3.0

Average Course Rating

Learning-enabled decision-making systems—driven by advances in AI and machine learning—are reshaping engineering fields such as mobility, robotics, energy, and healthcare. In many of these settings, multiple decision-makers repeatedly interact and adapt their behavior based on environmental feedback. These decision-makers can be autonomous algorithms, LLM-based agents, embodied AI systems, humans, or institutions. Their actions influence one another via physical interaction, competition for shared resources, or information exchange. As a result, overall system behavior emerges from the coupled learning and decision processes of many agents rather than from any single centralized controller. This course focuses on rigorous methods to model, design, and analyze Learning-Enabled Multi-Agent Systems (LEMAS). It pursues four main objectives: Build mathematical models for multi-agent decision making that capture interaction, adaptation, and feedback. Study algorithmic approaches to multi-agent learning and provide performance guarantees where possible. Characterize system-level outcomes—efficiency, fairness, and stability—that arise from interacting LEMAS. Explore how adaptive incentives and market mechanisms can steer collective behavior toward desirable outcomes. The course integrates techniques from AI and machine learning, control and dynamical systems, algorithmic game theory, and optimization. Emphasis is on mathematical modeling, rigorous analysis, and algorithmic design. Applications are used to motivate modeling choices and demonstrate practical relevance. This is a graduate-level, research-oriented course. Expected preparation includes undergraduate-level AI/ML, probability, optimization, linear algebra and differential equations, and multivariable calculus. Prior exposure to game theory or reinforcement learning is useful but not required. Students should be prepared for significant self-study and engagement with current research literature.