Ambarella's Oculii 4D Imaging Radar Uses Next-Gen CVflow to Boost Performance, Slash Power Draw

Designed around the next-generation CV3 system-on-chip, the Oculii system makes for dramatically simpler 4D imaging radar implementations.

Edge AI specialist Ambarella has announced the launch of what is claimed to be the "world's first centrally-processed 4D imaging radar architecture" — and is hoping to find it a home in autonomous vehicles and more.

"No other semiconductor and software company has advanced in-house capabilities for both radar and camera technologies as well as AI processing," claims Ambarella's president and chief executive Fermi Wang. "This expertise allowed us to create an unprecedented centralized architecture that combines our unique Oculii radar algorithms with the CV3's industry-leading domain control performance per watt to efficiently enable new levels of AI perception, sensor fusion and path planning that will help realize the full potential of ADAS [Advanced Driver Assistance Systems], autonomous driving, and robotics."

Ambarella has launched its latest edge AI chips with a new use case: power-efficient 4D imaging radar. (📹: Ambarella)

As Wang explains, the Oculii system is built around the company's latest CV3 system-on-chip range, which provides the performance required to carry out real-time on-device analysis of incoming radar data — while having enough power spare to handle sensor-fusion, path planning, perception, and more. More importantly, it's capable of dramatically reducing the hardware requirements for 4D imaging radar — slicing the number of antennas required down to a more manageable number.

Using Oculii's artificial intelligence system, Ambarella claims, the "thousands of MIMO [Multiple-Input Multiple-Output] antennas" required by traditional 4D radar imaging systems are reduced to an array of just 14 — six for signal transmission and eight for reception. Despite this, the system is competitive on angular resolution — but draws "considerably less power" and cuts the bandwidth required for data transportation to a sixth of its competitors, while boasting "up to 100x" the performance.

The secret sauce behind the company's claims: the next-generation CVflow architecture in the CV3, which now includes both a neural vector processor and a general vector processor in order to boost performance over the company's earlier architectures. Using the latest version CV3 chips, Ambarella is hoping to encourage adoption of its Oculii system for autonomous vehicles, robotics, and more.

Ambarella has confirmed plans to demonstrate the Oculii system at the Consumer Electronics Show (CES) in January, and requests anyone interested in sampling or evaluating the technology to get in touch with a representative.

Gareth Halfacree
Freelance journalist, technical author, hacker, tinkerer, erstwhile sysadmin. For hire: freelance@halfacree.co.uk.
Latest articles
Sponsored articles
Related articles
Latest articles
Read more
Related articles