Antenna Selection for uplink CoMP in dense small-cell clusters

We consider a small-cell cluster consisting of N receive radio-heads connected to a Remote Central Processor (RCP), which jointly decodes K users. Linear filtering across receive antennas is assumed, where the filters are estimated from pilot symbols according to a Least Squares Error (LSE) criterion. Given a limited number of pilot symbols, the Antenna Selection (AS) problem is to determine the subset of radio-heads to include in the filter to maximize the user´s Signal-to-Interference-plus-Noise Ratio (SINR). We introduce L₁ and L₂-norm constraints on the linear filter and show that there can be a substantial performance loss relative to an exhaustive search over the filter order. We then introduce a weighted-norm constraint, which biases the antenna selection towards those with higher path-gain. The optimized weighting profile is characterized numerically, and the performance is observed to be close to exhaustive search. Finally, we present an algorithm for sequentially adding antennas according to decreasing path-gain, which incurs a modest performance loss relative to exhaustive search.