ABSTRACT
This chapter summarizes some of the central ideas of finite-set statistics (FISST), a systematic, unified approach to multisensor–multitarget detection, tracking, identification, and information fusion. It compares the "random finite set" (RFS) paradigm of finite-set statistics with the conventional multitarget tracking approach: report-to-track association (MTA). The single-target Bayes filter is usually implemented using moment approximations. In a single-sensor, single-target system, the modeling of the sensor begins with a measurement function ηk+1(X). This indicates that the sensor will collect the measurement ηk+1(X) at time tk+1 if a target with state x is present. In multitarget systems, besides individual target motions one must also model target disappearance and target appearance. In the multitarget case, one must address not only sensor noise but also the possibility that targets may not be detected and that measurements may be due to clutter rather than targets. The multisensor–multitarget recursive Bayes filter is computationally intractable in general.
