A prediction and scheduling framework in centralized cognitive radio network for energy efficient non-real time communication

摘要

Advent of Internet of Things led to an exponential rise in battery-operated sensors transmitting small non-real time (NRT) data regularly. To this end, this work proposes a framework for centralized cognitive radio network (CRN) that facilitates better spectrum utilization and low-cost opportunistic NRT data transfer with high energy efficiency. The novelty of this framework is to incorporate Hidden Markov Model-based prediction within the traditional cognitive radio sensing-transmission cycle. To minimize the prediction time, we design a Hardware-based Hidden Markov Model engine (H2M2) to be used by the cognitive base station (CBS). CBS exploits the H2M2 engine over high primary user (PU) activity channels to minimize the collisions between PUs and NRT secondary users, thereby reducing the SU energy consumption. However, this is at the cost of reduced throughput. Taking this into account, we propose an Intersensing-Prediction Time Optimization algorithm that identifies the predictable PU activity channels and maximizes the throughput within a PU interference threshold. Furthermore, to minimize the total battery consumption of all the SUs within CRN, a Battery Consumption Minimizing Scheduler is designed at the CBS that efficiently allocates the predictable PU channels to the NRT SUs. By exploiting the unutilized high PU activity channels, the proposed Centralized Scheduling, Sensing and Prediction (CSSP) framework improves the spectral efficiency of the CRN. Exhaustive performance studies show that CSSP outperforms traditional nonpredictive sensing techniques in terms of energy efficiency and interference management. Finally, through a proof of concept, we validate the ability of CSSP framework in enabling NRT communication.