Simulation-based learning and generative AI for infrastructure safety training

Operators in civil infrastructure systems—such as inspectors of heavy-duty vehicles or managers of utility systems—often rely on tacit, experience-based knowledge to make split-second safety decisions. Unfortunately, this type of insight is challenging to capture and transfer to new personnel. This project explores how simulation-based learning environments enhanced with generative AI (GAI) agents can accelerate the training of infrastructure operators by uncovering and sharing this “hidden” expertise.

The student will help design interactive simulations of high-risk scenarios (e.g., system malfunctions, inspection tasks, or maintenance decisions), where both human players and AI agents engage in decision-making tasks. A key feature of this project will be integrating models that represent the latent interdependencies between machine operations—complex cause-effect relationships that experienced workers often understand intuitively. Drawing on reinforcement learning frameworks, the student will work on encoding state-action relationships that help novice learners recognize how one decision may influence multiple system states over time. The AI agent can learn from these dynamics and offer targeted feedback, enhancing the learner’s ability to reason about consequences, risk propagation, and safe action planning.

The student will have the opportunity to explore simulation development, behavior modeling, and adaptive learning system design. This project is ideal for students interested in human-AI teaming, digital twin environments, or engineering education, and offers real-world impact on safety training in the civil infrastructure domain.