We present work-in-progress to develop a hierarchical-control architecture for distributed autonomous systems. The new architecture comprises (i) a decentralised system-level control loop that partitions the goals of the distributed autonomous system among its components under conservative assumptions, and (ii) component-level control loops responsible for achieving the component sub-goals. The operation of the two control loops is underpinned by formal models that are small enough to enable their verification at runtime. To illustrate the use of our approach, we briefly describe its application to a multi-robot search and rescue mission.