Multi-User Opportunistic Spectrum Access Using Reinforcement Learning

This paper studies the channel exploration problem for the distributed opportunistic spectrum access (D-OSA) system, where multiple secondary users (SUs) sequentially sense multiple licensed channels and utilize one of idle channel. However, channel sensing order can affect the system performance seriously. When using a better sensing order, the SU can find faster a free channel with high quality and the less collisions among SUs can happen. In this paper, we propose a mechanism using reinforcement learning to find dynamically out a sensing order for improving the system performance. In the proposed mechanism, the interactions among SUs are considered. Simulation results are provided to show the effectiveness of the proposed mechanism and the significant improvement of the system performance.