Eoxys, a leading IoT solution provider, introduces the XENO+ NB-IoT ML SOM (System-On-Module) module, a production-ready SOM module targeting intelligent and secure IoT/AIML devices for applications such as smart metering, smart lighting, smart tracking, and industrial automation. Setting an industry first, Eoxys’ XENO+ NB-IoT ML SOM integrates Renesas’s RH1NB200 LTE Cat NB1/NB2 modem with Syntiant’s NDP120 Neural Decision Processor.
This integration brings together AI and Deep Learning capabilities with cellular connectivity, unlocking unparalleled opportunities for innovation in the IoT ecosystem. By leveraging the intelligent compute and cellular connectivity features of the XENO+ NB-IoT ML SOM, customers can accelerate the development of their IoT smart metering and smart tracking products. With the freedom to focus solely on adding sensors and their unique value proposition, customers can expedite their product development while ensuring optimal security and performance.
Figure-1: XENO+ NB-IoT ML SOM module top and bottom view
The Renesas’s RH1NB200 LTE NB-IoT modem integrated into the XENO+ NB-IoT ML module is designed to operate seamlessly on the networks of all major Indian telecommunications carriers and mainly supports the B1, B3, B5 and B8 bands. It is an ideal solution to serve the Indian smart metering, smart lighting and smart tracking markets. The RH1NB200 modem offers extremely low power consumption, significantly extending battery life. It also provides an integrated EAL5+ (Evaluation Assurance Level) Secure Element (SE) that provides zero compromise on security to ensure the safety of end applications, particularly smart energy and smart water metering systems.
“Our new NB-IoT ML SOM module is ideal for developing smart metering, smart lighting and smart tracking devices quickly with built-in machine learning capability for battery-powered devices” said Prabhakar Annavi, CEO of Eoxys Systems. “Our ML SOM module is readily usable, allowing customers to quickly build target-specific NB-IoT enabled devices by building only their carrier board, enabling data to be uploaded to the cloud via the networks of all major Indian telecommunications carriers.”
“The convenience and expedited time to market of the XENO+ NB-IoT ML SOM module, combining the Renesas’s RH1NB200 LTE NB-IoT modem adds tremendous value to Indian customers with accelerated NB-IOT enabled products that require security and extremely low-power,” said Amit Bavisi, VP of Renesas.
The XENO+ NB-IoT ML module is packaged with the Syntiant NDP120 Neural Decision Processor™ that can run multiple audio and time-series sensor data machine learning applications simultaneously with minimal power consumption. The NDP120 processor is designed to natively run multiple deep neural networks on a variety of architectures, including CNNs, RNNs and fully connected networks up to 256 layers.
“Eoxys’ new NB-IoT ML SOM module utilizes NDP120 with Syntiant’s second-generation Core 2 neural network engine bringing highly accurate, low-power inference to smart metering devices with integrated audio and sensor machine learning capability,” said Mallik Moturi, CBO of Syntiant.
The key benefits of NB-IoT ML SOM module are,
- The customers can save up to 40% of hardware and software development time.
- The Software SDK offered along with ML SOM modules helps the Customers with reduced Embedded SW development time.
- Common pin mapping across all XENO+ ML SOM modules brings SW compatibility among all connectivity variant devices.
- The customers can leverage FOTA feature offered with XENO+ ML SOM to upgrade MCU Application binary, ML Model binaries and Modem binaries.
- The NDP120 can run multiple Deep Neural Networks (DNN) on variety of architectures such as CNNs, RNNs and fully connected networks up to 256 layers.
- The customers can first start with IoT use cases and upgrade with ML use cases on the same device in future.
- The following eco-system components helps the developers during their embedded SW development time:
- XENO+ Flashing PC tool
- XENO+ SDK with Networking and Machine Learning APIs
- Bootloader for Serial Flashing, FOTA and secure boot
- FOTA Server for host MCU application binary and ML model for NDP120.
To demonstrate the capability of NB-IoT ML SOM on real embedded devices, an Audio Classification Device (ACD) is designed based NB-IoT ML SOM module.
Figure-2: Audio Classification Device
This ACD device is loaded with Human Alert sounds ML model for classifying the human alert sounds coming from the environment in real-time. The list of Human Alert sounds that can be detected by this device are listed below:
- The Baby cry
- Gun shot
- Dog bark
- Drilling
- Car theft alarm
- Jackhammer
- Fire alarm
- Glass break
- Hammer
- Men’ scream
- Women’ scream
The NDP120 processor classifies the human alert sounds and the classified outcomes are sent to IoT Dashboard application via LTE NB-IoT connectivity. The LTE NB-IoT connectivity uses both NIDD and HTTP based data protocol.
An AIML Evaluation Kit (EVK) is also available for this NB-IoT SOM module. The objective of AIML EVK Kit is to create a new IoT product development experience for developers, researchers, product design engineers and students and provided them with hands-on experience with sensors such that they can carry-out experiments on AIML based IoT use cases.
Figure-3: XENO+ NB-IoT ML SOM Evaluation Kit
Using this EVK kit, the developers can write embedded C based applications to collect data from the sensors via peripheral interface drivers of I2C/ UART/ SPI/ ADC/ DAC interfaces. They can build the Tiny ML models and load the same into AIML EVK Kit to run the model on the audio or time-series sensor data for advanced ML classifications of various target applications. The developers can push the ML classification outcomes to cloud server via HTTP/ MQTT/ TCP data protocols via NB-IoT connectivity.
