80Top/s AI vision and real-time control MCU for industrial robots and machinery

Called RZ/V2H, the ICs include the company’s DRP-AI3 ‘dynamically reconfigurable processor’ AI accelerator – recently revealed at ISSCC in San Francisco – in 10Top/s/W form.

“Pruning technology employed in the DRP-AI3 accelerator improves AI computing efficiency, boosting inference performance to 80Top/s,” according to the company.

Also on the die are four 1.8GHz Arm Cortex-A55 CPUs for Linux application processing, two 800MHz Cortex-R8 cores for real-time processing, and one Cortex-M33 sub-core.

This core count puts the new MCUs at the top of the RZ/V2 vision processing MCU family

“RZ/V2H will facilitate the development of next-generation autonomous robots with vision AI capabilities, that have the ability to think independently and control movements in real time,” claimed the company. “Renesas has applied its DRP technology to develop the OpenCV accelerator that speeds up the processing of OpenCV, enabling real-time execution of visual SLAM used in applications such as robot vacuum cleaners.” – OpenCV is an open-source library for vision processing, SLAM is simultaneous localisation and mapping.

It is also offering a library of pre-trained AI models (dubbed ‘AI Applications’) and a software development kit (‘AI SDK’).

There are so far six variants: with or without security hardware, with or without an Arm Mali-G31 graphics accelerator and, for those with a Mali-G32, the option of adding a Mali-C55 image signal processor.

Interfaces include PCIe, USB 3.2 and Gigabit Ethernet.

In all cases, packaging is a 19mm, 0.5mm pitch, 1,368 ball, BGA.

Working with Renesas, debug company Segger pre-integrated RZ/V2H support into its J-Link debug probe.

The RZ/V2H product page can be found here and the company has a decent white paper on its DRP-AI3 accelerator