MTHE 472 Optimization and Control of Stochastic Systems Units: 3.50
This course concerns the optimization, control, and stabilization of dynamical systems under probabilistic uncertainty with applications in engineering systems and applied mathematics. Topics include: controlled and control-free Markov chains and stochastic stability; martingale methods for stability and stochastic learning; dynamic programming and optimal control for finite horizons, infinite horizons, and average cost problems; partially observed models, non-linear filtering and Kalman Filtering; linear programming and numerical methods; reinforcement learning and stochastic approximation methods; decentralized stochastic control, and continuous-time stochastic control.
(Lec: 3, Lab: 0, Tut: 0.5)
(Lec: 3, Lab: 0, Tut: 0.5)
Requirements: Prerequisites: MTHE 351 or permission of the instructor
Corequisites:
Exclusions:
Offering Term: W
CEAB Units:
Mathematics 6
Natural Sciences 0
Complementary Studies 0
Engineering Science 18
Engineering Design 18
Offering Faculty: Faculty of Arts and Science
Course Learning Outcomes:
- Understand stochastic stability notions for Markov chains, including recurrence, positive Harris recurrence, and transience.
- Establish the existence and structure of optimal policies for finite horizon, discounted infinite horizon and average cost infinite horizon problems.
- Investigate the applicability of computational or learning theoretic methods for stochastic control problems.
- Understand linear and non-linear filtering theory and their use in engineering systems.
- Arrive at solutions for optimality, near-optimality, or stability in decision making under uncertainty via alternative analytical, numerical or simulation methods.
- Document, present, and critically review a scientific paper, method or algorithm on a subject involving stochastic control theory or applications.