Artificial Intelligence for Control of Environmental Systems

3.0 credits · Undergraduate

The course has the dual objective of introducing students to the control problems characterizing modern environmental infrastructures while exposing them to real-world examples of how modelling and control analytics are used. The course emphasizes the use of artificial intelligence techniques— particularly from reinforcement learning and approximate dynamic programming—alongside more traditional control methods. Specific topics include: (1) dynamic programming and its main extensions (stochastic and approximate DP), (2) policy search for single- and multi-objective problems, (3) deterministic and stochastic model predictive control, (4) model-order reduction and its application to large-scale control problems. Emphasis will be given not only to the theorical aspects underpinning the control problems, but also to their practical use. Examples will be drawn primarily from water-related domains—such as reservoir operations, urban water supply, and groundwater management—but will also include other environmental systems, such as forest fire suppression or ecosystem management, to illustrate the generalizability of the methods. Emphasis will be also placed on the role that data analysis plays in the formulation of these control problems. Class periods will be a mixture of in-person lectures and group discussions.