Indian Roads Need India-Specific ADAS Data Pipelines

Every road tells a story. In India, that story often involves a school bus making an unannounced U-turn, three motorcycles racing each other in a lane built for one, and a fog bank appearing out of nowhere. Any technology that claims to assist a driver on Indian roads has to first accept this reality – not simulate it, not approximate it, but actually learn from it.

Advanced Driver Assistance Systems, or ADAS, are becoming increasingly visible on Indian roads. According to JATO Dynamics, ADAS adoption in India reached 8.3% in the first half of 2025, with more advanced systems growing at over 70% year-on-year.¹ That is a meaningful shift. But adoption is only part of the story. The bigger question is simpler: has this technology actually learned from Indian roads, or is it still relying on foreign models?

The Indian Road Is Its Own Scenario Library

ADAS systems learn by using cameras and sensors on a test vehicle to capture what happens on the road, and that footage is used to teach the system how to respond. How well it responds later depends entirely on the quality of this data. Most of the footage used to train these systems around the world comes from roads with clear lanes, orderly traffic, and predictable behaviour. That is not urban India, and it is certainly not rural India.

Traffic on Indian roads does not really follow lanes. Two-wheelers are the dominant vehicle on the road and move unpredictably. Pedestrians, cyclists, auto-rickshaws, cattle, and construction debris all share the same stretch of road with no real separation. The data confirms the scale of this complexity: two-wheelers accounted for 48.3% of all road fatalities in India in 2024, and pedestrians for a further 14.7%, according to the National Crime Records Bureau.² A system that struggles to anticipate these road users simply is not ready for Indian conditions, no matter how well it performs elsewhere.

This is where standard global test scenarios fall short. A typical global test scenario might show a pedestrian crossing at a marked crossing, in good light, with a clean background. What an Indian validation engineer actually has to prepare for looks nothing like that: a child running out from between parked vehicles, a group of workers crossing a highway at dusk, a wedding procession spilling onto the road. In India, these are not rare edge cases, they happen every day. 

Add the mix of weather, from monsoon downpours to blinding summer glare, and it becomes clear that no off-the-shelf dataset can capture what Indian roads actually look like. The only real way to build ADAS systems that work here is to keep collecting and learning from actual Indian driving, continuously.

The Quiet Layer Behind Every ADAS Decision

Most conversations about ADAS focus on sensors and processing power. The less glamorous part, the systems that capture and hold on to all that data, gets far less attention, even though it matters just as much.

A modern vehicle generates a huge amount of data every second, but most of it doesn't need to be retained. What matters is capturing the right moments: sudden braking, near-misses, situations where the system got confused. These are the moments that actually teach the system something. If they are lost before they can be reviewed, that opportunity to improve is gone too.

Why Storage Deserves More Attention

If data is what makes ADAS smarter, then reliable storage is what makes sure that data survives long enough to be useful. It is one of the least discussed parts of the whole system.

Inside a vehicle, conditions are tough: heat, constant vibration, and the need to record critical moments without missing a beat. This calls for storage built specifically for automotive use, not the kind found in everyday consumer devices.

There is also the challenge of getting all this data off the vehicle and into a system where engineers can actually use it. With fleets generating this much data, how quickly and reliably that transfer happens becomes a real engineering concern, not a minor detail.

Of course, storage alone doesn't make a car safer. But what is does, is that it makes sure that when something goes wrong, the moment is reliaby captured, saved, and available for engineers to learn from. That is a distinction worth remembering.

Bringing It Back to the Road

Once the data is off the vehicle, it needs a place where engineers can look back at specific driving moments, rebuild them, and test whether a fix actually works. For India, that library of moments needs to be built from Indian roads and Indian traffic. Otherwise, the testing happening in the background stays disconnected from what actually happens on the ground, and that gap eventually shows up on the road.

The road ahead for ADAS in India will not be decided only by sensors or software. It will be decided by how well the industry builds the quieter systems that connect what happens on the road to what gets fixed in the model, and that is where the real work still lies.

Prasad Sandireddy is the Vice President of Product Development Engineering at SanDisk. All views expressed are the author's own.