We address the issue of self-adaptation of a sensor network with mobile sensors to better observe and track events of interest in an urban scenario. We present middleware called Event-driven Network Controller (ENC) that resides at every sensor node in the network and is independent of the sensor type. ENC translates the requirements of the application layer (focused on improving of the quality of the heterogeneous data fusion) into messages that are diffused throughout the wireless network with the intent of clustering heterogeneous sensor nodes in the vicinity of an event and dynamically changing the local network topology to meet the fusion needs of the cluster. ENC combines the cluster control and the resource management into a unified framework. We implement ENC in NS-2 and we show its applicability for localizing and tracking a mobile target in an urban scenario using a heterogeneous sensor network, formed by pressure, video, and magnetic sensors.