Transmitter optimization for the network MIMO downlink with finite-alphabet and QoS constraints

We consider a network MIMO downlink communication where base stations (BSs) can cooperate on transmission of data symbols to their intended mobile stations (MSs). We assume the data symbols for each user come from a finite alphabet. We formulate a novel power optimization problem for the downlink communication starting with the binary inputs: we take a symbol detection point of view, impose minimum Symbol Error Probability (SEP) as Quality-of-Service (QoS) constraints at the MSs, and show that the resulting problem is convex. We first tackle the problem via centralized approaches and then propose a distributed algorithm to solve the power optimization problem, where only local communication among the base stations is required. We compare with state of the art linear precoding techniques, and demonstrate significant reduction in consumed transmit power attained by the proposed schemes. We also provide an extension of the problem to general finite-alphabet inputs.