Adaptive Systems Using Generative AI
• LLM invokes the digital twin for xx KPI specification .
• Digital twin retrieves xx KPI specification from design context store .
• LLM gives response with KPI value and specification to user . Technical Service : Technical service is invoked when a problem is detected either by operator or customer . The technical service agent needs to troubleshoot which might involve looking at the KPI values , identifying root cause of problem and perform resolution tasks . Let ’ s look at the steps involved :
• Problem reported : yy KPI is wrong .
• LLM invokes digital twin for yy API & yy specification documents .
• LLM identifies probable cause and resolution steps from the documents .
• LLM gives the response with root cause and resolution steps .
There are many such use cases possible by integrating GenAI with digital twins for industrial systems . However , we will look at how these two use cases can be extended to build adaptive systems .
6.3 ADAPTIVE SYSTEMS WITH GENAI
As shown in the previous use cases , GenAI systems are being used to help in diagnosing problems and answer queries ( Refer to Figure 6-1 ). Leveraging the monitor and analyze modules on top of the digital twin , the modules can now identify whether there is a problem and identify the root cause and possible resolutions .
Based on the Generative Robotics frameworks 21 and the adaptive multi agent frameworks 22 we have the following options to solve “ unknown knowns .”
• Generate a plan of steps with an option tree for the business user to resolve the problem . Using the information provided by the digital twin , the system can accurately pinpoint the root cause . The resolution steps can be generated and validated by a business user . After validation the steps could be executed by the business user or the system to resolve the problem .
• Generate code to resolve the problem . This requires the system to be able to generate code and integrate into the production code base . This is not a recommended option for now as this will require a mature process to handle automatic code integration .
21 https :// corporate-innovation . co / 2022 / 03 / 01 / the-benefits-and-opportunities-of-software-definedvehicles /
22 https :// arxiv . org / abs / 2305.19352 80
March 2024