[Paper Review] A Satisfactory Power Control for 5G Self-Organizing Networks

Hajar El Hammouti, Essaïd Sabir+1 more|arXiv (Cornell University)|Jun 25, 2016

Advanced MIMO Systems Optimization20 references3 citations

TL;DR

This paper proposes a satisfaction-based power control mechanism for 5G self-organizing networks to balance quality of service (QoS) and energy efficiency. Using game theory, it introduces the efficient satisfaction equilibrium (ESE) under stationary fading and the long-term satisfaction equilibrium under fast fading, both solvable via distributed learning algorithms, ensuring energy savings without degrading network performance.

ABSTRACT

SmallCells are deployed in order to enhance the network performance by bringing the network closer to the user. However, as the number of low power nodes grows increasingly, the overall energy consumption of the SmallCells base stations cannot be ignored. A relevant amount of energy could be saved through several techniques, especially power control mechanisms. In this paper, we are concerned with energy aware self organizing networks that guarantee a satisfactory performance. We consider satisfaction equilibria, mainly the efficient satisfaction equilibrium (ESE), to ensure a target quality of service (QoS) and save energy. First, we identify conditions of existence and uniqueness of ESE under a stationary channel assumption. We fully characterize the ESE and prove that, whenever it exists, it is a solution of a linear system. Moreover, we define satisfactory Pareto optimality and show that, at the ESE, no player can increase its QoS without degrading the overall performance. Under a fast fading channel assumption, as the robust satisfaction equilibrium solution is very restrictive, we propose an alternative solution namely the long term satisfaction equilibrium, and describe how to reach this solution efficiently. Finally, in order to find satisfactory solution per all users, we propose fully distributed strategic learning schemes based on Banach-Picard, Mann and Bush Mosteller algorithms, and show through simulations their qualitative properties. fully distributed strategic learning schemes based on Banach Picard, Mann and Bush Mosteller algorithms, and show through simulations their qualitative properties.

Motivation & Objective

• To address energy efficiency in ultra-dense 5G networks with increasing numbers of low-power base stations.
• To replace traditional QoS-maximization with satisfaction-based power control, reducing energy consumption without sacrificing user-perceived performance.
• To develop a distributed, energy-efficient power control mechanism that guarantees target QoS levels while minimizing interference and power usage.
• To establish theoretical conditions for existence and uniqueness of the efficient satisfaction equilibrium (ESE) in stationary fading channels.
• To propose a long-term satisfaction equilibrium for fast fading channels, where robust solutions are too restrictive.

Proposed method

• Models the network as a non-cooperative game where base stations choose power levels to meet target QoS while minimizing energy use.
• Introduces the efficient satisfaction equilibrium (ESE) as a solution concept ensuring no player can improve QoS without degrading overall performance.
• Derives the ESE as the unique solution to a linear system of equations under stationary fading, with closed-form expressions for power levels.
• Proposes the long-term satisfaction equilibrium for fast fading, using time-averaged channel statistics to avoid overly conservative robust solutions.
• Employs distributed learning algorithms—Banach-Picard, Mann, and Bush-Mosteller—enabling decentralized convergence to the ESE.
• Proves that at the ESE, the system achieves satisfactory Pareto optimality, meaning no user can improve its QoS without harming others.

Experimental results

Research questions

• RQ1 Under what conditions does the efficient satisfaction equilibrium (ESE) exist and is unique in a 5G self-organizing network with stationary fading?
• RQ2 How can energy efficiency be maximized while guaranteeing a target QoS in ultra-dense 5G networks with small cells?
• RQ3 What alternative equilibrium concept can be used in fast fading environments where robust satisfaction equilibria are too restrictive?
• RQ4 Can distributed learning algorithms converge to the ESE without centralized coordination?
• RQ5 Does the ESE ensure satisfactory Pareto optimality, meaning no unilateral improvement is possible without system-wide degradation?

Key findings

• The ESE exists and is unique when the determinant of the system matrix is non-zero, which holds under specific channel and QoS parameter conditions.
• The ESE is the unique solution to a linear system, and closed-form expressions for power levels are derived using matrix determinant analysis.
• The ESE ensures satisfactory Pareto optimality: no user can increase its QoS without degrading the overall network performance.
• In fast fading scenarios, the robust satisfaction equilibrium is too conservative, so the long-term satisfaction equilibrium is proposed as a more practical alternative.
• The Banach-Picard, Mann, and Bush-Mosteller algorithms converge to the ESE in simulations, demonstrating robust and distributed learning behavior.
• The derived power control solution achieves energy savings by targeting satisfactory QoS levels rather than maximizing performance, reducing unnecessary interference and consumption.