomous vehicles. Could you comment a little bit on how these systems are going to be validated against these new regulations that are going to be developed?

[33:21] Matthieu Worm: The first step, and that’s what you also see in the new protocols that are being developed, for example, Euranka is a combination between virtual and physical testing. So, where today, all testing in these consumer tests is physical testings, very detailed protocols that describe how to execute a test, with which steering input, with which objects that you drive towards. So, a very tight set of boundary conditions to ensure that these tests are repeatable, and to ensure that you can compare the performance of one vehicle to another. However, referring back to what we discussed earlier. With increasing levels of automation, the big challenge is not in reacting safely in a very specific and well-described scenario. The big challenge is in the large variety of scenarios that the vehicle needs to react to. What Eurankaenvisions is that they describe a very large scope of possible conditions that they ask the automotive industry to supply virtual results, covering all those individual possible scenarios and that they choose then 20-30 specific cases out of these hundreds of cases to test on the proving ground and to repeat on the proving ground. 

[34:42] Nand Kochhar: So, what the regulators will look for is ways to extend the scope of possible testing conditions, while at the same time ensuring comparable, repeatable test procedures, and that’s a massive challenge. And I think even a bigger channel is the fact that these Over-The-Year software updates have an impact on the safety performance of vehicles. And there is a known Tesla update of the battery management system that impacted the braking distance of the vehicle. And the battery capacity is a contributor in these braking systems. 

[35:18] Nand Kochhar: So, if you optimize for full battery charging, your braking capacity will be less. So, there has been an adaptation in the battery management system to improve the braking distance of these vehicles. And that’s such a fundamental safety performance indicator that you can question if a vehicle that got a one-time type approval as we do today and is allowed on the road, but gets software updates that in that way can impact the safety performance. But you also need continuous regulation or a continuous homologation process. And the regulators definitely work on that. There’s also discussions that we have with these institutions on how to handle that. And my bet is that also their virtual validation or virtual verification then will be a key element in improving that your new software release is as safe as the release before was.

[36:14] Conor Peick: Taking everything we’ve talked about into consideration, what do you see, at least initially, is the most plausible scenarios for deploying autonomous vehicles at scale?

[36:24] Nand Kochhar: I think some of that is already happening today. And now the question is, how do we grow the scale. So, in terms of technically speaking, we’ve been saying, SAE levels of autonomy, and we said, “Yeah, the next level is an SAE level for autonomy.” And the examples of that would be robo taxis or the shared vehicles. And those are running into certain conditions that are called geofencing. So, in other words, you have an operational domain, you put some kind of constraint. And under those conditions, those records will be running. And it also depends on the other elements. We touched on the policy, and the legal regulations, and the openness of the legal system in a country. 

[37:14] Nand Kochhar: So, the answer varies from region to region and country to country as well. In my mind, a short term plausible way is growing the SAE level for vehicles, which are already running as pilots in several countries, including, let’s say, in North America, and I think it’s Phoenix area, Google running those vehicles where people sign up, and people who are willing, they share the rides and they get picked up in those. In countries like Asia Pacific, several countries, they’ve got the robo taxis; you can order a taxi under certain conditions that the geofence could be a certain city or within a city, a block. A good example, right in our backyard in Detroit, Ann Arbor area, I think Domino’s Pizza had run pilots of delivering pizza. And that was the SAE level four, but it was geofenced from certain locations to certain regions. So, I think the maturity of that as the technologies we talked on, that will continue to grow. So you will see an adaptation moving from Level 3 three to Level 4 vehicle of autonomy all over the globe over a shorter period of time. That would be the most plausible scenario.

[38:39] Conor Peick: Matthieu, do you have any further thoughts on how these vehicles will be deployed and then scaled?

[38:44] Matthieu Worm: So, I very much share the perspective that Nand just put on the table. What I hope is that the big value of self-driving vehicles is not in the self-driving part. The true value comes if we can improve our cities. And if you look at the city today, no matter where you are, you will see cars everywhere. Streets are packed with cars. There are specific parking full of cars. There is traffic jams. There’s people spending useless time in cars. There is research from Bosch that says that in urban environments, 30% of the driving time is spent on looking for parking places. It’s that kind of nonsense that we have, by now, accepted as a reality and that we have accepted as being normal, which is not normal. And autonomous vehicles can really help at solving that issue. And if we can take away vehicles from the road by reducing the number of vehicles and share them with more people – the typical car is standing still for 95% of its time so, if we can just increase the usage of vehicles, if we can put vehicles where we don’t see them, if they can automatically park themselves in an area in a garage where we don’t have them in front of our houses all the day, that would be excellent. 

[40:08] Matthieu Worm: And lastly, if we can share cars with multiple people, we will also have less kilometers driven in total. And that’s I think, where the true value of self-driving vehicles is. I hope that the deployment of self-driving vehicles is not something that’s just about increasing levels of automation in privately owned cars. But that it actually is the ignition of a change towards mobility as a service, where we really start to travel from A to B, rather than that we look for our next new car or status symbol.

[40:41] Conor Peick: If I could reduce the amount of time I have to spend looking for just parking, I think that would be a huge benefit to my life. 

[40:50] Matthieu Worm: A massive win, right? 

[40:51] Conor Peick: Yeah. Not to mention everything else that could come along with it. Guys, thank you so much for joining us today. I think this has been a really fascinating discussion. Nand, Matthieu, thank, you guys. And we will hopefully talk to you guys again soon. 

[41:04] Matthieu Worm: Many thanks. 

[41:05] Nand Kochhar: Thank you, Conor. 

[41:06] Conor Peick: AV development programs are regaining momentum across the automotive industry. Progress is being made in many aspects: self-driving technology, public perception, and regulations or legislation to manage the deployment of AVs. While challenges remain, the industry can count on advanced closed-loop development methodologies to overcome any roadblocks along the way. I’d like to thank both of our experts for offering their insight today. And I do hope that you’ll join us again on The Future Car. Bye for now.

Leave a Reply

This article first appeared on the Siemens Digital Industries Software blog at