Engineering of Sensor Network Structure for Dependable Fusion

: The primary objective of this research is to develop the theory and operation of heterogeneous sensor networks that can provide a desired quality of sensor fusion for creating actionable situation awareness. This goal is being achieved by developing (i) Mathematically rigorous and novel language theoretic sensor data representation and multi-level heterogeneous sensor fusion techniques that require substantially less sensing and communication resources as compared to conventional techniques, and (ii) Fusion-driven dynamic control and adaptation of heterogeneous sensor networks. In addition our research also involves experimental validation of the individual theoretical research problems as well as integrated research. For this purpose, we have created a sensor network test bed consisting of a sensor network simulator integrated with real sensor nodes and real sensor networks. This test bed has been successfully used to test the Heterogeneous Dynamic Space Time Clustering (HDSTC) for target tracking. The HDSTC also integrates research ideas from all the MURI team members. Major innovations of this year, outlined in following sections of this report, have been: (i) contextual semantic reasoning, learning and adaptation, making use of influence diagrams and dynamic decision networks; (ii) exploitation of cross-modal sensor dependencies; (iii) semantic fusion for upper layer control; and (iv) complete coverage of search area for a single robot. This research has led to formal techniques for multi-level fusion of heterogeneous sensor data and have furthered efforts to design engineered sensor networks whose structure is simultaneously adaptive, near optimal and resilient to events caused by either the sensed environment or the inherent network behaviors. The outcomes of this research when incorporated into real DoD sensor systems will lead to systems capable of robust context-adaptive and dependable surveillance with minimal human dependence.