Information-Based Decision Making in Sensor Planning and Management
Lunch and beverages will be served
Friday, April 6, 2007 - 12:00pm
Hudson Hall Room 216
Elizabeth Irish email@example.com or Justin Jaworski firstname.lastname@example.org
Semester & Year:
Information-based sensor management aims at making optimal decisions regarding the sensor type, mode, and configuration in view of the sensing objectives. In this talk, an approach is developed for computing two information-theoretic functions, the expected discrimination gain and the expected entropy reduction, to optimize target classification accuracy based on multiple and heterogeneous sensors fusion. The measurement processes are modeled by means of Bayesian networks (BNs). The two objective functions are then applicable for designing decision-support systems that optimize the effectiveness of the sensors’ search sequence while minimizing the measurement cost. New theoretical solutions are presented for the efficient, iterative computations of these objective functions with respect to BN representations of the underlying probability distributions. The Dempster- Shafer fusion rule is embedded in the computations to exploit the complementarity of multiple heterogeneous sensor measurements. A Bayesian network framework is developed to integrate different sensor measurement processes and facilitate the understanding of independent relationships between them. The efficiency of the two objective functions is demonstrated and compared using a landmine detection and classification application where the path planning of a robotic platform and management of GPR sensor mounted on this platform is implemented based on a priori IR sensor data.