Link Gain Matrix Estimation in Distributed Wireless Networks

In the planning of large-scale distributed wireless networks, the knowledge about the link gain matrix is required to facilitate a better match between system design and channel characteristics. Nevertheless, stochastic path loss models fail to produce an accurate estimation of link gains since they cannot faithfully capture the propagation characteristics of a specific environment. Furthermore, exhaustive measurement of all links in a large-scale network is unrealistic due to the complexity involved. This motivates us to develop an accurate link gain estimation methodology based on a small fraction of measurements. We propose here to partition all transmit-receive links into mutually exclusive categories - with each category defined by the number of obstructions on the direct path - and then derive a separate link gain model (equivalently, pathloss model) for each category. Thus, for each category, one would measure only a subset of all transmit-receive paths (selected on the basis of a "maximum entropy" principle) and, for each such path, record the link gain and distance. A least-squares linear fit of dB gain vs. long-distance for the measured links produces the model for that category, which can then be used to predict the remaining "unmeasured" path gains. We use ray- tracing programs to evaluate this approach for different network geometries, and we discuss its superiority to spatial interpolation methods for determining the network's link gains.