General Motors validates its advanced driver-assistance and automated driving systems by testing millions of possible scenarios in simulation before any vehicle reaches the road. In a profile of engineer Aamir Ali, the automaker described virtual testing as one of its most important tools for developing autonomy features across brands including Chevrolet, Cadillac, and Corvette. Ali, who previously built simulation environments at Cruise, says the approach lets engineers model edge cases and accelerate development cycles across a global vehicle portfolio.
Highlights
- GM tests millions of possible scenarios in simulation before a vehicle reaches the road.
- Super Cruise is in market today, with customers using it across hundreds of thousands of miles of compatible roads in North America, according to GM.
- Simulation work targets long-tail, rare scenarios that are difficult or unsafe to stage in the real world.
- GM plans to extend simulation across its full vehicle portfolio, supporting Active Safety, Assisted Driving, and Super Cruise.
Why Simulation Outpaces Road Testing
GM’s lineup spans many vehicle types and driving conditions, which the company says demands validation tools that cover far more situations than road testing alone. Simulation lets engineers model edge cases, shorten development cycles, and build confidence in system performance before vehicles reach customers.
“GM has an incredible scale and breadth of vehicles, which creates a unique challenge,” Ali said. “You’re not solving for one type of product or one type of customer. You’re solving for millions of drivers across many environments and use cases.”
That breadth is why the team is expanding virtual testing across GM’s autonomous vehicle organization, where simulation supports both new autonomy features and improvements to systems already on the road.
How GM Tests for Rare Scenarios
The harder validation problem, Ali says, is the long tail — the uncommon situations a vehicle must still handle correctly. Everyday driving is straightforward to test; rare events are not.
“One of the biggest challenges is handling the unknown,” Ali said. “It’s easy to test the situations that happen every day on the road. The hard part is preparing for rare scenarios that still have to be handled correctly.”
To close that gap, GM combines real-world data with proactive testing, building simulation environments that recreate varied scenarios and reveal how the system responds. The company also uses machine learning to generate new situations its software might encounter, pushing the system’s limits in a controlled way and reaching a level of coverage that would be difficult to achieve on the road.
Building on Super Cruise
GM’s large on-road fleet gives it access to real-world data and repeated opportunities to study how its systems perform, according to the company. Its Super Cruise hands-free driver-assistance system is already in market, with customers using it across hundreds of thousands of miles of compatible roads in North America. GM says that installed base provides a foundation to refine current features while developing the next generation, with the goal of extending simulation-driven development across Active Safety, Assisted Driving, and Super Cruise.
From Google to GM
Ali’s route into autonomous driving was not direct. He studied mechanical engineering in India, then spent more than a decade at Google building internal platforms for products including AdWords and YouTube, followed by work on trust and safety systems at Zoom. At Cruise, he helped develop simulation environments used to validate self-driving software as the team shifted a large share of testing from physical vehicles into virtual ones.
“If we do this right, it can change how vehicles are developed, how cities are designed and how people move from place to place,” Ali said.
Sign up for our weekly emails to stay up to date on the latest news!
