Explainable and Causally Enforced Reinforcement Learning

This PhD project is concerned with developing methods for representing
RL strategies so that they can be interpreted and inspected by human domain
experts to gain confidence in the behaviour they exhibit and maybe even help
explain hidden dynamics and causal relations of the system under consideration.
Building upon the vast research on formal verification and explainable machine
learning, the project will seek to combine model learning, causal structures
and state-of-the-art RL methods to advance the applicability of RL for cyber-
physical systems.