FogHorn Systems Inc., a provider of edge intelligence software for industrial “internet of things” deployments, today updated its Lightning platform to enable more machine learning models to run at the network edge.

The platform provides intelligent software to enable edge computing in the industrial sector. The company’s customers include businesses in the transportation, manufacturing and petroleum fields, as well as smart cities and buildings. They use the Lightning platform for tasks such as predicting and monitoring conditions in real time and predictive maintenance.

The idea is that by performing these kinds of computing tasks at the edge, companies can analyze data in real time that would otherwise be sent to the cloud to be processed. That’s an important capability for time-sensitive applications, which include self-driving cars that need to make split-second decisions and don’t have time to wait for a response from a cloud server.

FogHorn’s software is a “closed-loop inference engine that enables data to be sent to the cloud and updated models sent back to the edge,” said Keith Higgins, vice president of marketing. Cloud service providers “can’t deliver a lot of value close to the sensor layer. That’s the primary reason they’ve partnered with us.” Google LLC announced a partnership with FogHorn last spring and both Microsoft Corp. and Amazon Web Services Inc. have less-formal agreements to support the company’s products with their cloud services.

FogHorn’s Lightning software isn’t the only edge intelligence platform around, but it does stand out for its embedded machine learning capabilities, which are powered by a “highly miniaturized complex event processing (CEP) engine,” the company said. Machine learning is a subset of artificial intelligence that involves training algorithms to become more accurate at predicting outcomes without programming them to do so. It involves feeding those algorithms with data and the use of statistical analysis to improve their accuracy.

With the update, FogHorn is further augmenting its machine learning capabilities with support for Predictive Model Mark Up Language, which allows predictive models to be shared across multiple applications. The result is that customers can now run any PMML-compliant machine learning model at the edge.

“A lot of models are built with the cloud as the playground, but when you want to run something at the edge it has to be small and efficient,” said Ramya Ravichandar, FogHorn’s director of product management. “We optimize to run on small footprint.” The company said its engine can run most popular machine learning libraries on a device as small as a Raspberry Pi and that its optimization engine shrunk one 500-megabyte machine learning model down to 2.5 megabytes.

These capabilities enable a new concept called “sensor fusion,” which is a technique that enables analysis of data from disparate sources in order to improve application or system performance. It works by automatically correcting for the deficiencies of individual sensors.

Lightning 2.0 also gains new capabilities for automating and scaling sensor deployments, such as automatic sensor-discovery, sensor fusion, edge device auto-registration, and single-click deployment to thousands of devices at once. In addition, FogHorn is updating Lightning’s user interface in order to try to shorten the learning curve for new users and make it easier to execute common operations.

Pierce Owen, principal analyst at ABI Research, said the update was timely because many companies have already seen success with their initial edge intelligence deployments.

“Now, the task at hand is to efficiently deploy these projects at large scale,” Owen said. “This latest release from FogHorn focused on going from pilot to broad commercial rollout is well-timed with the state of the market’s needs.”

FogHorn Chief Technology Officer Sastry Malladi appeared on theCUBE, SiliconANGLE Media’s mobile livestreaming studio, at the BigData SV event in San Jose, California last March, where he discussed how a number of specific customers were using the Lightning platform:

With reporting from Paul Gillin

Image: FogHorn Systems