The emerging applications using sensor network technologies constitute a new trend requiring several different devices to work together and partly autonomously. However, the integration and coordination of heterogeneous sensors in these emerging systems is still a challenge, especially when the target application scenario is susceptible to constant changes. The setup and adaptation of these systems are challenging, considering that their nodes are distributed and must respect operational constraints, such as energy consumption. Due to the dynamicity of the scenarios in which they are deployed and the nature of their operations these systems require autonomous decisions that have to be taken without any human operator intervention. This paper presents a reflective middleware that supports heterogeneous sensor networks deployed in dynamic scenarios. This middleware presents specific handling of users’ requirements by representing them as missions that must be accomplished by the network. These missions are then translated to network parameters and activities that are distributed to network nodes by means of the autonomous reasoning that takes into account network nodes’ capabilities and environment conditions. A multi-agent framework is proposed to provide the necessary support for missions’ dissemination and for the required reasoning to allow autonomous behaviour related to mission allocation and network adaptation.