Active Learning in Cognitive Radar
Abstract
Cognitive radars (CRs) offer advanced sensing capabilities by simultaneously optimizing both transmit and receive processing in response to the changes in the target environment. The advent of CRs present an inherent conundrum - they use environmental information to adapt their transmit/ receive strategies, while the information gleaned about the environment depends, in turn, on the transmit/ receive strategy. An avenue towards solving this conundrum is to investigate the well-known active learning paradigm which presents an iterative procedure involving (i) environment analysis and (ii) probing waveform design. The alternative approach to cast the cognitive radar performance as an optimization problem incorporating the probing waveform leads to highly non-linear formulation without tractable approaches towards its solution. This research will involve the environment analysis estimating the target parameters in the presence of clutter. While traditional clutter modeling has focused on treating the clutter returns as random vectors whose covariance matrices depend on the transmitted waveforms, this research will use available public databases and develop and accurate physics based stochastic transfer function clutter model that facilitates site specific simulations. This method promises to characterize clutter independent of the transmit waveform that simplifies the design of the probing waveform.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Jan 21, 2022
- Source ID
- FA86552217172XX0
Entities
People
- Björn Ottersten
Organizations
- Air Force Office of Scientific Research
- United States Air Force
- University of Luxembourg