Autonomous driving dominates the discussion around artificial intelligence in the auto sector, but it’s just one of several applications carmakers are exploring for the technology. FogHorn Systems Inc. and Porsche AG have come up with another use case: enabling vehicles to recognize their owners.

FogHorn detailed its collaboration with the sports car maker this morning. The startup, which is backed by more than $47 million in funding from investors that include Intel Corp., sells software that helps enterprises to process sensory data from connected devices. It teamed up with Porsche through Startup Autobahn, an accelerator program that connects emerging technology companies with major players in the auto sector.

FogHorn partnered with the carmaker on a 100-day project to build an automatic car unlocking system. The startup put together a prototype that enables users to access their vehicles without a key fob. Leveraging an infrared camera, the system scans the user’s face when they walk up to their vehicle, confirms that they are in fact the car owner and then automatically opens the door.

The facial verification is paired with multifactor authentication to reduce the risk of abuse. Before they can access the vehicle, users have to confirm their identify a second time using their phone or another device.

The computational heavy lifting involved in the verification process is carried out by FogHorn’s analytics software. The startup’s offering, a platform called Lightning, can run real-time data processing workflows and AI models on connected devices with limited computing resources. Moreover, the software performs the analysis locally without sending data to the cloud, which allows it to work even in the absence of a reliable internet connection. 

“Our stack can run in any hardware, we’re completely OS-independent – it’s a full-up software layer,” FogHorn chief executive David King detailed in November interview (below) on theCUBE, SiliconANGLE Media’s video studio. “The whole stack is about 100 megabytes of memory with all the components.”

That’s many times smaller than the several gigabytes of memory often required to run traditional stream processing systems such as Apache Spark, according to King. The result, he said, is an “order-of-magnitude of footprint reduction and speed of execution improvement” in data analytics. 

Photo: Porsche