Embedded systems, ubiquitous computing and networked architectures are getting more and more important within our society. System parts are often completely implemented as integrated circuits (SoC = System on chip). Consequently, their complexity and heterogeneity have grown dramatically in the recent past. Moreover, embedded systems are used in environments where parameters are subject to continuous changes. Hence, they have to respond to environmental requirements and changes of their own system parameters in a robust manner. To gain this robustness and to cope with the design methodology, formal measures and metrics are of great importance. Such measures need to be combined with the still increasing requirement for computing performance. The implementation of robust features requires adap-tivity by reconfiguration and parallelism. We will call the corresponding sys-tems adaptive computing systems (ACS). The ACS class offers the opportunity to adapt the whole architecture or parts of the architecture to the changing needs of applications or changing environments. The paper addresses some of these aspects and presents some ideas for mod-elling and designing adaptive computing systems (ACS). Especially measures, metrics and taxonomies for reliability, adaptivity and robustness are analysed and discussed. Robust behaviour of electronic systems will contribute to significantly higher trust of the society in modern technology. Therefore it is of very high eco-nomical relevance for industry and commerce.