The human engineered control architectures have evolved towards classical stabilization (feedback control) and feedforward (trajectory planning) loops that are “put together” to perform the control task. These architectures come in many forms and a general theory is lacking on how to setup such control architectures that can work in a reliable manner across various spatiotemporal scales. A starting point for such theory is the fundamental aspect of what control is, namely the child of two parents: algorithmics and physics. There is no other discipline where both come so close together. Using physics-based and data-driven models we can understand how to interact with the actual physics, using algorithmics we can drive the physics to perform tasks. In this project we aim to develop methodologies that can design control architectures to accommodate for multi-rate dynamics. The coupling between feedback control and feedforward control is studied and how their coupling can be mathematically derived, inversely, starting from known models and control algorithms. We will apply and validate these concepts on robot quadruped tasks and robotic manipulation tasks.