Resource allocation in interference limited systems

In this work we address two problems which deal with resource allocation. The first deals with scheduling and power control over the digital subscriber line (DSL) system. The second deals with scheduling data transmissions on the downlink of UMTS WCDMA. The former is a wired system, while the latter is wireless. We study these systems, abstract out their essential features, and formulate problems based upon these abstractions. This allows our solutions to be applied to any system that exhibits these properties.; Our first problem is motivated by the DSL system. DSL technology leverages existing telephone infrastructure to provide high-bandwidth internet access. The dominant problem in DSL is interference. Although each subscriber has his own dedicated line from the central office to the customer premises, multiple lines are grouped together into cable bundles. Close proximity to neighbouring lines results in crosstalk. We address two shortcomings of currently employed transmission schemes for DSL.; First, we observe that existing schemes use a single static power allocation for each user. However, we note that when interference is high, significant gains in system performance are obtained by employing time-divisioning between highly interfering users. Next, we note that current schemes treat all users equally. However, DSL subscribers fall into a wide variety of classes, some expecting higher data rates than others. We present a formulation that accounts for different data rate requirements, while introducing the possibility of time-divisioning between users. We show that this scheme leads to appreciable gains in system performance.; Our second problem deals with scheduling multiple users over a channel with time-varying capacity. This problem is motivated by the shared data channel of UNITS WCDMA. The system model is such that it is possible for the base station to accurately predict the achievable data rate in each slot. As in the case of DSL, different users have different data rate requirements. The scheduler must decide which user to serve in each slot. We present a solution that makes the optimal use of channel conditions. This solution yields significant performance gains over solutions that do not account for channel conditions while making scheduling decisions.