Reducing the bias of a bearings-only TLS target location estimator through geometry translations

Despite its simplicity, the pseudolinear estimator for bearings-only target localization is known to suffer from severe estimation bias. To overcome this problem, the method of total least squares (TLS) can be employed which attempts to correct errors not only in the data vector but also in the measurement matrix. The paper examines the relationship between the target localization geometry and the TLS estimation bias for the target location. The paper shows that translations of the target localization geometry through simple shift and/or rotation operations do affect the bias of the TLS estimator and can be exploited to reduce the TLS estimation bias significantly.