Software-defined vehicle adoption in India growing fast

Senior Architect, Transportation Business Unit, Tata Elxsi discusses with Shahkar Abidi how the adoption of this technology is taking place in the country and its likely impact on the auto ecosystem.

• Software-defined vehicles are more connected, with features like telematics, vehicle-to-everything (V2X) communication, and connectivity to the cloud. This connectivity enables real-time data exchange, remote monitoring, and enhanced safety features.
• The development of autonomous vehicles relies heavily on sophisticated software algorithms and sensors. Software-defined vehicles play a crucial role in the advancement of autonomous driving capabilities.
• 'There's no single-fit solution for all OEMs. We have to study the use cases and we need to understand how you migrate to the new architecture, and that itself is a consultation service'
• Tata Elxsi is a global design and technology services company that provides product engineering and design services to various industries, including automotive. The company works on solutions for connected cars.
• A lot of the SDV depends on cloud storage, India's policy roadmap is still not very clear, though there is some unanimity with regard to saving the data within India.

Large deals and strong traction in Software Defined Vehicle (SDV) engagements have helped Tata Elxsi, one of the top providers of design and technology services to the automotive and other industries, experience an impressive surge in its transportation business. Sreeja KS, Sr. Architect, Transportation Business Unit, Tata Elxsi talks about the Indian OEM's transition towards SDVs and how its adoption in the country is growing at a greater pace than globally. Tata Elxsi is a global design and technology services company that provides product engineering and design services to various industries, including automotive. They work on technologies such as connected cars, autonomous vehicles, and human-machine interfaces.

SDV is a newest technology application that is coming up, but it's not that it wasn't there earlier. It was like the industry was concentrating much on decoupling the hardware from the software. So the issue was how to migrate and upgrade the software where hardware was the limiting factor. Then came the concept that everything is software-centric, but that doesn't mean that the hardware doesn't play a role there. So now the hardware has to be selected so that the software upgrade can take place in the future.

"In a software-defined vehicle, if you look into the multiple drivers that actually resulted in this software- defined vehicle, as we all know, we have electrification, autonomous driving, powertrain systems, and connected cars. So all these are the main drivers for the software- defined vehicle," says Sreeja.

To make the system a success, high computational powers are needed for making technologies like electrification, non-autonomous driving, and connectivity work seamlessly. Alternatives include cloud computing.

These days customer want more and more experience and wants to use it like an app or a mobile device. There's a limitation with the current hardware in the vehicle that may not be able to support the set-up, so carmakers will go for a high-computing system, and then the most important thing is connectivity to the cloud. The cloud concept was already there, but then that was not connected to the vehicle that runs now, reckons Sreeja.

Benefits of cloud computing

OEMs have realised that it's really important to harness the ability that is there in the cloud as it actually can allow a lot of computation work to be done. However, in the cloud set-up, there are some restrictions on the hardware. So the OEMs are now going towards the monitoring part and the analytics part of the data.

This is to help enable increase the customer experiences. "You want to have feedback on the data that you are actually getting out of your car. So you have a lot of sensors connected to that; you are getting a lot of data, but then how do you use it as feedback to improve the customer experience?" asks Sreeja.

Till now, most OEMs have been running through the legacy architecture; they have the distributed environment, the distributed architecture of ECUs, and the domain-centralised ECUs. Now it's not possible for them to quickly or suddenly shift to a new architecture and also reduce dependancy on ECUs. So OEMs now have a phased approach to enabling that software-defined architecture, wherein they are actually looking for collaboration with specialists for solutions.

Collaborative efforts

Electric cars, autonomous cars and technologies are becoming important aspects for OEMs now to consider. "So as you know that everything is getting electric, you want better battery health, and you want to predict how far your vehicle will go. So how much of that is possible? So the prognostic part is everything. For that, the SDV is one of the main concepts through which all of this could be achieved, says Sreeja.

Reducing the number of ECUs

OEMs globally are looking for is to reduce the number of ECUs by at least 10 percent of their existing number of ECUs. Also, wiring weight is one of the main things that they want to actually reduce to, say, 20-25 percent, and for this, they have to actually go through or adapt to the high- performance computer-based architecture, or the E2E architecture.

It's not just limited to the global OEMs; even the Indian OEMs have the same aspirations. They want to reduce the number of ECUs; they want to achieve a lower reduction in the weight of the wiring. The SDV market is growing at a rate of 20 percent globally.

Influence of EVs

The EV adoption in Indian market and its growth is even greater than that of the global market. "So, as we all know, how important are electric vehicles now running in India So electrification has actually touched the Indian marketvery significantly, and that is one of the major drivers of SDV as well, says Sreeja.

Predictive analysis in EVs

Accurate range predictions and the the health of the battery are two major considerations for EVs. "We are actually collaborating with some of the OEMs on that. I may not be able to disclose that, but then within, let's say, three to five years from now, the Indian cars will also be adapting to the new technology trends of, let's say, the computer platforms and the AI-ML models that we use for the digital twin for better battery health monitoring," says Sreeja.

She estimates that it might take three to five years for the Indian OEMs to bring that, but then the concept is already in place like the connectivity to the cloud and the use of 5G, and predictive analysis.

Where are the opportunities?

Sreeja reckons that there are a lot of opportunities or potential within the Indian market where Tata Elxsi can offers customers with consultation services. For example,an OEM migrating to the new architecture may look for a befitting SDV solution. There's no single-fit solution for all OEMs, she adds. So the company has to study the use cases individually and then suggest the migration route to the new architecture.

