The founders of Argo AI, the self-driving startup backed by Ford, are giving back to their alma mater to the tune of a $15 million investment in Carnegie Mellon University to fund the creation of a new research center.
The Carnegie Mellon University Argo AI Center for Autonomous Vehicle Research (phew) will use those funds to “pursue advanced research projects to help overcome hurdles to enabling self-driving vehicles to operate in a wide variety of real-world conditions, such as winter weather or construction zones,” the company and university announced on Monday. Argo was founded in 2016 by a team of CMU alumni.
Argo — which, along with Ford, is testing its vehicles in Miami, Washington, DC, Palo Alto, and, most recently, Detroit — will support research into advanced perception and next-generation decision-making algorithms for autonomous vehicles. In other words: the software and hardware that power a self-driving car’s ability to “see” and “think.”
But this isn’t a navel-gazing project by Argo, nor is it a benevolent gift by a handful CMU alums who made it big. (In 2017, Ford said it would spend $1 billion on Argo over five years.) Rather, this research project is geared toward enabling the “large scale, global deployment” of self-driving cars. This is money to get self-driving cars on the roads faster and at scale.
Autonomous vehicles are being tested in small batch deployments in cities around the world, but they are still a long way from “global deployment.” To get there, the cars need to be proven safe to operate in all types of road and weather conditions. People need to trust the technology — which they currently do not — and they need to be cheaper and more efficient than taxis, ride-hailing apps like Uber and Lyft, and personally owned vehicles.
Argo offered a taste of the types of projects it hopes to sponsor through this new partnership with CMU:
For instance, how can autonomous vehicles “see” their surroundings and operate safely in adverse weather such as very heavy rain, falling snow and fog? How can we reduce or eliminate reliance on high-definition maps without sacrificing safety and performance? How can autonomous vehicles reason in highly unstructured broken-traffic conditions commonly found in some big international cities, where actors on the road completely ignore any road rules? How can we reduce our need for labor intensive high definition map data when moving to new cities? Once fleets of vehicles are deployed, how do we efficiently leverage the experiences of an autonomous fleet to ultimately obtain exponential improvements beyond the original launch capabilities?
Last week, Argo released Argoverse, its HD maps dataset. Making datasets like these available to the research community for free “helps compare the performance of different (machine learning – deep net) approaches to solve the “same” problem,” said Raj Rajkumar, an electrical and computer engineering professor at Carnegie Mellon University who is not affiliated with Argo. “In other words, they provide some sort of a standard benchmark.”
Argo isn’t the only company focused on bolstering the research community. Last year, Intel launched the Institute for Automated Mobility in the driverless testing hotbed of Phoenix, Arizona. The institute combines the three state universities with the state’s departments of Transportation, Public Safety, and Commerce as well as the companies working on automated cars, trucks, and drones. Intel did not disclose its financial commitment to the institute.