Optimal Mobile IRS Deployment for Empowered 6G Networks

Abstract

The development of cellular networks is driving the rapid growth of wireless communications. With the advent of the 5th Generation (5G) towards the future of the 6th Generation (6G) dedicated to achieving strong growth in traffic while reducing energy consumption, there is a need to solve the problems facing leveraging of these networks’ advantages and support both operators and mobile users. The main challenges for wireless communications are power consumption, Quality of Service (QoS), and the blind areas of a Non-Line-Of-Sight (NLOS) between mobile users and the Base Station (BS). The Intelligent Reflective Surface (IRS) of a reconfigurable meta material is a promising solution for solving some of the challenges of wireless communications. Additionally, it enhances the QoS of the received signals without the need for a power source to operate. Hence, it does not constitute an additional burden as it consists of passive elements. From the other hand, it provides the perfect solution to cover mobile users in blind areas without the need to deploy extra expensive BSs. In this work, we propose to equip buses by IRS allowing them to act as mobile IRS. These buses will become a relay for the surrounding moving vehicles, represented as taxies in the performance Section of this paper. Practically speaking, not all buses have to be IRS equipped. We propose various approaches for selecting the best buses equipped with IRS. In the first optimization approach, we adapt the classical IRS selections methods used in static context to the mobile case. It uses a Multi Integer Linear Programming (MILP) which gives optimal results but with a very long processing time. Thus, we propose a neural-network to learn the result of the MILP. As an alternative solution, another approach is proposed using a Markov decision problem (MDP) relying on Long Short-Term Memory (LSTM) to predict the positions of the surrounding moving vehicles. It is used to solve the optimization problem with the performance criteria targeted for each session. The performance of the proposed approaches are validated based on bus and taxi dataset for the city of Rome in Italy.

Keywords

• IRS-Bus
• cellular network management
• ML
• MILP
• MDP
• LSTM