Registration in a distributed multi-sensor environment

To gain any benefit from a network of multiple sensors, it is essential that each sensor be correctly integrated into a global frame of reference. This process is termed REGISTRATION. Any error in the global coordinate system introduced through the sensor reports or the nature of the coordinate system itself has the potential to introduce sufficient ambiguity to compromise the utility of the multi-sensor network. Techniques for the solution of registration among two sensors are well established in the literature, but when the network involves a larger number of sensors (N sensor case), techniques for 2 sensor registration cannot be efficiently adapted because of their algorithmic growth rates, insufficient sensor visibility, dissimilar sensor types and the resulting multi modal cost functions. This paper examines sensor registration techniques adapted for networks with a large number of sensors using traditional least square estimation and non traditional evolutionary computation in the context of registering a netted radar system. The analysis is supported by extensive simulation including recorded radar data from a network of distributed radar sensors.