Recogni introduces compute platform for autonomous mobility

2 mins read

Recogni, a company working with AI-based perception purpose-built for autonomous vehicles, is making available the Phoenix ADAS/AD ECU system.

This is the first complete system with processing capacity for high resolution sensor data with low latency and low power consumption for autonomous vehicles.

The Phoenix system combines the company’s Scorpio AI processor and the Renesas R-Car V4H ASIL D support processor, both leading edge 7nm designs, offering a more efficient and scalable solution for the high compute needs of various autonomous driving levels, from SAE levels L2+ to L3 to L4.

“Recogni was founded on the premise that true autonomous driving can only be achieved when a system collects and processes environmental sensor data far better than humans," said Marc Bolitho, CEO of Recogni. "The Phoenix system is unmatched in its ability to process high resolution data with low latency and extremely low power consumption. It is a feature rich solution with an open software architecture offering OEMs the ability to customize their autonomous driving technology stacks to deliver the desired ADAS features to customers and scale their offerings over time for L2+ to L4."

Combining both Recogni and Renesas technologies, the Phoenix system is able to deliver AI compute performance across a wide range of application needs and demands – and is scalable from 500, 1000 to 2000 TOPS, while also having the lowest latency and zero jitter.

Having the capability to process multiple high resolution high frame rate video streams simultaneously, the Phoenix ECU is designed to be future proof, having the ability to be remotely updated with new features and expanded capabilities and functions without any upgrades of hardware.

This capability will allow OEMs to deliver software-defined vehicles over an extended period of time.

Recogni offers an open software platform along with a comprehensive L2+ capable AI vision perception stack that will include a wide variety of object classification and detection features - vehicles, VRUs (vulnerable road users), traffic signs and traffic light detection, lane detection, etc.

The open nature of the platform enables partners and customers to run their own perception and driving function software.

Phoenix, a safety centric purpose-built ASIL compliant system is capable of delivering AI convolutional neural network performance with the low latency, while simultaneously supporting the vision Transformer Networks. Phoenix's low power consumption allows for passive cooling which effectively reduces system cost.

Commenting Daniel Cisco, Senior Director, Automotive Digital Products Marketing Division at Renesas, said, "Our collaboration with Recogni demonstrates the power of our combined solutions. With Recogni's Scorpio AI processor, our already notable AI capability on the R-Car V4H SoC has increased substantially.  Our combined system provides automotive Tier 1's and OEM's a scalable architecture with the highest level of compute at the lowest power envelope."