Optimal Spectrum Partitioning and Licensing in Tiered Access under Stochastic Market Models

TL;DR

This paper develops an optimization framework for spectrum partitioning and licensing that maximizes utilization in tiered access markets, using a game-theoretic approach with efficient algorithms and simulations.

Contribution

It introduces a novel two-stage Stackelberg game model for spectrum partitioning and licensing, with an algorithm to find optimal parameters M and P.

Findings

Optimal M and P values maximize spectrum utilization.

The proposed algorithm efficiently solves the game and finds optimal spectrum partitioning.

Numerical simulations validate the effectiveness of the approach.

Abstract

We consider the problem of partitioning a spectrum band into M channels of equal bandwidth, and then further assigning these M channels into P licensed channels and M-P unlicensed channels. Licensed channels can be accessed both for licensed and opportunistic use following a tiered structure which has a higher priority for licensed use. Unlicensed channels can be accessed only for opportunistic use. We address the following question in this paper. Given a market setup, what values of M and P maximize the net spectrum utilization of the spectrum band? While this problem is of fundamental nature, it is highly relevant practically, e.g., in the context of partitioning the recently proposed Citizens Broadband Radio Service band. If M is too high or too low, it may decrease spectrum utilization due to limited channel capacity or due to wastage of channel capacity, respectively. If P is too high (low), it will not incentivize the wireless operators who are primarily interested in unlicensed channels (licensed channels) to join the market. These tradeoffs are captured in our optimization problem which manifests itself as a two-stage Stackelberg game. We design an algorithm to solve the Stackelberg game and hence find the optimal M and P. The algorithm design also involves an efficient Monte Carlo integrator to evaluate the expected value of the involved random variables like spectrum utilization and operators' revenue. We also benchmark our algorithms using numerical simulations.