On ML estimation for automatic RSS-based indoor localization

We consider the problem of RSS-based indoor localization with Maximum Likelihood (ML) estimation techniques in low-cost Wireless Sensor Networks (WSN). In the perspective of fully automated methods, we consider the problem of channel and position estimation as coupled problems. We compare via simulations the approaches of separate and joint ML estimation, plus a third method based on multi-lateration. We find that channel estimation via simple linear regression combined with ML localization has the potential to achieve good accuracy while keeping a very low level of computational and implementation complexity. We also find that in 3D localization the vertical error on the z-axis is considerably larger than the horizontal error on the xy-plane. This is due to the limited vertical offset that can be imposed to anchor beacons in “flat” buildings where the height is considerably smaller than the horizontal dimensions.