LL Logo



Post-STAP Detection
Performance Under
Non-Ideal Conditions

Steven Smith
MIT Lincoln Laboratory
244 Wood Street
Lexington, MA 02173-9108
tel: (781) 981-3106
email: stsmith@ll.mit.edu

Adaptive filtering affects radar system detection performance in a variety of ways. If enough ideal assumptions are made, namely independent, identically distributed Gaussian samples used in the AMF GLRT, or ACE detectors, then closed form detection statistics are obtainable. However, it is easy to find examples where each or all of these assumptions fails to hold. For example, pre-Doppler STAP coherently combine adaptive outputs determined from overlapping data sets, i.e., the assumption of independence is invalid. Furthermore, clutter is notoriously inhomogeneous in range; if training samples are drawn from that dimension, then the assumption of identically distributed data is invalid (up to a scale factor). Finally, all closed-form analysis of standard adaptive detectors involves the Wiener equation w = R\v; however, diagonal loading is oftentimes used for sidelobe control, which seems impervious to closed-form analysis. These oft encountered issues raise the following interesting questions: Do pre- and post-Doppler STAP algorithms exhibit the same detection performance given fixed sample support and degrees of freedom? What can be said about the detection performance of the AMF, GLRT, ACE, and CLED detectors in the presence of an inhomogeneous scale factor in range? How does diagonal loading affect the detection performance of pre- and post-Doppler STAP algorithms? This talk will answer these questions using Monte Carlo analysis, possible alternate simplifying assumptions, and comparison to previous closed-form analysis.



LL Logo Disclaimer

Direct comments and questions to: webmaster@ll.mit.edu

MIT Lincoln Laboratory. All rights reserved.