Research Projects

LLMs in Multi-Agent Systems

(UW: Danielsson, RIT: Alm)

This individual project will study how a multi-agent-based system can leverage large language models (LLMs) to enhance the negotiation process among diverse resources in the multi-agent system and, when needed, constrain it on human operator processes for generating robust and efficient process plans. Recent advances in LLMs have enabled new ways to interact with automated agents. This will allow for more naturalistic conversations between humans and machines to complete complex planning tasks together. The IRES scholar will investigate frameworks for developing advanced multi-agent systems, that can be used as a key tool for agent negotiations leveraging LLMs. The research outcomes can yield a shift towards better integrated and more flexible manufacturing systems, beyond state-of-the-art interaction methods, which are essential for modern, sustainable, and human-centered manufacturing solutions.

Deep Learning for Machine Recovery in Plug and Produce Production

(UW: Belenki, RIT: Ororbia)

In this project, the student will investigate and compare state-of-the-art deep learning architectures to capture the techniques and actions employed by skilled maintenance engineers when investigating and addressing unexpected stops in AI-controlled Plug and Produce. An important research question is what data from the control system, from existing machine-embedded and externally added sensors, including sensors worn by operators in the shape of gloves and glasses, are most valuable for generating robust and interpretable ML models that can later be queried by operators in need for guidance.

Design of Traceability Signatures for AI-Supported Product Quality Assurance 

(UW: Sikström, RIT: Bailey)

The quality of produced items partially depends on their production history–which different machines in an automated production line have affected the item, and the traceability of this history in the production process. The IRES scholar will design and compare ML-supported quality inspection methods that integrate AI in a generalizable traceability signature process. UW’s Production Technology Center has a sensor infrastructure that will be used for this project. This will include studying how traceable historical features correlate with product quality issues, and how an energy-efficient ML system can filter configurations to provide early warnings or, alternatively, use interpretable graph-based optimization to redesign a production line flexibly and efficiently.

Real-Time ML-based Deviation Control

(UW: Ali, RIT: Shi)

In additive manufacturing, production deviations in real-time remain an open research issue that has yet to be successfully automated. The IRES scholar will extend research at UW on anomaly detection using a hybrid ML approach with both reinforcement and supervised learning, and combine historical and real time data. Furthermore, the project will leverage federated ML to enhance the collaborative aspects of the anomaly detection. Federated ML enables integrating multiple local sources to collectively improve performance, without sharing sensitive data, in a joint approach to address production deviations.

AI in Human Resource Management

(UW: Olsson, RIT: Ororbia)

AI offers new opportunities to manage employees and strengthen organizational performance. Employee occupational safety and well-being in AI-automated environments is an essential social sustainability factor in industry. The IRES scholar will study ML methods to forecast and prevent safety incidents (avoiding accidents, minimizing mental health-related sick-leaves), including in human re-skilling, using data sourced from privacy-preserving infrastructure (shop floor, office environments) that comply with regulation, including EU’s General Data Protection Regulation (GDPR), and its integration with an LLM-based architecture able to learn continuously that generates cues for safety-promoting workplace practices.

AI for Human Resource Knowledge Management

(UW: Eriksson, RIT: Bailey)

AI presents new opportunities for Human Resource Management Systems (HRMS) and Human Resource Information Systems (HRIS), used in employee recruitment, performance evaluation, satisfaction, compensation and benefit analysis, and competence development. An IRES scholar will develop an ML-based recommendation algorithm that simulates scenarios when competencies and workloads are rearranged in an organization, studying how competence metrics compare with metrics based on formally acquired education.