OEMs generally engage specialist service companies, Tier-1s, and SOC vendors to analyse and find what could be the optimal design strategy that they should adopt for moving towards the software-defined vehicle. Besides giving consultation services on migrating to this new architecture, Tata Elxsi also offers advise on the cloud platform, integrating cybersecurity among others.

Different verticals

Tata Elxsi has different verticals for autonomous driving; and for electric vehicles separate units like AUTOSAR, Classic AUTOSAR, and Adaptive AUTOSAR. The connected vehicle platform (CVP) has been developed in-house at Tata Elxsi.

"So we have all the different components available with us, and that was there even earlier, but then once we started looking towards a software-defined architecture, we understood that it's not something new. We don't need to reinvent the wheel; we don't need to develop everything from scratch," says Sreeja. At Tata Elxsi, the work isn't just confined to automotive sector. The company has an internal roadmap wherein prototyping the connected vehicle platform with AI-ML technologies is possible.

SDV tech in commercial vehicles

Almost 80 percent of SDV business comes from passenger vehicles. However, not so much in commercial vehicles. "You don't need those many abilities that you want in a passenger vehicle. So that could be one of the reasons why the complete software-defined vehicle part is not yet coming in the commercial vehicle part," says Sreeja.

"If you look into the prognostics or, for example, the safety part of the vehicle, So, if you compare it, the commercial vehicles, for example, have the J1939 new protocol over CAN. It's not the same CAN, but it's another protocol, and then bringing safety to that is not yet considered because you don't need so many features in the commercial vehicles as you look for them. Because the commercial vehicles are not running, the drivers don't need the same kind of experience that a normal passenger vehicle driver needs," she says.

So, that is the reason that as of now, the SDV part is not yet fully into the commercial vehicles, but maybe the cloud connectivity part to do the prognostics of the data is there, but then it will take maybe a few more years for these software-defined vehicles to be brought into the commercial vehicles, she reckons

Cloud policy not so clear

Since a lot of the SDV depends on cloud storage, India's policy roadmap is still not very clear, though there is some unanimity with regard to saving the data within the confines of the boundary in India. Sreeja explains that here are actually two aspects: the safety part and the non-safety part. "In our vehicles, if you look into the number of sensors that are connected, there may be some hundreds of sensors that are connected to the vehicle, and they generate a lot of huge data. If you look into, let's say, 2-4 years from now, there will be almost 10 million terabytes of data that will be generated in a month. It's a huge amount of data, and that's almost a thousand times the amount of data we have now," she says.

Bandwidth and wi-fi

The huge size of the data and the bandwidth capability for connectivity, it's possible that it will reach a bottleneck. So, for example, if a vehicle is running, it's not like a normal mobile app. "If you have Wi-Fi and the connectivity gets lost, it’s okay. Maybe the app download doesn't take place, but that's fine; it doesn't have any critical implications.

But if you look into the car domain, if the connectivity gets lost, then your safety-critical applications won't run, and that's really critical, and you cannot afford that," she explains. So what do the system designers do with the most suitable design? "But then that depends on the use cases. So is it a safety-critical application? Maybe you don't run it on the cloud; you can run it on the vehicle itself, and then the most important thing is, as I mentioned, the connectivity part—the huge data that you're looking for," she says and suggests that it's not very wise to push all the data to the cloud.

Avoid putting all applications in one basket

The company begins any consultation job with the first level of analysis on the vehicle itself to determine whether the application is really critical, then do it on the vehicle itself. Sreeja is of the opinion that it is not advisable to push everything to the cloud, but then there are exceptions like entertainment features that are not safety-critical. So even if cloud computing sis used, that's fine, but there has to be a possibility of checking the data that you receive back. For example, there has to be a handshake between the control centre and the vehicle. For example, whatever data you are getting, are you getting the right data? Are you getting it within this much time? Otherwise, the latency will increase, and time-critical things won't work, she asks.

So basically, she says that if you look into the cloud part, it's always wise not to push everything to the cloud, and the OEMs are also thinking in that way. They think of the design, they think of the performance that they need, which critical applications they want to run, what is the latency, and how much time they have for this specific application to run. Is it really possible to run it on the cloud? Let's say the safety application is running on a cloud, and if your connectivity is lost, you don't want your car to be going in a haphazard direction. So you then use the other cloud as well. So that could be one of the possibilities that the customers might be looking into in the near future.

ADAS applications

Infotainment or the ADAS applications are related to the customer experience, could be updated multiple times; that's okay, but if you're looking for an airbag or a braking system, you do not go for a frequent update. So you don't need cloud connectivity for such features. So it depends on the use case—what kind of applications you are running on your ECU or your complete vehicle compute platform—that you decide on the cloud side or what kind of data you want to process exactly.

Addressing software failures

There have been cases of GPS malfunctioning; one happened in Kerala itself in India, wherein three doctors were driving back home when suddenly the GPS malfunctioned and their car ran into the river instead of the bridge. The question that arises is that how can a design architect for vehicles address the issue? "Even though we might have come across such situations where you have your GPS and then you just come to a dead end, you are not able to proceed further. So maybe one of the possibilities would be to use the cloud to analyse the environmental conditions at that point in time and see how critical they are," she explains.

For example, if it's a rainy season, maybe the GPS signal or the data that you use may not always be appropriate. So use a redundant way to analyse and do a plausibility check on the data that you have received. Maybe you have two service providers, and then make sure that the data that you received is plausible. You have to have redundant checks for that data and a different path to make sure that the data is correct or not. For that, you may even use the AI-ML models and do the prognostics to understand the environmental conditions to have and also to see whether there have been multiple such occasions earlier.

Author: Sreeja KS - Sr Architect, Transportation Business Unit, Tata Elxsi