Sensor Bias Estimation Based on Ridge Least Trimmed Squares

A robust sensor bias estimation approach, named as the ridge least trimmed squares (RLTS), is proposed. Combing the advantages of ridge regression and least trimmed squares, RLTS can solve the sensor bias estimation problem with the presence of misassociations and ill-conditioning. Simulation results verify the effectiveness of the proposed approach.

I. Introduction

In the distributed multisensor multitarget data fusion system, each sensor transmits its local tracks of multiple targets to the fusion center, which performs associating and fusing the data. However, sensor biases, a kind of systematic biases, can make the target state estimates reported by local trackers deviate away from the true values. If not properly estimated and corrected, sensor biases may cause the final fusion results to be unreliable and even lead to ghost tracks.