## Final Report Summary - DENSE4GREEN (Dense Deployments for Green Networks)

The biggest challenges for next generation wireless communication systems (5G) are to support the ever-growing demands for higher data rates and to ensure a consistent quality of service throughout the entire network. To meet these demands, the network area throughput (in bit/s/km2) needs to increase by a factor of 1000 over the next 15 years. At the same time, the power consumption of the information and communication technology (ICT) industry and the corresponding energy-related pollution are becoming major societal and economical concerns. Credited sources foresee that, to meet such a 1000× higher data traffic without increasing the ICT footprint, new technologies that improve the overall energy efficiency by 1000× must be developed. Hence, higher network area throughput on the one hand and less power consumption on the other are seemingly contradictory 5G requirements. There is a broad consensus that these goals can only be achieved with a substantial network densification.

Two promising technologies towards network densification are small-cell networks and massive MIMO systems. The first technology relies on an ultra dense and irregular operator-deployment of low-cost and low-power base stations, with higher density where the user load is higher. Bringing the base stations and user equipments closer to each other can increase the area throughput, while significantly reducing the radiated signal power. In contrast, the massive MIMO technology aims at evolving the conventional base stations by using arrays with a hundred or more small dipole antennas. This allows for coherent multi-user MIMO transmission where tens of users can be multiplexed in both the uplink and downlink of each cell. It is worth observing that, contrary to what the name “massive” suggests, massive MIMO arrays are rather compact; 160 dual-polarized antennas at 3.7 GHz fit into the form factor of a flat-screen television.

Within this context, the main objectives of the DENSE4GREEN ''Dense deployments for green cellular networks'' project lie in identifying and posing in the right modeling perspective the theoretical performance of dense cellular networks in which the above technologies interplay between each other to improve the spectral efficiency of the network while reducing the power consumption costs. To this end, the research activities were conducted at the Large Networks and Systems Group of CentraleSupelec in close collaboration with Prof. Merouane Debbah and his team (composed of Prof. R. Couillet, Prof. M. Kountouris, Prof E. Bjornson, Prof. A. L. Moustakas, Prof. S. Lasaulce and Dr. A. Zappone) that provided the fellow the right means to acquire deep knowledge of the required mathematical models and tools.

MAIN PROJECT ACHIEVEMENTS

The main project achievements were pursued through the following main research activities and collaborations.

• The strong collaboration and interactions of the fellow with Prof. M. Debbah, Prof E. Bjornson, and Prof. M. Kountouris, were very fruitful to derive a new and refined model of the total power consumption of cellular networks that accounts for the power consumed by different analog components, digital signal processing, backhaul signalling, and other overhead costs (such as cooling and power supply losses). Such a refined was used to formulate an energy efficiency maximization problem under the assumption of a stochastic base station deployment based on Poisson point processes. The energy efficiency was maximized analytically with respect to the density of base stations, the transmit power levels, the number of base station antennas and users per cell, and the pilot reuse factor for channel acquisition. The closed-form expressions provided general guidelines on the optimal operating regimes and exposed the fundamental interplay between the optimization variables, hardware characteristics, and propagation environment. The analysis showed that smaller cells is undoubtable the way towards high energy efficiency, but the positive effect of increasing the base station density saturates when the circuit power dominates over the transmission power. A further leap in energy efficiency can be achieved by adding extra antennas (in a massive MIMO flavour) to multiplex several user equipments per cell.

• The research activity carried out by the fellow in collaboration with Prof. M. Debbah and Prof. A. L. Moustakas were focused on two different problems. The first one was to study the energy consumption dynamics in a single-cell network with a variable number of antennas when user equipments move around and linear processing is used at the base station to guarantee target rates. Recent and standard results from large system analysis and large-deviation theory were used to provide a deterministic approximation of the energy consumption and to study its fluctuations around this value. These results were used to approximate the probability that a battery-powered base station runs out of energy and also to design the cell radius for minimizing the energy consumption per unit area. The second problem was focused on the power consumption in the uplink and downlink of a heterogeneous network in which a massive MIMO macro tier (serving medium-to-high mobility user equipments) is overlaid with a dense tier of small cell access points using a wireless backhaul for traffic. A reverse (inter-tier and intra-tier) time division duplex protocol was proposed to let massive MIMO macro tier simultaneously handle the traffic of macro users and small cells without causing much interference to the overlaid tier. Results from random matrix theory have been used to investigate the impact of user mobility on the power consumption. Analytical and numerical results showed that for a given set of target rates there is a critical value of user mobility beyond which the power of all transmitters rapidly increases (and eventually diverges).

