Hopfield Neural Network Based Uplink/Downlink Transmission Order Optimization for Dynamic Indoor TDD Femtocells

Abstract

The Uplink/Downlink transmission mode or Transmission Order (TO) optimization has recently appeared as a new optimization domain in radio resource management. Such optimization is a combinatorics problem and requires good heuristic algorithm to be approximately solved within short time for the dynamic radio environment. This paper shows how the TO optimization problem in Time Division Duplex (TDD) indoor femtocells can be formulated and solved by the Hopfield Neural Network (HNN) based TO schedulers. Both centralized and distributed versions are analyzed in the context of indoor femtocells. We also examine proposed TO schedulers’ system performance in TDD indoor femtocells environment by extensive simulation campaigns. Our simulation results for a large 3-story building including 120 femtocells show that (i) the indoor femtocell system performance is improved up to 13 to 20 percent by the proposed HNN schedulers depending on the number of femtocells, (ii) the proposed TO schedulers converge within the first few epochs. (iii) The performance of the proposed schedulers are justified by a time-consuming but a thorough Genetic Algorithm Scheduler.

Keywords

• Dynamic-TDD
• femtocell networks
• hopfield neural network
• transmission order optimization
• genetic algorithm