“Where traffic rules are not followed, you can't train an AI model to figure out what to actually do”

Snapdragon will be a familiar name if you are an Android phone user, but lesser known is their deep involvement with the automotive industry. Back in 2022, Maruti Suzuki began incorporating Snapdragon chipsets for the connected car features on the Baleno facelift, and now the upcoming eVitara will use the Snapdragon Digital Chassis. The same is also used on the Tata Harrier EV. And Mahindra’s association goes even further back, with the 6150 chipset integrating the twin screens, cameras and sensors enabling ADAS on the XUV 700. And that was so successful that Mahindra upgraded to the 8295 chipset for the BE 6 and XEV 9e which has a whole bunch of features that can progressively be unlocked as infrastructure improves. On the 2-wheeler front everyone from Hero to Bajaj, TVS, Royal Enfield, Ather and Ultraviolette, are working with Snapdragon on making their vehicles safer and more connected. But can ADAS reduce accidents in India? We caught up with Nakul Duggal, group general manager, automotive at Qualcomm Technologies during the Snapdragon Auto Day to understand their involvement with the Indian automotive industry.

On ADAS and infrastructure

“One of the challenges with adding safety technology into a vehicle is that you have to be able to have enough sensors that can deal with the environment within which that platform is implemented. Part of the challenge is the limited infrastructure or the non-uniform availability of infrastructure. Where it's not available, where traffic rules are not followed, then you can't really train an AI model to figure out what to do. At that point in time, the system stops working.”

Developing driver assistance features for a specific market

“The approach we have taken is to develop for the globe then to localise for the markets where we feel we have to do something different. We have the good fortune of having a large team in India. We obviously work very closely with many different customers."

“The offer that we make to our OEM partners is tell us what you would like to do for your specific needs. We'll figure out a way to do that. It's an overlap of infrastructure availability, affordability, the evolution of new infrastructure, and new capabilities becoming available. As that happens, we can offer more and more capability. One thing that is quite unique about ADAS is that it depends upon quality of data collected. What we are starting to do is to do a lot of local data collection and a lot of local data processing. We'll be able to create a corpus of information, of models, of training, that then becomes very beneficial to this local market. That creates a couple of different benefits. The same data can be used for informational safety.

“If I detect a hazard in front of me, and I know that I am a new driver, I am a teenager, I just learned how to drive, I would like my car to be able to share with me more information about hazards as opposed to ignore them. If I'm a seasoned driver and I see a cow on the road, that's fine. If I see a child on the road, and I'm a brand-new driver, it's probably something that’ll warrant a different sort of input from the car as well as the driver."

“This is a very simple way for technology to be able to provide more alerts, but it requires you to be able to go to the next level of detail. It's not possible to do this if you just take some global product and apply it locally. You actually have to do something on the ground level.”

Safety in the context of situational awareness

“Something we have started to think about is in the event of a crash, how do you alert the authorities, the automaker, your emergency contact, about what is actually going on? It's like the black box feature. I want to be able to get a snapshot of what is actually going on so I can take some action. It’s not complicated to do. The technical barrier is actually not all that high. But it has to be something where the automaker has to say, this is something I would like to figure out how to implement. If they [manufacturers] can implement features that improve the quality of safety through technology, that has a chance of becoming more broadly adopted.”

On V2X and how it can be implemented in India

“V2X has turned out to be more complex than we initially thought it would. Everybody [now] understands what frequency is needed, what technologies work. The second thing is, how do you deploy it? Vehicle to vehicle means you have to have enough vehicles that have implemented the tech. But if you do infrastructure to vehicle, I feel like that this network effect benefit is very powerful. If you think about smart infrastructure, where the infrastructure has cameras, the infrastructure is able to look at the status of the highway at any point in time and broadcast that information. And every vehicle has the ability to receive it and then act on it. That reduces congestion, reduces collisions and even fatalities. This is an area that we want to spend a lot of time on with the government to figure out, as new infrastructure gets built.

“From an investment perspective, this is a very simple technology. We wouldn't even have to think too much about what investment is needed to make it happen. But it has to be something where you have to have the vision that this is something that has to be implemented. It has to be managed. There has to be a management aspect, because somebody has to administer this.”

Cost sensitivity of the Indian market

“We are not interested in the people that say no. We are interested in people that say yes because they will then get other people to make those changes. We drove this integration strategy very heavily and today the level of integration is unbelievable. We are talking about integrating ADAS and cockpit onto the same platform, which we will launch by middle of next year. The rate at which the auto industry has evolved and the need for competing, the need for doing better than the other guy, the ability for people to experiment, that has driven the integration mandate."

“For India we are able to bring a lot of value when we bring integration into these vehicles because the message that we bring is, not only are we integrating, but we are providing you a platform that has a lot of headroom to scale. You don't have to keep changing your platform, the same platform will last you a long time."

“The other thing that we introduced was this feature called a soft SKU. This allows you to only use the amount of capability in the chip that you need and you only pay for that. Let's say I can give you 40 per cent of the capability, you're only paying for 40 per cent. If you want to get to 60 per cent, you pay me when you need it, otherwise you don't pay for it.

“We have the ability to unlock more capability after the vehicle is deployed. We've done a lot of different things that allow us to be able to be very sensitive to what customers' needs are. Every automaker is cost-sensitive. We tend to think that India is a price-sensitive market but every automaker is a price-sensitive automaker. There is no automaker that says that this is not an issue. The question really becomes, what levels of flexibility are available?

On ADAS with two-wheelers

“It's mostly informational. One feature that we learned about in India is that a lot of accidents happen when somebody opens the door of a car suddenly and it's too late for the two-wheeler to recognise it and that creates a lot of accidents."

“It's basically a camera in the two-wheeler that literally detects that somebody has opened a door and it sends an alert, an audible alert if you're wearing your headphones or an alert on your dash that will tell you here's an obstacle ahead or even a haptic alert to the handlebars. The technology is not all that complex."

“You have to think about a car that suddenly showed up. Here's a pothole that you don't see or the water on the road, or here is an accident. All of these things, if you make the rider situationally aware, you are providing them information and this is where a lot of people are thinking about ADAS in the context of infrastructure."

“The rider has to be engaged but now we’re providing the rider with information such that they are more alert to their surroundings. That's really the gradient."

“Royal Enfield has selected us as a global partner now. Every Royal Enfield platform, starting 2026, is going to use our infotainment systems, our cluster systems, our connectivity and then our bike to cloud platform.”

Adapting ADAS to Indian traffic

“The false positives are a data problem. The more training that you can do, the smarter the model can become. In all of these problem statements, all it really comes down to is what is it that the system is trained to do and if that training is completely different from the environment that it is exposed to, then users are not going to get a good result. One of the reasons why we believe that we must build models just for the Indian market is because if you train for the local market, then the results will be based upon what it is trained for."

"You're always going to have misses. For example, if somebody's driving on the wrong side of the road, you can’t train for that. If there is somebody who's breaking the rules how do you plan for something like that? With these types of things, it just comes down to where is it making sense to use the system, and where does it not make sense."

“We are trying to collect data to learn. Whatever learning is there, we feed it into a central database such that we improve those learning algorithms and everybody benefits from it.”