Although broadly referred to as AI (Artificial Intelligence), the specific functions and performance of AI vary considerably based on the technology used and the implementation platform, such as microcontrollers. This page provides an in-depth look at the implementation of ROHM’s Solist-AI™ and the benefits it delivers to customers.
Technical Details
The Solist-AI™ solution combines the Solist-AI™ MCU, equipped with proprietary on-device learning AI, with a wide range of support tools, including utility software and evaluation boards.
1. Edge Computing AI Microcontroller (Solist-AI™ MCU)
The Solist-AI™ microcontroller integrates ROHM’s proprietary hardware AI accelerator AxlCORE-ODL that utilizes a simple three-layer neural network architecture. The product lineup is being expanded with variations based on Arm® Cortex® processors, ROM capacity, and pin count to deliver on-device AI solutions across a broad spectrum of applications.
For detailed specifications and options, please visit the AI MCU product information page.
2. Solist-AI™ AI Algorithm and AxICORE-ODL
The AI algorithm used in Solist-AI™ is a form of Extreme Learning Machine (ELM) based on a three-layer neural network. This architecture is optimized for hardware implementation, allowing for efficient AI processing with minimal memory usage.
The AxlCORE-ODL hardware accelerator enables high-speed, low-power execution of AI algorithms. Beyond accelerating AI computations, batch execution of the entire AI workflow is possible, significantly reducing the load on the MCU. Furthermore, by reconfiguring batch operations, tasks such as FFT can also be accelerated.
3. AI Operation Verification Simulator Solist-AI™ Sim
Solist-AI™ Sim is a PC-based simulation tool that allows users to validate the functionality of Solist-AI™. AI learning and inference results can be easily verified, facilitating the rapid assessment of AI effectiveness. Each simulation requires only a few seconds, making it possible to efficiently evaluate performance under a variety of conditions.
Simulation tools are currently available for both anomaly detection (unsupervised learning) and prediction/parameter estimation (supervised learning).
4. Solist-AI™ Scope Utility for Verifying Operation on Actual Devices
Solist-AI™ Scope is a visualization tool that displays internal data processed by the Solist-AI™ microcontroller as waveforms. Sensor input data and AI-generated anomaly scores can be monitored in real time, making it easy to verify whether the AI is functioning as expected.
*Solist-AI™ Scope is included with the reference software.
5. Embedded Software Development
Reference software (peripheral drivers, libraries, sample software)
The bundle includes a range of peripheral drivers, AI libraries, and sample programs for the Solist-AI™ MCU. Functionality can be verified using the LEXIDE-Ω integrated development environment along with reference board.
Also included is Solist-AI™ Scope, a real-time viewer for monitoring Solist-AI™ operation. Please note that a connection adapter (MM-FT232HC*1) is required and must be prepared separately.
Sample AI Application Software
Sample software is available for developing an anomaly detection AI application using the Solist-AI™ microcontroller. Initial learning and inference are conducted based on acceleration data collected from the accelerometer to calculate an anomaly score.
Running the sample software requires the LEXIDE-Ω integrated development environment, a reference board, a accelerometer, and a PC connection adapter (MM-FT232HC*1). These components must be prepared separately by the customer.
*1 The MM-FT232HC is a product of Sunhayato Corp.
*2 FT232H/FT2232H is a product of Future Technology Devices International Limited (FTDI).
6. ARM® Core Integrated Development Environment
LAPIS Development Tool LEXIDE-Ω is an application program development environment for ROHM microcontrollers featuring proprietary 8-bit RISC (nX-U8/100 core), 16-bit RISC (nX-U16/100 core) processors as well as Arm® Cortex®-M cores.
Built on the open-source Eclipse platform with CDT plug-in, this integrated development environment enables efficient, streamlined software development.
For those working with Keil® MDK, please contact us separately.
7. Microcontroller-Equipped Reference Board (Provided by ROHM)
A dedicated reference board is available for verifying the functionality of the Solist-AI™ MCU. It supports software development, debugging, and Flash memory programming. In addition, by connecting Solist-AI™ Scope, users can monitor AI operation in real time.
