Physics-based signal processing in the presence of environmental model uncertainties
Physics-based signal processing methods for radar and sonar utilize computational acoustic or electromagnetic models to improve target detection and localization. These techniques are motivated by the fact that using an accurate model of the full multipath wavefield can often theoretically provide large performance. A key difficulty in practically realizing these gains, however, has been the need for sufficiently accurate knowledge of the propagation environment. An approach which offers improved robustness to environmental modeling errors has been to treat uncertain environmental parameters as random variables with known statistics. Two applications of statistical environmental modeling in physics-based signal processing are reviewed: (1) matched-field beamforming with environmental perturbation constraints for passive sonar, and (2) maximum a posteriori multipath track association for skywave over-the-horizon radar. In both applications, real-data results support the use of statistical environmental models to achieve improved robustness.
