AI at the edge enables real-time failure prediction without the need for a cloud server
SANTA CLARA, Calif. & Kyoto, Japan, Nov. 29, 2022 (Globe Newswire) — ROHM Semiconductor Today they announced that they’ve developed machine learning An AI chip (SoC with AI accelerator for on-device learning) for cutting-edge IoT computing endpoints. The new AI chip uses artificial intelligence to predict failure (predictive failure detection) in electronic devices equipped with real-time actuators and sensors with extremely low power consumption.
In general, AI chips perform learning and inferences to achieve AI functions, as learning requires capturing a large amount of data, compiling it into a database, and updating it as needed. So, the AI chip that does the learning requires a lot of computing power and necessarily consumes a lot of power. Until now, it has been difficult to develop AI chips that can learn in the field and consume low power for high-end computers and endpoints to build an efficient IoT ecosystem.
Based on the “on-device learning algorithm” developed by Professor Matsutani of Keio University, ROHM’s newly developed AI chip is mainly composed of an AI accelerator (a dedicated hardware circuit for artificial intelligence) and ROHM’s 8-bit high-efficiency “tinyMicon MatisseCORE™” CPU. The combination of an ultra-small AI accelerator with a capacity of 20,000 gates and a high-performance CPU enables learning and inference with extremely low power consumption of a few tens of megawatts (1000 times smaller than traditional learning-capable AI chips). This allows real-time failure prediction in a wide range of applications, as the “anomaly detection results” (anomaly score) of unknown input data can be digitally output at the site where the equipment is installed without involving a cloud server.
Going forward, ROHM plans to integrate the AI accelerator used in this AI chip into various actuator and sensor IC products. Commercialization is scheduled to begin in 2023, with mass production planned for 2024.
Professor Hiroki Matsutani, Department of Information and Computer Science, Keio University, Japan
“With the advancement of IoT technologies such as 5G communications and digital twins, cloud computing will be required to evolve, but processing all data on cloud servers is not always the best solution in terms of load, cost and energy consumption. With the “learning on the device” we are researching and “algorithms On-device learning” that we developed, we aim to achieve more efficient data processing at the edge side to build a better IoT ecosystem. Through this collaboration, ROHM showed us the way to commercialization in a cost-effective way by further developing on-device learning circuit technology. I expect the AI chip prototype to be integrated into ROHM IC products in the near future.”
About tinyMicon MatisseCORE
tinyMicon MatisseCORE (Matisse: Micro aunit of account for tNew York seasy sequencer) is a ROHM proprietary 8-bit CPU developed with the purpose of making analog integrated circuits smarter for the IoT ecosystem. The improved instruction set for embedded applications, combined with the latest compiler technology, provides fast computational processing in a smaller chip area and program code size. Highly reliable applications are also supported, such as those requiring a qualification in accordance with ISO 26262 and ASIL-D functional vehicle safety standards, while a built-in “real-time debugging function” prevents the debugging process from interfering with the operation of the software, allowing errors to be debugged only during operation. app.
ROHM AI chip details (SoC with on-device AI learning accelerator)
The AI chip prototype (prototype part number BD15035) is based on the on-device learning algorithm (three-layer neural network AI circuit) developed by Professor Matsutani of Keio University. ROHM has shrunk the AI circuit size from 5 million gates to just 20,000 (0.4% by volume) to reconfigure it for commercialization as a proprietary AI accelerator (AxlCORE-ODL) controlled by ROHM’s 8-bit micro-CPU enabling AI learning and inference With a very low power consumption of a few tens of megawatts. This makes digital outputs of “anomaly detection results” possible for unknown input data patterns (eg, acceleration, current, brightness, sound) at the location where the equipment is installed without involving a cloud server or requiring prior AI learning, allowing in-time Real-time failure prediction (detection of predictive signs of failure) by on-site AI while keeping cloud server and connection costs low.
For AI chip evaluation, ROHM provides an evaluation board with Arduino compatible terminals that can be attached to an expansion sensor board to connect to an MCU (Arduino). Wireless communication modules (Wi-Fi and Bluetooth®), along with a 64k-bit EEPROM, is installed on the board. By connecting modules such as sensors and attaching them to the target equipment, it will be possible to check the effects of the AI chip from the screen. This rating board will be on loan from ROHM sales. Please communicate ROM sales for more information.
Demo video of the AI chip
Here is a demo video showing this AI chip being used with the scoreboard: https://youtu.be/SVn5CKFX9Uo
tinyMicon MatisseCORE™ is a trademark or registered trademark of ROHM Co., Ltd.
bluetooth® is a trademark or registered trademark of Bluetooth SIG, Inc.
 On-device learning: Run learning and inference on the same AI chip
#ROHM #developing #ultralow #power #Edge #chip #ondevice #learning