Building an effective situational picture requires the fusion of potentially heterogeneous information coming from multiple sources. In this work, we propose the architecture of a Resource Management Module (RMM) meant to refine and optimize the performance of a multi-sensor fusion engine. Specifically, the RMM is designed to improve performance of the tracking and classification tasks performed by the engine. Here, attention will be focused on assisting the tracking process, while taking into account the current state of things in the observed environment and the available contextual information.