Browse
Log in / Register
Share
Every stretch of Indian road is a living library of chaos: a school bus cuts across three lanes without warning, three motorcycles squeeze into a space meant for one, and a sudden fog bank swallows the highway. Technology that claims to assist a driver here must first learn from this reality—not simulate it, not approximate it, but ingest it, mile by mile.
Advanced Driver Assistance Systems (ADAS) are gaining traction in India. JATO Dynamics reports that ADAS adoption hit 8.3% in the first half of 2025, with more advanced systems growing over 70% year‑on‑year.¹ That jump signals real momentum. Yet beneath the adoption numbers lies a critical question: are these systems actually trained on Indian roads, or are they still relying on sanitized foreign datasets?
Why India’s Roads Are a Unique Scenario Library
ADAS systems learn by capturing real‑world footage through cameras and sensors on test vehicles. The quality of that footage determines how well the system later responds to hazards. Most global training data comes from roads with clear lane markings, orderly traffic, and predictable behavior. That description doesn’t apply to urban or rural India.
India’s traffic is fundamentally different. Two‑wheelers are the dominant vehicle type—they accounted for 48.3% of all road fatalities in 2024, according to the National Crime Records Bureau.² Pedestrians added another 14.7%. These road users weave unpredictably, often intermingling with auto‑rickshaws, cyclists, cattle, and even stray debris. A system trained only on European or American highway footage will fail to anticipate a child darting between parked cars or a group of workers crossing a dark highway at dusk.
Standard global test scenarios are no help. A typical Euro NCAP scenario might show a pedestrian crossing at a marked crosswalk in good light. An Indian validation engineer, by contrast, must prepare for a wedding procession spilling onto the road, a cow standing in the middle of a lane, or a sudden two‑wheeler swerving from the shoulder. In India, these are not rare edge cases—they are daily occurrences.
Add the climate: monsoon downpours reduce visibility to near zero, while summer glare bounces off dust‑covered windshields. No off‑the‑shelf global dataset can capture this complexity. The only real path to effective ADAS in India is continuous, ground‑up data collection from Indian roads.
The Quiet Layer Behind Every ADAS Decision: Data Storage
Most conversations about ADAS focus on sensors and processing power. The less glamorous but equally vital layer—how data is captured, stored, and transferred—rarely gets the attention it deserves.
A modern vehicle generates massive data every second, but not all of it is valuable. What matters is capturing the moments that teach the system: sudden braking, near‑misses, system confusion. If those moments are lost before engineers can review them, the opportunity to improve is lost too.
Reliable storage becomes a silent enabler. Inside a vehicle, conditions are harsh: heat, vibration, and the need to record critical events without skipping a beat. That demands automotive‑grade storage, not consumer‑grade flash memory. Equally important is moving that data off the vehicle quickly and reliably—a nontrivial engineering challenge when fleets generate terabytes per day.
Storage alone doesn’t make a car safer, but it ensures that when something goes wrong, the moment is captured, saved, and available for engineers to learn from. That distinction is worth remembering.
Building a Local Data Pipeline
Once data leaves the vehicle, it needs a platform where engineers can reconstruct specific driving moments, test fixes, and validate improvements. For India, that platform must be built entirely on Indian traffic patterns and road conditions. Otherwise, the testing happening in labs stays disconnected from reality, and that gap eventually shows up as missed detections or false alerts on the road.
Several global ADAS companies, including Mobileye and NVIDIA, have begun collecting Indian driving data through local field tests. But the scale is insufficient. A few hundred hours of highway footage cannot capture the diversity of India’s 6.4 million km road network, which ranges from multi‑lane expressways to single‑track village roads. What India needs is a dedicated pipeline—similar to what China built with Baidu’s Apollo platform—that continuously ingests data from thousands of vehicles across the country.
What’s Next for ADAS in India?
The road ahead for ADAS in India will not be decided solely by better sensors or faster chips. 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. That means investing in local data collection fleets, automotive‑grade storage designed for high‑temperature environments, and data‑management platforms that let engineers replay and learn from real Indian crashes and near‑misses.
The technology exists. What’s missing is the willingness to treat Indian roads as a first‑class use case—not an afterthought.
Prasad Sandireddy is Vice President of Product Development Engineering at SanDisk. All views expressed are the author’s own.
¹ JATO Dynamics, H1 2025 ADAS Adoption Report (India).
² National Crime Records Bureau, Road Accidents in India 2024.
