Bias Estimation for Distributed Radars in 3D

An important property of a multisensor data fusion scheme in a target tracking scenario is to be able to improve tracking by adding more sensors. This property can be difficult to implement because of unknown sensor measurement biases. Not accounting for these may lead to more false tracks and poor performance. It is desirable to find all of these biases from measurements from common targets. We consider a network of 3D sensors observing targets which are moving at constant speed, heading and altitude. The measurements are subject to biases due to uncertain sensor alignment and location, and biases from sensor imperfections. The observability of these biases are discussed, and most importantly it is shown that absolute, as opposed to relative, alignment biases can be estimated when taking into account the curvature of the Earth. Simulations show that the presented model is not satisfactory to estimate the biases, and a new model is proposed.