Reinforcement Learning Group Homepage

Reinforcement Learning (RL) is a method to enable an autonomous computer system to learn from experience by receiving rewards and punishments for performed actions (i.e., a carrot and stick approach). An RL agent is therefore learning by trial and error while exploring the environment, a learning model that is ideally suited for many agent tasks, e.g. robotics. Nevertheless, if the environment is very complex, and there are many actions which can be performed in each state, then learning by exploration can take too much time to be of practical use. This scaling-up problem is one of the major challenges to RL techniques, despite their general appeal and high application potential.

The RL group at York is studying methods to overcome the scaling problem, specifically Knowledge-Based Reinforcement Learning. In addition, and complementing this principal research, we are working on applications of Reinforcement Learning, ranging from network security to distributed mobile sensor management.