Utilization of OIM for Measurement Selection in Multistatic Target Tracking

Abstract

The sensor management problem can be expressed as obtaining the state estimation with desired accuracy by utilizing the resources effectively. In the literature, there are two principal approaches to this problem, namely task-driven and information driven sensor management. Performance metrics for both task-driven and information driven sensor management frameworks suffer from the heavy computational burden due to the evaluation of expectations or are available only in simulation. In this paper, the Observed Information Matrix (OIM), which is widely used in statistical practice as a surrogate for the Fisher Information Matrix (FIM) in difficult problems, has been proposed as a metric that can be used in sensor management. Recursive computation of OIM has been derived for the cases with linear and nonlinear system dynamics corrupted with additive Gaussian noise. The usefulness of OIM in sensor selection in multistatic target tracking has been demonstrated via simulations.

Keywords

• Sensor management
• observed information matrix
• multistatic sensor networks
• particle filter