REDESIGN is an Individual Fellowship (IF) of the Marie Skłodowska-Curie Actions (MSCA) funded by the European Commission, which aims to support the career development and training of researchers in all scientific disciplines through international and intersectoral mobility. Research fields are chosen freely by the applicant. The long-term vision of this project is to kick-start the wireless communications paradigm toward developing distributed, self-adaptable, and scalable fog networks and guaranteeing requirements of a multitude of Internet of Things applications including high energy efficiency, high data rate, and high reliability. The goal is to develop wireless fog networks (WFNs) by using network slicing technology and deep learning techniques to integrate ground and drone fog nodes into cellular networks. The project will design WFNs integrated into cellular networks composed of smart cells which are able to continuously sense the network topology and to autonomously learn how to configure network parameters and to slice their own network resources to guarantee the required quality of services by fog nodes. Each cell could also add configuration parameters to enable device-to-device (D2D) and multicast communications among wireless fog nodes themselves. This project enables a shift from centralized core-centric cellular networks toward distributed, software-based, and self-adaptable cell-centric ones to support new fog computing applications. These advancements will make the vision of smarter cities by applying ground and drone fog nodes, almost zeroing the operational expenses related to network configuration and hence revolutionizing the existing business models for fog computing by radically reducing energy and operational costs. This will also facilitate the rise of new fog network markets and applications including healthcare, security, smarter power grids, and disaster management.
Keywords: Fog-based wireless networks, multi-objective resource allocation, intelligent & reconfigurable network slicing and RAN sharing, enabling multicast/Broadcast and D2D messages
The purpose of REDESIGN is to aid the design of new communication techniques for wireless fog‐based networks and to examine and verify their usefulness to improve the network performance in terms of self‐adaptability, energy efficiency, and service latency. During REDESIGN, the application of game theory and machine learning to fog‐based networks were investigated. Resource allocation problems in fog‐based networks, reconfigurable intelligent surface‐based communications, and dense millimeter‐wave communications were studied, and novel efficient algorithms were devised. The efficiency of the proposed algorithms was assessed with the aid of numerical simulations and analytical frameworks.
In particular, the research work was focused on the following specific macro objectives:
a. A practical power consumption model for wireless devices
In the problem of resource allocation in wireless networks, existing works did not consider a practical model for power consumption of the electronic circuits of wireless devices. They usually set a fixed amount for it, while the power consumption of the electronic circuits involved in communication, e.g., radio frequency (RF) and baseband (BB) electronic circuits, are variables and are a function of the transmit power level. In particular, the power consumption of RF electronic circuits is a function of the transmit power level and that of BB is a function of achieved data‐rate. Based on our conducted study, by considering practical models, we have understood that the spectral and energy efficiencies change significantly compared with considering non‐practical models.
b. Joint optimization of service latency and energy consumption in fog‐based networks
Existing works did not investigate the ideal purpose of fog‐based networks in resource allocation that is the joint minimization of both service latency and energy consumption metrics. In fact, a weighted‐sum function does not guarantee joint minimization of both metrics, not even if one optimizes the weight coefficients. A bi‐objective minimization problem was introduced. To tackle it, the weighted Tchebyshev method was used to derive the Pareto boundary between service latency and energy consumption metrics. The Pareto boundary is an ideal tool that can help network designers to design the network based on the available power budget subject to the constraint on service latency. The competition among devices was modelled using the cooperative Nash bargaining game. To compute the unique cooperative Nash equilibrium, an algorithm based on block coordinate descent was developed.