8. Microcontroller-Equipped Reference Board (Provided by an Ecosystem Partner)
A system evaluation board equipped with an AI MCU is available through one of our ecosystem partners. For more information, please visit our ecosystem partners page.
Implementation Flow and Use Cases
Solist-AI™ supports customers throughout the evaluation, verification, and deployment stages by combining proprietary on-device learning AI technology with a robust set of support tools. Here we will outline the implementation flow and use cases of Solist-AI™.
Solist-AI™ Use Case 1: Predictive Maintenance
Detecting Motor Bearing Damage: Utilized in condition-based monitoring (CBM) to enable predictive maintenance of production equipment/machinery
This includes various peripheral drivers, AI libraries, and other resources for the Solist-AI™ MCU.
Background and Challenges
- Predictive maintenance is required to detect anomalies and potential failures in production equipment and machinery before they occur
- A shortage of skilled personnel (experienced engineers) capable of performing regular inspections remains a significant challenge
Uses of Solist-AI™
- During initial equipment installation, on-site learning is conducted using accelerometers to capture vibration data under constant-speed operation, establishing a reference profile for normal conditions.
- The vibration state during normal operation is continuously compared to the learned baseline data, and when deviations beyond a certain threshold are detected, the system issues an alert indicating a potential anomaly (= 'different from usual').
Expected Benefits
- Leverage AI to detect ‘different than usual’ at an early stage and facilitate root cause analysis
⇒ Helps prevent unexpected equipment repairs/replacements as well as both minor and major unplanned downtimes - Shift from manual TBM to AI-driven CBM
⇒Significantly reduces overall maintenance costs
*The above is one example of predictive maintenance. Applicability extends beyond motor systems to various other types of systems as well.
Solist-AI™ Use Case 2: Condition/Degradation/Parameter Prediction
Battery Degradation Prediction/Condition Change Detection: Used to forecast future indicators of product degradation and estimate parameters that cannot normally be measured
Background and Challenges
- Customers operate battery-powered devices under a wide range of scenarios
- The inability to track degradation of the battery’s full charge capacity imposes limitations on how devices can be used
Uses of Solist-AI™
- Standard degradation profiles (SOH) are pre-trained prior to shipment. Supplementing the model with real-world usage data (SOH) enables high accuracy prediction of battery degradation
- The standard charge/discharge characteristics of each battery cell are pre-learned before shipment. Additional learning is performed during actual use to detect changes in condition, taking individual cell variations into account.
Expected Benefits
- Learn customer-specific usage patterns and environmental conditions
⇒Predict future degradation to extend the operational lifespan - Incorporate cell-to-cell variations into the learning process
⇒Enable early detection of charge capacity loss, imbalance anomalies, and condition changes
*This example illustrates a use case in condition/degradation prediction, which can be extended beyond battery systems to various other applications.
Solist-AI™ Use Case Portfolio
Beyond use cases like motor anomaly detection and battery degradation, we have compiled comprehensive documentation covering a broad spectrum of implementation examples spanning industrial equipment, electrical devices, and production systems. Please refer to this as a resource when evaluating and implementing Solist-AI™.
Collaboration with Ecosystem Partners
Solist-AI™ is delivered in collaboration with a diverse network of ecosystem partners. From hardware and software development to infrastructure deployment, we offer end-to-end support, enabling immediate implementation by customers.
To further drive AI innovation through Solist-AI™, we are actively expanding our ecosystem and welcome inquiries from prospective partners. Interested parties are encouraged to contact us for more information.
Contact Us / Join Solist-AI™ Ecosystem
We welcome companies that share our vision for the Solist-AI™ ecosystem and aspire to drive technological innovation together. Will you join us as a Solist-AI™ ecosystem partner and create new value side by side?
Note: Solist-AI™ is a trademark or registered trademark of ROHM Co., Ltd.
ARM®, Cortex®, and Keil® are trademarks or registered trademarks of Arm Limited (or its subsidiaries).