• The collaboration with Prof. Debbah and Prof. Couillet was fruitful for the design of linear precoding algorithms to minimize the total power consumption when different degrees of cooperation among base stations is considered: no cooperation between base stations; only channel state information is shared between base stations; and channel state and data cooperation is envisaged. Stating and proving new results from large-scale random matrix theory allowed the fellow to provide concise approximations of the precoder parameters, the powers needed to ensure target rates and the total transmit power. Such approximations turned out to depend only on the long-term channel attenuations of the users, the relative strength of interference among base stations, the target rates and the quality of the channel estimates. Applied to future dense cellular networks, such results may lead to important insights into the system behavior, especially with respect to the benefits of base station cooperation, target rates, channel state information quality and induced interference. Moreover, they can be used to simulate the network behaviors without to carry out extensive Monte-Carlo simulations.

• In collaboration with Dr. A. Zappone, using tools from fractional programming, sequential convex optimization, and game theory the fellow developed a framework to design centralized and decentralized power control algorithms for energy efficiency maximization in cellular networks. Unlike most previous related works, the provided framework allows to encompass most of the emerging 5G technologies including small-cell networks and massive MIMO systems. Applied to practical networks, the developed framework provides innovative solutions that allow to maximize the energy efficiency of future networks through procedures that can be implemented in a fully decentralized fashion solutions as they only depends on large-scale fading components, which can be accurately estimated and easily exchanged between base stations as they change slowly with time. Within this context, particularly helpful were also the research activities and collaborations pursued together with Prof. S. Lasaulce, Prof. V. Belmega, Dr. P. Mertikopoulos and Dr. I. Stupia, which allowed the fellow to further improve his knowledge on game-theoretic tools and quasi-variational inequality theory for maximal energy efficiency in cellular networks.

POTENTIAL IMPACT OF THE PROJECT

In addition to the scientific results published in international peer-reviewed IEEE journals and conferences, the socio-economic impact achieved by DENSE4GREEN lies in patentable algorithms and techniques, that can be employed by the networks nodes of mobile wireless systems. This includes base stations and user mobile terminals in existing 4G networks, and in future 5G networks, with possible interest from most companies and service providers working in developing next-generation broadband wireless communication systems.

Furthermore, the international network of collaborations initiated between the project team and many research centres and universities across Europe during the project shows potential for future collaborations for European public projects and for some industrial spin-offs in the field of 5G networks.

Two promising technologies towards network densification are small-cell networks and massive MIMO systems. The first technology relies on an ultra dense and irregular operator-deployment of low-cost and low-power base stations, with higher density where the user load is higher. Bringing the base stations and user equipments closer to each other can increase the area throughput, while significantly reducing the radiated signal power. In contrast, the massive MIMO technology aims at evolving the conventional base stations by using arrays with a hundred or more small dipole antennas. This allows for coherent multi-user MIMO transmission where tens of users can be multiplexed in both the uplink and downlink of each cell. It is worth observing that, contrary to what the name “massive” suggests, massive MIMO arrays are rather compact; 160 dual-polarized antennas at 3.7 GHz fit into the form factor of a flat-screen television.

Within this context, the main objectives of the DENSE4GREEN ''Dense deployments for green cellular networks'' project lie in identifying and posing in the right modeling perspective the theoretical performance of dense cellular networks in which the above technologies interplay between each other to improve the spectral efficiency of the network while reducing the power consumption costs. To this end, the research activities were conducted at the Large Networks and Systems Group of CentraleSupelec in close collaboration with Prof. Merouane Debbah and his team (composed of Prof. R. Couillet, Prof. M. Kountouris, Prof E. Bjornson, Prof. A. L. Moustakas, Prof. S. Lasaulce and Dr. A. Zappone) that provided the fellow the right means to acquire deep knowledge of the required mathematical models and tools.

