Fundamental Bounds on Information Fusion with Focus on Waveform-based Intent Detection and Avoidance.

摘要

In radar systems one may have a priori knowledge of the scene or its statistics. The supported research sought to exploit this additional knowledge available to a radar to improve its performance along two main lines: (1) past radar waveform returns and knowledge of scene statistics allows a radar to adapt its subsequent waveforms to these extra sources of information, and (2) knowledge of the geometry of the radar scene may allow one to exploit multi path reflections to improve radar performance such as target detection and localization. In our first direction, it is known that adapting radar waveforms to best extract information exploits prior scene knowledge - for example knowledge of multipath - and improves performance. How to 'best' design and adaptively select waveforms, however, remains an open question. Central to answering this is how to properly incorporate feedback, or 'close'the loop. In a more theoretical direction, we have proposed a novel, information theoretically optimal metric which properly incorporates feedback which will allow for the more efficient and effective scheduling of radar waveforms and shown the resulting designed or scheduled waveforms in simulations.