Self-adaptation is often defined as the ability of systems to alter at runtime their behaviour in response to changes in their environment, capabilities and goals that were unforseen at design time. I will discuss why I believe one interesting route to achieving self-adaptation is to endow systems with the capability of synthesising at runtime discrete event controllers.I will provide an overview of the work we have done in this direction and some of the challenges that remain, covering software architectures, modelling and learning approaches for defining novel discrete event cotrol problems, and algorithms to solve them.