Location-Domain Channel Representation for Estimating Distributed MIMO Channels

We propose a location-domain channel representation and apply it to channel estimation for distributed MIMO (D-MIMO) networks. The approach contrasts with various angle-domain formulations that appear well suited for a collocated large array where the far-field assumption allows an angle-domain channel representation but are ill-suited to geographically distributed arrays (large and small), as in densely deployed 5G cellular networks. Our alternative location-domain representation avoids such difficulties by indexing the multipath channel by the locations, instead of angles, of user terminals or access points. Furthermore, it naturally incorporates 3D surrounding information in the form of a ‘channel database,’ achieving scene-specific SNR gain and ease of machine learning. We demonstrate the efficacy of our proposal through simulation of simple channel estimation algorithms over a narrowband channel. The new channel representation is applied to direct positioning in a companion publication [15].