I investigate the computational aspects of an agent’s internal, action-oriented representation of its own body—commonly referred to as the body schema. My research focuses on how this model is continuously updated by integrating multimodal sensory information. To explore this dynamic process, I leverage contemporary frameworks in predictive processing and active inference. These frameworks posit that the brain continually generates predictions about incoming sensory signals and minimizes the resulting prediction errors through both perception and action. In my work, I examine how such predictive mechanisms, grounded in hierarchical Bayesian inference, underlie the ongoing creation and refinement of the body schema. I test these ideas by implementing computational models inspired by system-level neural dynamics on in-house developed humanoid platforms, thereby bridging theoretical neuroscience with embodied robotics.