Spectrum sharing for unlicensed bands

TLDR

The study investigates spectrum sharing among multiple systems in an unlicensed band, examining both cooperative and selfish scenarios to assess how cooperation affects performance. The authors analyze cooperative power allocation under Gaussian signaling and interference‑as‑noise assumptions, then model non‑cooperative interactions with one‑shot and repeated game theory to explore outcomes and reputation‑based incentives. They find that Pareto‑efficient rates are achievable with piece‑wise constant power, frequency‑division multiplexing is optimal under strong interference, and that extending to repeated games mitigates inefficiencies, yielding only small performance loss and tight bounds on non‑cooperative performance.

Abstract

We study a spectrum sharing problem in an unlicensed band where multiple systems coexist and interfere with each other. We first analyze a cooperative setting where all the systems collaborate to achieve a common goal. Under the assumptions that the systems communicate with Gaussian signals and treat interference as noise, we study the structure of the optimal power allocations. We show that any Pareto efficient vector of rates can be achieved with piece-wise constant power allocations. Moreover, if a strong interference condition among all the systems is satisfied, we show that frequency division multiplexing is optimal. We then consider a non-cooperative situation, where the systems act in a selfish and rational way, and investigate how the lack of cooperation can affect performance. Using game theory, we first analyze the possible outcomes of a one shot game, and observe that in many cases an inefficient solution results. We show that by extending the game definition to that of a repeated game, the possibility of building reputations and applying punishments allows to enlarge the set of achievable rates. We present examples that show that in many cases, the performance loss due to lack of cooperation is small. We also provide a converse theorem that proves that our results are tight and quantify the best achievable performance in a non-cooperative scenario