Background
Our project aims to improve processes in industries by combining chatbot technology with digital twins. The goal is to create a chatbot that can help people who work in factories find information easily, make better decisions and improve performance. To train the chatbots we should use process data, logged information and documents from the factory. Factories in this context means process plants in pulp and paper, aluminium, chemical, food and beverage, dairy or electrical power industries.
The thesis aims to upgrade traditional methods used in process plants by merging digital twin models and chatbot technology. Recent trends show a growing interest in enhancing Language Foundation Models (LFMs) using Retrieval-Augmented Generation (RAG) frameworks. New projects like AutoGPT, BabyAGI, CAMEL, and Generative Agents are leading this change, with the LangChain community eagerly integrating features from these projects into their system.
Research Focus
The thesis goal is to understand how chatbot technology can work seamlessly with digital twins and assess the benefits that machine learning algorithms can bring to the digital twin model. This approach indicates a significant change, moving towards an "open-book" method where responses are generated based on information gathered from verified data sources.
Activities:
- Literature Review: Carry out a study to include insights from the recent progress related to LLMs and agents
- Prototyping: Create the first version of a chatbot that works well with the digital twin, using ideas from recent LLM projects like BabyAGI, AutoGPT, CAMEL and Metaprompt with RAG.
- Evaluation: Use the Design Science Research process (DSR) to evaluate the digital twin chatbot with stakeholders from one or several companies that we have contacts with.