Sensor array processing algorithms for source localization often requi
re precise knowledge of the sensor locations, which is often not avail
able in practice. Errors in the sensor locations are known to degrade
severely the performance of these algorithms. This paper proposes iter
ative damped Gauss-Newton maximum likelihood array sensor localization
algorithms using known calibration source signals, Cramer-Rao bound (
CRB) expressions for the sensor location parameters are found for this
problem. Two calibration source signal models are considered. One ass
umes that each signal waveform is known up to a scaling constant and p
hase, while the other requires complete knowledge of the signal wavefo
rm. For uncorrelated, completely known calibration signals, a necessar
y and sufficient condition for the optimal choice of the source bearin
gs that minimizes the sensor position CRB variances is presented. Nume
rical results illustrate that the maximum likelihood calibration algor
ithms attain the CRB asymptotically.