Joint Optimization of Detection and Tracking in Adaptive Radar Systems - Advances in Heuristic Signal Processing and Applications - page 139

approximately 64 times more efficient than the iterative search algorithms. Note that
a very crude stopping tolerance usually obtains an arbitrary point near the mid-point
of the initial search interval rather than the true maximum point. This behavior can
be expected to impact performance especially when the maximum lies close to the
interval boundaries, such as the one illustrated in Fig. 6.12 .
6.5 Conclusion and Future Directions
In this chapter, a theoretical and experimental framework has been presented for
joint optimization of detector and tracker subsystems. This exciting problem, which
can be considered within the context of a more general topic of cognitive radar ,is
called tracker-aware detection threshold optimization by the authors.
The problem and possible improvements are presented in non-maneuvering tar-
get tracking domain, particularly for the PDAF case. There were some prior attempts
[ 20 , 22 , 48 ] to this problem, but a comparison of these solution schemes in a uni-
fied framework was not available in the literature until [ 2 ]. After categorizing these
attempts as static and dynamic optimization schemes, a comprehensive compari-
son of these schemes is presented in a unified experimental and theoretical frame-
work. Contrary to expectations, the results concluded that only marginal gains can
be achieved by HYCA-based approaches as compared to MRE-based ones. More-
over, it is observed that there exists a trade-off between having low track loss per-
centage (TLP) and having low steady-state tracking error. 18 The dynamic schemes
are found to be well-located in this trade-off by providing considerably low TLP
and low level of steady-state estimation error. The cost paid for this achievement
is the computational complexity. An approximate closed-form solution proposed in
[ 5 ] partially overcomes this issue for the MRE-based dynamic optimization scheme.
Although the solution is given only for the Neyman-Pearson detector case, in au-
thors' opinion, it can be applied for other practically used detection systems, which
mimic asymptotically the NP detector, such as the Cell Averaging Constant False
Alarm Rate (CA-CFAR) system. Apart from its computational efficiency, the pro-
posed closed-form solution also gives some useful insights into the problem. The
most important implication is that it provides a theoretical lower bound on the de-
tection SNR concerning when the whole tracking system should be switched to the
track before detect (TBD) mode.
For the future research directions, the NSPP algorithms for other tracking filters,
such as NNF or SNF, which are already available in the literature can be applied to
the detection threshold optimization problem. Furthermore, new NSPP algorithms
can also be proposed. Especially, the one for the IMM-PDAF for tracking maneu-
vering targets deserve some attention.
An interesting and also a challenging research direction is for the case of tracking
multiple targets. When two tracks corresponding to two targets overlap, optimal
determination of the detection threshold seems to be a challenging problem.
18 These measures can be seen also as transient vs. steady-state performance, respectively.