c. Resource allocation in ultra‐dense mm‐Wave communications
The problem of joint transmitter‐receiver beamwidths and power allocation problem in ultradense millimeter‐wave network deployments was investigated. Unlike existing works, the optimization problem was formulated in order to jointly maximize the overall data‐rate and the overall energy‐efficiency (in terms of bits/Joule) while guaranteeing the constraints on the transmit power, minimum and maximum possible beamwidths on each individual wireless terminal. Also, a novel constraint was introduced on transmit beamwidth and power allocation to ensure that the every possible combinations of the transmit beamwidth and power comply to the regulatory limit on the maximum equivalent isotropic radiated power. This is highly important from the practical deployment point of view. To formulate the introduced bi‐objective utility function, the Tchebyshev method was used to derive the Pareto boundary between two objectives. The competition among wireless millimter‐wave transmitters was modelled using non‐cooperative game theory. The uniqueness of Nash equilibrium (NE) was proved. To compute the NE, an algorithm based on deep Q‐learning was developed.
d. Deep Q‐learning for computing Nash equilibrium
In general, there does not exist a specific algebraic method to solve mixed‐strategy best response equations, and solving such problems is typically NP‐hard, when the number of strategies and players is large. The question we sought to answer was: “How to find a mixed strategy Nash equilibrium when there are a large number of strategies and players?” To address the question, we used learning methods, that are able to let the players interact so that they can learn about the game and gather information about each other in the course of playing, to finally end up with computing the mixed‐strategy Nash equilibrium point. We developed a new data efficient deep Q‐learning methodology for learning of Nash equilibria of a non‐cooperative game. The algorithm is a generalization of classical Q‐learning that is parametrized by distributed deep neural networks to give it sufficient flexibility to learn the environment without the need to experience all state‐action pairs.
e. Resource allocation in reconfigurable intelligent surfaces‐based communications
Reconfigurable intelligent surfaces have emerged as a promising technology for future wireless networks. Given that a large number of reflecting elements is typically used and that the surface has no signal processing capabilities, a major challenge is to cope with the overhead that is required to estimate the channel state information and to report the optimized phase shifts to the surface. This issue has not been addressed by previous works, which do not explicitly consider the overhead during the resource allocation phase. To fill this gap, an overhead‐aware resource allocation framework for wireless networks was developed where reconfigurable intelligent surfaces are used to improve communication performance. An overhead model was proposed and incorporated in the expressions of the system rate and energy efficiency, which were then optimized as a function of the phase shifts of the reconfigurable intelligent surface, the transmit and receive filters, the power and bandwidth used for the communication and feedback phases. The bi‐objective maximization of the rate and energy efficiency was investigated too. The proposed framework characterized the tradeoff between optimized radio resource allocation policies and the related overhead in networks with reconfigurable intelligent surfaces.
Fog computing is an emerging domain building a glue between today’s control and automation systems and the cloud, supporting new services for modern communication systems and smart cities. The ER innovated on emerging services based on fog computing, around artificial intelligence technologies, tactile internet, and enhanced 6G connectivity based on the reconfigurable intelligent surfaces.
REDESIGN’s innovations resulted in a holistic solution enabling the dynamic instantiation in the fog computing facilities of highly demanding personal cloud services by third‐parties for end‐users with guaranteed quality of service and quality of experience, anywhere, anytime, while ensuring high network performance and low complexity interfaces. More specifically, the ER had the opportunity to work on cutting edge topics in the area of wireless networks and attended many courses on hot topics related to machine learning and artificial intelligence. During the execution of REDESIGN, the ER received encouraging responses from colleagues in academia and industry, since realizing energy‐efficient wireless networks has major economic and environmental advantages. The scientific results obtained during REDESIGN are already published (one journal paper) or under review.
Alessio Zappone, Marco Di Renzo, Farshad Shams, Xuewen Qian, Mérouane Debbah:
“Overhead‐Aware Design of Reconfigurable Intelligent Surfaces in Smart Radio Environments”,
IEEE Trans. Wirel. Commun. 20(1): 126‐141 (2021).
REDESIGN has received funding from the European Union’s Horizon 2020 research and innovation programme under the Marie Sklodowska-Curie grant agreement No. 789260 https://cordis.europa.eu/project/id/789260.