Beijing University of Posts and Telecommunications
This paper proposes the separate-SMDP approximation technique to solve the dimensionality problem of the Semi-Markov Decision Process (SMDP) model for Radio Resource Management (RRM). The proposed separate-SMDP model approximates the original complex SMDP Linear Programming (LP) problem by building a series of simple SMDP LP problems. The estimated policy results are obtained by solving these LP problems in a distributed manner. Related algorithms are proposed to support the model in actual HWN RRM process. This technique not only greatly decreases the computational complexity of SMDP and solves its dimensionality problem, but also adapts well to actual RRM and provides a new horizon for utilizing SMDP.