TY - JOUR
T1 - An Information Elasticity Framework for the Adaptive Matched Filter
AU - Narayanan, Ram M.
AU - Liu, Andrew Z.
AU - Rangaswamy, Muralidhar
N1 - Funding Information:
This work was supported by the US Air Force Office of Scientific Research Contract No. FA9550-17-1-0032.
Publisher Copyright:
© 1965-2011 IEEE.
PY - 2020/12
Y1 - 2020/12
N2 - The adaptive matched filter (AMF) uses a number of training samples observed by the radar to estimate the unknown disturbance covariance matrix of a cell under test. In general, as the number of homogeneous training samples increases, the detection performance of the AMF improves up to a theoretical limit (defined by the performance of a matched filter detector where the disturbance covariance is known). However, radar data are nonhomogeneous in practice. Consequently, a high number of training samples is typically undesirable, since nonhomogeneous training data cause detection performance to suffer. Thus, a decision maker (DM) must consider these tradeoffs when selecting this number of training samples, along with other decision parameters for the AMF. Using the concept of information elasticity, this tradeoff behavior is characterized for decisions that are relevant to a DM. A simple user defined constraint function is proposed, characterizing the relative cost of selecting different decisions. Using a multi-objective optimization (MOO) technique known as compromise programming, information overload is observed, in that increasing the cost of decisions improves performance up to a point, beyond which increasing the cost no longer provides meaningful benefit. Using this framework, a cost-efficient solution is selected.
AB - The adaptive matched filter (AMF) uses a number of training samples observed by the radar to estimate the unknown disturbance covariance matrix of a cell under test. In general, as the number of homogeneous training samples increases, the detection performance of the AMF improves up to a theoretical limit (defined by the performance of a matched filter detector where the disturbance covariance is known). However, radar data are nonhomogeneous in practice. Consequently, a high number of training samples is typically undesirable, since nonhomogeneous training data cause detection performance to suffer. Thus, a decision maker (DM) must consider these tradeoffs when selecting this number of training samples, along with other decision parameters for the AMF. Using the concept of information elasticity, this tradeoff behavior is characterized for decisions that are relevant to a DM. A simple user defined constraint function is proposed, characterizing the relative cost of selecting different decisions. Using a multi-objective optimization (MOO) technique known as compromise programming, information overload is observed, in that increasing the cost of decisions improves performance up to a point, beyond which increasing the cost no longer provides meaningful benefit. Using this framework, a cost-efficient solution is selected.
UR - http://www.scopus.com/inward/record.url?scp=85097768656&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85097768656&partnerID=8YFLogxK
U2 - 10.1109/TAES.2020.3009508
DO - 10.1109/TAES.2020.3009508
M3 - Article
AN - SCOPUS:85097768656
VL - 56
SP - 4916
EP - 4929
JO - IRE Transactions on Aerospace and Navigational Electronics
JF - IRE Transactions on Aerospace and Navigational Electronics
SN - 0018-9251
IS - 6
M1 - 9141395
ER -