Downlink scheduling in CDMA data networks

We identify optimality properties for scheduling downlink transmissions to data users in CDMA networks. For arbitrary-topology networks, we show that under certain idealizing assumptions it is optimal for a base station to transmit to only one data user at a time. Moreover, for data-only networks, we prove that a base station, when on, should transmit at maximum power for optimality. We use these two properties to obtain a mathematical programming formulation for determining the optimal transmission schedule in linear data-only networks, with time allocations playing the role of decision variables. The optimality conditions imply that there exist (i) subsets of outer users on either side of the cell that should be served when only the neighboring base station on the opposite side is on; (ii) a subset of inner users in the center of the cell that should be served when both neighbors are on; (iii) a subset of users in the intermediate regions that should receive transmissions when both neighbors are off. Exploiting these structural properties, we derive a simple search algorithm for finding the optimal transmission schedule in symmetric scenarios. Numerical experiments illustrate that scheduling achieves significant capacity gains over conventional CDMA.