On the application of the expectation-maximisation algorithm to the relative sensor registration problem

An important prerequisite for successful multisensor integration is that the data from the reporting sensors are transformed to a common reference frame free of systematic or registration bias errors. The relative sensor registration (or grid-locking) process aligns remote data to local data under the assumption that the local data are bias free and that all biases reside with the remote sensor. In this study, an algorithm based on the expectation-maximisation approach is proposed to estimate all the registration errors involved in the grid-locking problem, that is, attitude, measurement and position biases. Its statistical performance is investigated by Monte Carlo simulation and compared with that of a previously derived linear least squares estimator and to the hybrid Cramér-Rao lower bound.