MKM 511E - Special Topics in Mechatr.Eng.

Course Objectives

Reinforcement Learning (RL) is a powerful paradigm where an agent learns to take
sequential decisions in a complex environment in order to accomplish a goal. RL
has been applied successfully in a myriad of problems in various fields such as
robotics, games, finance and healthcare. In the maritime domain, applications
cover autonomous vessels, collision avoidance, navigation and maritime traffic
management. This course will provide a solid foundation to the participants in the
field of reinforcement learning.
During the introduction to RL, the course will cover main concepts such as Markov
Decision Processes, value functions and optimality. The course will then touch
upon Dynamic Programming, but mainly focus on sample-based learning methods
with function approximations. The use of deep learning techniques in RL (deep RL)
will be the central topic of the course. The course project will be about heading
keeping of ships in waves in numerical simulations using deep RL

Course Description

Reinforcement Learning (RL) is a powerful paradigm where an agent learns to take
sequential decisions in a complex environment in order to accomplish a goal. RL
has been applied successfully in a myriad of problems in various fields such as
robotics, games, finance and healthcare. In the maritime domain, applications
cover autonomous vessels, collision avoidance, navigation and maritime traffic
management. This course will provide a solid foundation to the participants in the
field of reinforcement learning.
During the introduction to RL, the course will cover main concepts such as Markov
Decision Processes, value functions and optimality. The course will then touch
upon Dynamic Programming, but mainly focus on sample-based learning methods
with function approximations. The use of deep learning techniques in RL (deep RL)
will be the central topic of the course. The course project will be about heading
keeping of ships in waves in numerical simulations using deep RL