The field of Embodied Artificial Intelligence highlights the significance of embodiment and situatedness. Agents sense and act within the sensori-motor loop, which defines the context of learning including their morphology. The development of cognitive capabilities cannot be understood detached from this context. In this project, we study the way in which learning principles for embodied agents shape behavioural patterns that reflect morphology and environmental constraints. Particular behaviours of interest are exploration and body coordination. The ultimate goal of the project is to develop an information-theoretic understanding of the sensori-motor loop. Causal aspects, such as causal effects of acting on sensing, and their role in the context of autonomy and control play a central role to this end.