Autonomous intelligence is the future of all “internet of things” applications. To drive that intelligence, developers are pushing the architectural concept of “digital twins” into more and more edge devices.

At heart, a digital twin drives the analytics that guide a physical entity’s ongoing adjustments to its environment. As the name implies, digital twins are data constructs that mirror specific physical entities and thereby help to manage them, either through remote connection or through autonomous local operations.

Essentially, a digital twin describes the current configuration, state, condition, behavior, location and other attributes of some physical device that possesses an IoT connection. It aggregates, manages and analyzes the sensor data emitted by its IoT-connected physical counterpart.

It also supports data-driven functions such as closed-loop simulation, monitoring, maintenance, diagnostics, remediation, and other life-cycle management functions vis-à-vis that analog entity.

Though digital twins have taken root first in the industrial IoT, this practice’s applicability is not limited to that domain. Depending on where they’re embedded and the sophistication of their analytic smarts, digital twins can perform many data-driven real-time optimization functions:

• On shop-floor physical devices, they can predictively manage equipment performance, prevent failures, and boost efficiency and throughput.
• On routers, controllers, and other networking devices, they can predictively optimize bandwidth, maintain quality of service and guarantee service level agreement metrics.
• On access devices, clustered servers, storage area networks and other enterprise information technology assets, they can predictive maintain high availability, reliability and security of all components.
• On shipping containers, warehousing systems, long-haul trucks, and other logistics and supply chain systems, they can predictively ensure real-time asset tracking, efficient handling, quality assurance and on-time delivery.
• On smart-city infrastructure, they can predictively ensure that public utilities, mass transit, emergency response, and other critical systems predictively adjust to changing weather, traffic and other conditions.
• On mobile gadgets, they can predictively ensure that intelligent virtual assistants serve users’ interests through embedding of personalized profile information, environmental context, interest graphs and so on.

Embedding of machine learning intelligence is key to the effectiveness of digital twins in performing sensor-driven predictions, pattern recognitions and other algorithmic feats. In the ML pipeline, sensor-infused digital twins provide a key source of training data that enables continual retuning of edge-embedded intelligence.

With all that in mind, Wikibon takes note of digital-twin startup SWIM.AI, which recently came out of stealth mode to announce the general availability of its SWIM EDX solution. This offering provides lightweight device-resident ML-driven software that enables on-the-fly ML-driven reduction, analysis and prediction from sensor data collected locally on edge devices. Self-training digital twins that run locally on edge devices interoperate in a mesh architecture to support a wide range of autonomous, cooperative and other edge-intelligence scenarios.

The mesh manages real-time dynamic state information across distributed devices that incorporate SWIM’s digital twin technology. As these distributed digital twins learn what data-streaming patterns are normal vs. anomalous within the mesh fabric, they can adaptively identify critical events, find hidden patterns, predict future behaviors and autonomously trigger fast actions at individual or multiple coordinated edge devices.

This announcement is a key milestone in the development of IoT infrastructures in which more AI is being embedded in edge devices that are retraining themselves dynamically through what Wikibon refers to as “federated” approaches. The operational benefits of using real-time streaming digital-twin technology for building and training edge-resident ML are undeniable. They include improvements in ML contextual accuracy, reductions in the need for manual ML training, and savings in data bandwidth, storage and cloud processing costs.

Wikibon, which recently discussed digital twin technology in one of our weekly Action Item calls, expects to see this digital-twin-based approach become standard in the growing edge-intelligence marketplace. It’s good to see that SWIM is working closely with customers and partners in many sectors to bring their dynamically adaptive edge intelligence into many real-world infrastructures and applications.

Wikibon’s Peter Burris recently spoke with SWIM Chief Technology Officer Simon Crosby and founder and Chief Architect Chris Sachs in the Palo Alto, California studios of SiliconANGLE’s theCUBE:

Image: geralt/Pixabay