Module overview
Unlocking the potential of AI requires autonomous systems that can independently learn to make good decisions. Deep reinforcement learning is a powerful framework for this, with particular relevance to robotics, enabling robots to autonomously acquire and refine complex skills through interactions with the world without manually programming the desired behaviour. This module will cover the principles and current advances of the field of deep reinforcement learning and its application to robotic skill learning. It begins by introducing the theoretical foundations of reinforcement learning and the role of deep neural networks as function approximators. The module then explores important methods and state-of-the-art approaches focused on learning to control robotic agents for game playing or performing physical tasks. This culminates in practical applications of deep reinforcement learning for key robotic control problems.