Channel Reconstruction in Intelligent Surface aided Communications

Reconfigurable Intelligent Surface: Design the Channel -- a New Opportunity for Future Wireless Networks

“…A key fact of [64], [65], [63] is that channel estimation is initiated from scratch in that no a priori side-information which can reduce overhead is assumed to be available. In this context, we note that exploiting subspace side-information together with other reconstruction methods can very significantly reduce the training overhead, as demonstrated by the recent results in [66]. Moreover, akin to machine learning based approaches, a location indexed dictionary can be maintained which can provide such side-information (cf.…”

“…We further note that effective algorithms for subspace estimation from low dimensional projections over massive MIMO systems have been proposed in [67]. One attempt to tailor these algorithms to RIS-assisted communications has been made in [66], wherein a particular type of inherent sparsity is exploited.…”

“…We recall that generalized Lloyd algorithms have traditionally been used for vector quantizer design and more recently for massive MIMO analog codebook design [71]. Such algorithms can also be leveraged to construct a set of RIS patterns suitable for training, where finite alphabet constraints on these pattern vectors can be imposed via a proximal distance based approach (see [66] for one such provably convergent algorithm).…”