Two techniques are presented in this paper that address the issue of hosting applications in a ubiquitous computing space: (1) context-aware component deployment and (2) self-adaptive component deployment configurations. The former focusses on deploying an application, which is composed of fine-grained components, in a mobile distributed computing environment based on resource information that is spread through context-data. The latter focusses on letting the component deployment configuration adapt itself when the resources in the environment change due to device mobility. Both techniques are solved using a single mechanism that is based on an extended version of collaborative reinforcement learning (CRL) in the design of our component deployment framework. In order to find local optimal solutions for deploying a component-based application in a distributed resource constrained environment, the CRL method balances the resource consumption of components on several nodes that are in each other?s proximity.