Nowadays, mobile user terminals are usually composed of multiple network interfaces and multiple processors. This means there are usually several independent ways of satisfying the immediate Quality of Service (QoS) expected by the user. Dynamic decision engines are used for selecting the most power-efficient interface and performance states in order to satisfy the QoS while increasing the battery life. Current decision-engines are not generic and need to be mostly re-written when a new processor or radio comes out. This is because there are no generic decision-making framework for real-time power and performance management that would allow creating re-usable bricks that could be shared by multiple decision engines. In this paper, we propose such a framework to ease the implementation of a run-time decision making with real-time requirements and a low memory/CPU footprint to increase the reliability of mobile devices.