Asymptotically Optimal Cross-Layer Schemes for Relay Networks with Short-Term and Long-Term Constraints

Convex optimization and dual decomposition have been successfully used to design cross-layer resource allocation algorithms for cellular access networks. However, less effort has been devoted to design optimal algorithms for systems equipped with relay stations. Presence of relay stations renders the design of the access schemes more difficult and requires consideration of additional constraints. The present paper relies on a sum-utility constrained maximization framework to design cross-layer algorithms that guarantee diverse quality of service (QoS) and consider different forwarding strategies at the relay stations. One of the main challenges in the design is the joint consideration of both long-term (elastic) and short-term (real-time) constraints. Such constraints account for diverse delay QoS requirements and relay forwarding strategies. A two-step methodology is proposed to efficiently deal with this challenge. Specifically, for each time instant it applies: a) an approximate online method to estimate the multipliers for the long-term constraints and the corresponding primal variables (resources), and b) a classical iterative method to calculate the multipliers for the short-term constraints and the corresponding primal variables. Our approach incurs an arbitrarily small loss of optimality, and can accommodate both static and fading channels.