Channel/Beam Handoff Control in Multi-Beam Antenna Based Cognitive Radio Networks

Abstract

A novel spectrum handoff scheme, called feature stacking (FEAST) is proposed to achieve the optimal “channel + beam” handoff (CBH) control in cognitive radio networks (CRNs) with multi-beam smart antennas. FEAST uses the online supervised learning based on the support vector machine to maximize the long-term quality of experience of user data. The spectrum handoff uses the mixed preemptive/non-preemptive M/G/1 queueing model with a discretion rule in each beam to overcome the interruptions from the primary users and to resolve the channel contentions among different classes of secondary users (SUs). A real-time CBH scheme is designed to allow the packets in an interrupted beam of a SU to be detoured through its neighboring beams, depending their available capacity and queue sizes. The proposed scheme adapts to the dynamic channel conditions and performs spectrum decision in time- and space-varying CRN conditions. The simulation results demonstrate the effectiveness of our CBH-based packet detouring scheme, and show that the proposed FEAST-based spectrum decision can adapt to the complex channel conditions and improves the quality of real-time data transmissions compared to the conventional spectrum handoff schemes.