Multitarget motion analysis in a DSN

The data association problem in multiple target bearings-only motion analysis is considered. The measurements are target bearings obtained from passive sonars in a distributed sensor network. These measurements, after being transmitted to a processing center, are used to estimate the initial states of targets. The overall problem is formulated using the maximum likelihood principle. A two-step iterative method is proposed to maximize the joint likelihood, a nonlinear least-squares method is used to fit the trajectory estimates to the measurements for given data associations, and the linear assignment algorithm is used to solve the data associations for given trajectory estimates. A stochastic relaxation method based on simulated annealing is proposed to augment the second step so that the global maximum of the likelihood function can be located. Monte Carlo simulations are included to evaluate the joint estimator. The Cramer-Rao lower bound of the initial state estimate is derived in the presence of data association uncertainties