In response to the ever-increasing complexity of technical systems and the corresponding challenges in their controllability, robustness, and safety, initiatives such as Autonomic Computing, Organic Computing, and Pro-Active Computing emerged within the last two decades. In the core, they all propose to change the way systems are developed by adding a self-adaptation unit on top of productive units in order to adapt the runtime behaviour to changing conditions and unforeseen events. This resulted in various design patterns and architectural concepts that mostly assume some kind of control loop. Those include steps such as (i) perceiving the environmental and internal status by sensors, (ii) analysing this information towards a situation-awareness model, (iii) using these models as a basis for planning the next adaptation steps, and (iv) finally executing these plans, maybe in a coordinated fashion within collectives of systems. In this chapter, we summarise these efforts with an emphasis on the Organic Computing and Autonomic Computing domains. We highlight that the control of large-scale systems that are made of a potentially large set of fully autonomous entities is limited when using these approaches and propose a hybrid control model. This combines system-wide and centralised planning capabilities with fully autonomous local behaviour of entities. The main impact is on how systems decide on their current behaviour and how this impacts their long-term learning behaviour. We use platooning as a running example and mention further applications sharing the same characteristics.