Edge Impulse is proud to announce that the Edge Impulse for Linux deployment integration with AWS IoT Greengrass is now available. AWS IoT Greengrass is a customizable deployment solution, part of the larger AWS IoT service family of IoT solutions, that enables scalable deployments of software to a large array of Greengrass-enabled edge devices.
Why Does This Matter?
Enterprises face significant challenges when designing, deploying, and managing machine learning (ML) models across thousands of edge devices. For instance, companies deploying real-time monitoring models on millions of wearables often struggle with delays, errors, and high operational costs, especially when updates are frequent. Similarly, manufacturers scaling predictive maintenance models across factories experience costly downtime and complexity when managing large collections of devices. These inefficiencies can lead to significant productivity losses.
Edge Impulse, together with AWS, can provide an effective solution by combining their strengths. Edge Impulse simplifies the development and deployment of ML models specifically crafted for the edge, while AWS IoT Greengrass and IoT Core automates and simplifies the management of fleets centrally as well as scales management operations, reduces complexity of management, and provides seamless scaling capabilities.
Why Edge and Cloud Work Better Together
Edge devices run ML models locally, reducing latency and bandwidth while focusing computational needs where they are most needed/important. This enables near real-time decisions and actions critical for device classes that make significant use of ML at the edge (i.e. wearables as well as factory automation/management). The cloud handles resource-intensive tasks like model training, model testing, advanced analytics, and generalized fleet management, ensuring good coordination across the network. The optimal balance between the edge and the cloud is that edge devices focus on the most immediate, localized tasks involved with the ML models specifically, while the cloud serves as an aggregation center for broader insights, strategic decisions, and refined improvements of the higher level decision making process.
The Edge Impulse and AWS IoT Greengrass Advantage
The integration of AWS IoT Greengrass and the Edge Impulse platform streamlines the entire edge-to-cloud experience. Businesses can update and manage ML instances across thousands of edge devices effortlessly. Edge device results can be communicated into the greater cloud via AWS IoT Core for deeper analysis and decision making processes. Inferences at the edge trigger immediate, localized actions, while the cloud not only serves as the coordination management point for ML management, but also as the higher level analytics and decision coordination systems. By uniting edge processing efficiencies with the scalability of cloud computing, Edge Impulse and AWS can empower businesses with greater decision making processes, reduced costs and optimized performance over time.
Our solution is also available on the AWS Marketplace, backed by the trusted AWS ecosystem. For more details about our listings or to discuss a private offer, feel free to reach out to us.
“Under the Hood” — Scalable Deployment Made Simple
To accomplish this integration, Edge Impulse’s Linux-based Runner capability has been augmented to serve as a key artifact within an AWS IoT Greengrass custom component. The component can be deployed across all Greengrass-enabled edge devices within an organization.
As an AWS IoT Greengrass component, the Edge Impulse Runner service is easy to deploy at scale and integrates seamlessly with AWS IoT Core. This integration allows all ML inferences made by the Runner service to be conveyed into AWS IoT Core for additional processing/action generation or to trigger cloud-based “actions” on the ML inferences.
The integration between AWS IoT Greengrass, AWS IoT Core, and Edge Impulse’s platform services is illustrated in the following architectural diagram:
If you are new to AWS IoT Greengrass, it is recommended that you consult the AWS documentation to learn about how the service works and what it is typically used for.
The Edge Impulse Runner runtime integration is straightforward with AWS IoT Greengrass:
- A “custom” Greengrass component recipe and artifact set is available for download from the edgeimpulse/aws-greengrass-components repository
- The recipe and its artifacts can be used to create a custom Greengrass component within your own AWS organization
- Once created, you can then create Greengrass deployments that target all/some/subsets of your deployed edge devices to receive and begin to run the Edge Impulse Runner runtime
- Once running, the Edge Impulse Runner runtime will automatically download and invoke the latest model that you have created within your Edge Impulse environment
- Greengrass will monitor the Edge Impulse Runner and keep you informed of the health of your ML fleet
Here is an example illustrating the new IoT Core integration. In the screen capture you see the MQTT messages produced by Edge Impulse Runner service as it executes an object detection ML model. The messages are relaying the inference result telemetry into IoT Core using an edge device-centric topic:
This is especially useful to create actions/rules within the AWS Cloud to “act/respond” to the inference results your Edge Impulse Runner runtime is conveying. AWS provides many mechanisms that can “hook” into AWS IoT Core data and create a bevy of “actions” targeting other AWS services as a result. AWS IoT Core Rules are an example:
To get started, please review the Edge Impulse AWS Greengrass documentation to begin with your own AWS organization and Edge Impulse.
Edge Impulse will provide new and innovative integrations over time as the core AWS IoT services continue to evolve. Please stay tuned for new updates and new exciting features!
If you would like to learn more about the Edge Impulse integration with AWS IoT Greengrass and IoT Core and are planning to attend AWS re:Invent 2024, please visit the AWS Village IoT kiosk and/or attend the 300-level workshop IOT-316R/R1: “Unleash edge computing with AWS IoT Greengrass on NVIDIA Jetson.”
Got questions or ready to take the next step? Contact our team today for personalized assistance and insights tailored to your needs.
Authors: Sophia Zimmerman (Edge Impulse) Doug Anson (Edge Impulse), Jenny Plunkett (Edge Impulse), Channa Samynathan (AWS), Joe Julicher (AWS), Yong Ji (AWS), Minsung Kim (AWS)