A Self-Organized Spectrum Assignment Strategy in Next Generation OFDMA Networks Providing Secondary Spectrum Access

Abstract

Abstract-This paper proposes a Self-organized Spectrum Assignment strategy in the context of next generation multicell Orthogonal Frequency Division Multiple Access networks. The proposed strategy is able to dynamically find spectrum assignments per cell depending on the spatial distribution of the users over the scenario, opening new spectrum access opportunities for secondary spectrum usage. Reinforcement Learning methodology has been employed to implement the strategy, which compared with other fixed and dynamic spectrum assignment strategies shows the best tradeoff between spectral efficiency and Qualityof-Service while releases spectrum in large geographical areas.

Learning mechanisms for achieving context awareness and intelligence in Cognitive Radio networks

“…In constrast, Adapt chooses channel in a random manner during channel switching, hence its network performance is lower than RL, AdaptWin and Win. This paper also shows that simple and yet pragmatic learning mechanims such as AdaptWin and Win achieve similar network performance provided by the RL in [1], [3]- [7].…”

“…In [6], RL is applied to identify channels with the most available white spaces at the BS in centralized networks. In [7], RL is applied in spectrum assignment strategy in OFDMA networks in order to improve the PU's performance metrics including spectral efficiency, users' Quality of Service (QoS) satisfaction, and the amount of licensed spectrum to be released to SUs. As a complement to [1], [3]- [7], this paper compares the RL results with other learning mechanisms.…”

A Survey of Self Organisation in Future Cellular Networks

“…[14] and [120]) provided us with principles and paradigms in designing SO systems, although no algorithms for cellular networks were demonstrated. The third generation, which includes contributions in [13], [113], [114] and [100], build on the previous work to present algorithms and solutions termed self organising. Although a careful analysis shows that some solutions in the literature are classic adaptive algorithms, others possess necessary features (scalability, stability and agility) required in any SO solution.…”

“…Thus various modified RL schemes are often used. A number of authors have found RL suitable for developing self optimisation algorithms [114], [126] and [127].…”

“…The authors in [114] designed a self organised spectrum assignment based on reinforcement learning that can help operators to define policies for the system to self organise its spectrum assignment and to achieve the best tradeoff between spectral efficiency and QoS thus creating more spectrum access for opportunistic secondary users. …”