Software in mobile and ubiquitous computing scenarios has to cope with a highly dynamic and heterogeneous environment. For tapping the environment’s full potential, software should be able to dynamically adapt on demand in a platform- and language-independent manner (e.g. by migration). Current systems have only limited support for the requested dynamics and heterogeneity. Thus, the development of such adaptive applications is still a highly complex and thus error-prone task. In the context of this thesis, the SAMProc middleware is designed. SAMProc supports software developers in creating such adaptive applications and thus eases development. In the SAMProc approach, applications are able to dynamically change the location and to adapt the provided interface, the current state and the implementation in use to the particular local execution context. Thereby, applications are transparently and permanently addressable. This allows continuous interaction with other applications. For easing application development SAMProc uses a model-driven approach. For this purpose, this thesis introduces the novel concept of a self-adaptive mobile process (SAMProc), which allows an abstract specification of the life cycle as well as distribution aspects of an adaptive application: First, software developers model the SAMProc; then, a code generator automatically maps this description to the SAMProc infrastructure and creates the necessary adaptation code. This thesis presents all parts of the SAMProc middleware in detail. Important contributions are an architectural design pattern for dynamic application adaptation and migration in a heterogeneous environment and novel infrastructure services supporting software in such environments, such as a service for dynamic management of code, a generic context service and a novel particularly lightweight Web service container. Innovative example applications complete this work by demonstrating the potential of the SAMProc middleware.