Abstract
This letter is concerned with distributed estimation in a Gaussian sensor network operating under a semi-orthogonal multiple-access channel (MAC). In such a semi-orthogonal MAC, all sensors are divided into groups and signals from the sensors in each group are directly combined (opposed to be coherently combined) and then transmitted on one orthogonal channel to a fusion center (FC). The behavior of the average distortion in terms of the number of sensors, namely the scaling law, is analyzed. The analytical derivation and simulation results show that the semi-orthogonal MAC with adaptive sensor grouping can achieve the optimal scaling law while requiring only a small amount of feedback from the FC to the sensors.
