Reciprocity inspired learning for opportunistic spectrum access in cognitive radio networks

Abstract

This paper addresses opportunistic spectrum access (OSA) in non-cooperative cognitive radio networks (CRNs). The sel sh behaviors of the secondary users (SUs) will cause a CRN to collapse. The SUs are thus enabled to build beliefs about how other SUs would respond to their decision makings. The interaction among the SUs is modeled as a stochastic learning process. In this way, each SU can independently learn the behaviors of the competitors, optimize the OSA strategies, and nally achieve the goal of reciprocity. Two learning algorithms are proposed to stabilize the stochastic CRNs, the convergence properties of which are also proven theoretically. Simulation results validate the performance of the proposed results, and show that the achieved system performance outperforms some existing protocols.

Opportunistic Spectrum Access with Discrete Feedback in Unknown and Dynamic Environment：A Multi-agent Learning Approach

“…It is assumed that there are five available channels and ten users. The error packet is 3 10 − and the average received SNR increases from 5 to 15 dB . The simulation result shows that the proposed multi-agent Q-learning algorithm achieves better throughput performance.…”

“…Opportunistic spectrum access (OSA) has become increasingly popular due to its potential to improve the efficiency of spectrum usage [3]- [4]. However, many authors studying the problem of OSA assumed that the wireless environment is static and does not vary with time.…”