Artificial intelligence (AI) at the edge enables real-time failure prediction without requiring a cloud server, said Rohm Semiconductor at the launch of its on-device learning AI chip. This is an SoC with on-device learning AI accelerator for edge computer endpoints in the IoT. The chip uses AI to predict failures (known as predictive failure detection) in electronic devices which are equipped with motors and sensors. It does this in real-time with low power consumption, said Rohm.
Generally, AI chips perform learning and inferences to achieve AI functions. Learning requires that a large amount of data gets captured, compiled into a database and updated as needed, which means that the AI chip which is learning requires substantial computing power.
Based on an ‘on-device learning algorithm’ developed by Professor Matsutani of Keio University, Rohm has developed an AI chip which mainly consists of an AI accelerator (AI-dedicated hardware circuit) and Rohm’s high-efficiency, eight-bit CPU, the tinyMicon MatisseCORE.
Matisse is the micro arithmetic unit for tiny size sequencer. The tinyMicon MatisseCORE was developed specifically to make analogue ICs more intelligent for IoT ecosystem.
Combining the compact, 20,000-gate AI accelerator with a high performance CPU enables learning and inference with low power consumption of just a few tens of mW (1,000 times smaller than conventional AI chips capable of learning, said Rohm). This allows real-time failure prediction in a wide range of applications, since ‘anomaly detection results (anomaly score)’ can be output numerically for unknown input data at the site where equipment is installed without involving a cloud server.
An instruction set optimised for embedded applications and the latest compiler technology delivers fast arithmetic processing in a smaller chip area and program code size, explained Rohm. It is also feasible for high-reliability applications, such as those requiring qualification under the ISO 26262 and ASIL-D vehicle functional safety standards. The proprietary onboard real-time debugging function prevents the debugging process from interfering with program operation, allowing debugging to be performed while the application is running.
Rohm has announced plans to incorporate the AI accelerator in this AI chip into various IC products for motors and sensors. Commercialisation is scheduled to start in 2023, with mass production planned in 2024.
The AI chip was evaluated used an evaluation board equipped with Arduino-compatible terminals that can be fitted with an expansion sensor board for connecting to an MCU (Arduino). Wireless communication modules (Wi-Fi and Bluetooth) along with 64kbit EEPROM are mounted on the board. Connecting units such as sensors and attaching them to the target equipment it will be possible to verify the effects of the AI chip from a display, said Rohm.
This evaluation board will be available on loan from the company.
