Information-theoretic sensor management for multimodal sensing
In recent years, sensor management algorithms have been studied for the purpose of providing intelligent, automated control of complex fielded sensor suites in remote sensing applications. In this paper, a framework for sensor management is presented that is based on the information-theoretic formulation of Kastella. The sensor manager searches for N targets using M multimodal sensing platforms and incorporates realistic features such as cost of motion and cost of use for the sensors as well as the availability of useful prior information about the region of interest. In all cases, the performance of the sensor manager is found to be superior to a direct search procedure in which the sensors methodically sweep through the cell grid. Sensitivity of the sensor manager to erroneous prior information is also examined, and the sensor manager performance is found to be robust to reasonable errors in the prior information. Finally, the sensor manager is demonstrated to perform successfully on a set of real landmine data.
