MAP-PF multitarget tracking with propagation modeling uncertainties

We consider the problem of multiple target tracking using the maximum a posteriori penalty function (MAP-PF) technique under propagation modeling uncertainties. The propagation model, which describes how the target state is perceived at the sensor, is assumed to be a function of one or more parameters that are not known exactly, but are characterized statistically by a probability distribution. Under this model, the penalized sensor likelihood function used in the measurement step of the standard MAP-PF algorithm and the measurement likelihood function used in the information update become robust penalized sensor and measurement likelihood functions, averaged over the propagation parameter distribution. Performance is demonstrated on a simulated underwater acoustic scenario with uncertainty in array orientation.