An Experimental Investigation of the Bayesian Passive Multi-Target Localization Algorithm

Passive localization using the time-of-arrival of a wideband signal has the potential to achieve high accuracy at a much smaller cost than active localization solutions, which makes it promising for a variety of applications. However, a major challenge in passive localization involves localizing multiple targets that have the same radar signature, where it is difficult to distinguish between reflections from different targets. This difficulty is compounded if there are an unknown number of targets along with multipath propagation and the possible (statistically dependent) blocking of the direct paths (going from the transmitter and reflecting off a target to the receiver). In this scenario, identifying the direct paths corresponding to each target (known as data association) is non-trivial. In an earlier work, some of the authors investigated this problem using a Bayesian estimation framework, where dependent blocking was modeled as a prior, and proposed the \emph{Bayesian multi-target localization algorithm} as a solution. In this work, we analyze the performance of this algorithm using experimental data, obtained from a commercially available localization testbed. In comparison with the cases where blocking is either completely ignored or assumed to be mutually independent across paths, our results show that for a fixed probability of false-alarm, the detection probability is highest when dependent blocking is taken into account.