IBM uses AI-enabled robots for improved facility monitoring at the edge

IBM uses AI-enabled robots for improved facility monitoring at the edge

By harnessing the strength of remote sensing, AI and edge computing, IBM, along with partners National Grid and Boston Dynamics are working to monitor facilities autonomously. The new inspectors: AI-enabled robots.

Though inspections can be grueling, dirty and potentially dangerous, they are essential for keeping company equipment and production operations functioning correctly. Despite the risk, many of these inspections are conducted manually. They require workers to enter dangerous, highly electrified areas that are off-limits unless they shut the station down. IBM says that robotics and AI technologies can help minimize the manual labor needed for infrastructure inspections while improving safety.

As mentioned, IBM is working with National Grid, a leader in transmission and distribution operations, to use robots and AI to automate inspection procedures. The robots, powered by the Boston Dynamics’ Spot robot platform, can be used in areas that would otherwise require costly shutdowns.

To tackle this problem, IBM Research devised AI models that equip robots with the capacity to read analog gauges and interpret thermal images. For example, IBM Consulting’s Robotics Solution for Asset Performance integrates with Spot, quickly alerting staff of any issues that need addressing.

“By turning autonomous devices like Spot into roaming edge devices, we can show the value of robots taking on dangerous yet critical tasks to get data vital to business operations. Some research challenges remain, however, including ensuring that models running on these robots are constantly improving and that all the models needed for inspections are managed properly from a distance,” said Mike Murphy, the editor-in-chief of the IBM Research blog.

IBM says that an AI model’s accuracy depends heavily on the quality of data used to train it, which can vary daily. To ensure that images are suitable for training and to prevent false positives, they developed an Out-of-Distribution Detection (OOD) system. This technique assesses whether an image is in focus and representative of the intended outcome, for example. Additionally, it can be utilized post-modeling to identify any modifications in the datasets processed by the model.

When a manufacturer must substitute suppliers because of an abrupt supply chain discrepancy, OOD (Object-Oriented Design) comes into play. A perfect illustration is an instance where the substituting parts differ in color from their original counterparts.

“An AI model without OOD would flag the new part as defective and cause a costly error. OOD will reduce the prevalence of indicating these subtle differences as false positives,” explained Murphy.

Leveraging artificial intelligence, Spot allows National Grid to swiftly and thoroughly assess critical gas and electrical facilities. In turn, staff can effectively manage their time and safely inspect dangerous zones.

Dean Berlin, the lead robotics engineer at National Grid, is enthusiastic about the advantages of this technology. He says, “this lets us monitor areas routinely without costly shutdowns.”

To ensure optimal accuracy and maintain the highest performance levels, IBM mentions AI models must be constantly retrained with new data. When a single model adjusts, it’s crucial to replicate these adjustments across every edge device in the organization; otherwise, mistakes in detection can occur and cause harmful false positives. To ease this complexity, organizations should adopt a hub-and-spoke management system.

The hub occurs where the models aggregate, retrain and redeploy to edge devices, which act as individual spokes. The spokes can send data back to the hub, allowing AI models to update in real-time with new information gathered from the field. This helps ensure that all edge devices in the system are running the latest model version and that users can quickly identify new defects or anomalies.

“Research’s advancements in AI and edge computing are bringing new possibilities to solve clients’ problems cost-effectively, with low touch ease of use, and high impact to productivity and safety improvements,” concluded Murphy.

Last month, IBM unveiled a trio of new Natural Language Processing (NLP) libraries, broadening its Artificial Intelligence software portfolio. IBM’s expansion helps its ecosystem partners and clients create cost-effective AI applications across a multi-cloud environment. The company said this advancement could also open up potential uses for language processing on edge devices.

Article Topics

 |   |   |   |   |   | 

Comments

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.

Edge Analysis & Opinion

Featured Company

Edge Market Reports

Explaining Edge Computing

Edge White Papers

Edge Computing Events

Latest News