Secondary Resource Allocation for Opportunistic Spectrum Sharing with IR-HARQ based Primary Users

We propose to address the problem of a secondary resource allocation when a primary Incremental Redundancy Hybrid Automatic Repeat reQuest (IR-HARQ) protocol. The Secondary Users (SUs) intend to use their knowledge of the IR-HARQ protocol to maximize their long-term throughput under a constraint of minimal Primary Users (PUs) throughput. The ACcumulated Mutual Information (ACMI), required to model the primary IR-HARQ protocol, is used to define a Constrained Markov Decision Process (CMDP). The SUs resource allocation is then shown to be a solution of this CMDP. The allocation problem is then considered as an infinite dimensional space linear programming. Solving the dual of this linear programming is similar to solving an unconstrained MDP. A solution is finally given using the Relative Value Iteration (RVI) algorithm.