Adaptive Prediction of Battlefield Signal Propagation and Detection
Abstract
Predictions of signal propagation and sensor system performance aid in understanding the vulnerabilities of friendly forces to detection and tracking by hostile forces, and vice versa. In practice, the accuracy of such predictions is limited by simplifications in the modeling physics and by uncertainties in the environmental inputs to the models. To address this problem, we consider the possibility of rapidly updating initial model predictions as more information on the actual sensor performance becomes available during an operation. Specifically, we propose a Bayesian sequential updating approach that incorporates realistic signal distributions for randomly scattered signals transmitted along a single path or multiple paths. We discuss how, in the Bayesian context, these various distributions represent likelihood functions, which are conveniently paired with their conjugate priors to provide efficient updates to the uncertain signal parameters. We also discuss how the Bayesian updating problem is closely connected to the problem of describing the impact of parametric uncertainties on the signal and noise distributions, and hence upon the receiver operating characteristic (ROC curve, a plot of the probability of detection vs. the probability of false alarm). In particular, we show that parametric uncertainties in quantities such as the mean signal and noise power substantially raise the tails of the distributions, which leads to severely degraded ROC curves. Formulation of automatic target recognition (ATR) algorithms, that use Bayesian updating to help compensate the signature for environmental effects, is also discussed.
Document Details
- Document Type
- Technical Report
- Publication Date
- Nov 02, 2017
- Accession Number
- AD1068846
Entities
People
- Carl R Hart
- Chris L. Pettit
- D. Keith Wilson
- Daniel J. Breton
- Edward T. Nykaza
Organizations
- Engineer Research and Development Center