On the impact of the observation strategy in a POMDP-based framework for spectrum selection

Dynamic Spectrum Access, based on the Cognitive Radio (CR) paradigm, is a promising solution to improve the currently inefficient spectrum utilization. In this respect, this paper deals with the spectrum selection problem when a number of radio links has to be established in a CR network. A novel strategy based on a Partially Observable Markov Decision Process (POMDP) is proposed combining partial observations of the interference state together with a statistical characterization of the interference dynamics. In this context, the use of an efficient observation strategy is a key element to account for the trade-off between achieved performance and measurement requirements. For that purpose, the aim of the paper is to propose and evaluate different observation policies for the POMDP-based spectrum selection framework. Results show that the proposed framework is able to achieve very similar performance than a strategy operating under full knowledge of the interference state requiring much less associated signaling.