MAIN PROJECT ACHIEVEMENTS

The main project achievements were pursued through the following main research activities and collaborations.

• The strong collaboration and interactions of the fellow with Prof. M. Debbah, Prof E. Bjornson, and Prof. M. Kountouris, were very fruitful to derive a new and refined model of the total power consumption of cellular networks that accounts for the power consumed by different analog components, digital signal processing, backhaul signalling, and other overhead costs (such as cooling and power supply losses). Such a refined was used to formulate an energy efficiency maximization problem under the assumption of a stochastic base station deployment based on Poisson point processes. The energy efficiency was maximized analytically with respect to the density of base stations, the transmit power levels, the number of base station antennas and users per cell, and the pilot reuse factor for channel acquisition. The closed-form expressions provided general guidelines on the optimal operating regimes and exposed the fundamental interplay between the optimization variables, hardware characteristics, and propagation environment. The analysis showed that smaller cells is undoubtable the way towards high energy efficiency, but the positive effect of increasing the base station density saturates when the circuit power dominates over the transmission power. A further leap in energy efficiency can be achieved by adding extra antennas (in a massive MIMO flavour) to multiplex several user equipments per cell.

• The research activity carried out by the fellow in collaboration with Prof. M. Debbah and Prof. A. L. Moustakas were focused on two different problems. The first one was to study the energy consumption dynamics in a single-cell network with a variable number of antennas when user equipments move around and linear processing is used at the base station to guarantee target rates. Recent and standard results from large system analysis and large-deviation theory were used to provide a deterministic approximation of the energy consumption and to study its fluctuations around this value. These results were used to approximate the probability that a battery-powered base station runs out of energy and also to design the cell radius for minimizing the energy consumption per unit area. The second problem was focused on the power consumption in the uplink and downlink of a heterogeneous network in which a massive MIMO macro tier (serving medium-to-high mobility user equipments) is overlaid with a dense tier of small cell access points using a wireless backhaul for traffic. A reverse (inter-tier and intra-tier) time division duplex protocol was proposed to let massive MIMO macro tier simultaneously handle the traffic of macro users and small cells without causing much interference to the overlaid tier. Results from random matrix theory have been used to investigate the impact of user mobility on the power consumption. Analytical and numerical results showed that for a given set of target rates there is a critical value of user mobility beyond which the power of all transmitters rapidly increases (and eventually diverges).

• The collaboration with Prof. Debbah and Prof. Couillet was fruitful for the design of linear precoding algorithms to minimize the total power consumption when different degrees of cooperation among base stations is considered: no cooperation between base stations; only channel state information is shared between base stations; and channel state and data cooperation is envisaged. Stating and proving new results from large-scale random matrix theory allowed the fellow to provide concise approximations of the precoder parameters, the powers needed to ensure target rates and the total transmit power. Such approximations turned out to depend only on the long-term channel attenuations of the users, the relative strength of interference among base stations, the target rates and the quality of the channel estimates. Applied to future dense cellular networks, such results may lead to important insights into the system behavior, especially with respect to the benefits of base station cooperation, target rates, channel state information quality and induced interference. Moreover, they can be used to simulate the network behaviors without to carry out extensive Monte-Carlo simulations.

• In collaboration with Dr. A. Zappone, using tools from fractional programming, sequential convex optimization, and game theory the fellow developed a framework to design centralized and decentralized power control algorithms for energy efficiency maximization in cellular networks. Unlike most previous related works, the provided framework allows to encompass most of the emerging 5G technologies including small-cell networks and massive MIMO systems. Applied to practical networks, the developed framework provides innovative solutions that allow to maximize the energy efficiency of future networks through procedures that can be implemented in a fully decentralized fashion solutions as they only depends on large-scale fading components, which can be accurately estimated and easily exchanged between base stations as they change slowly with time. Within this context, particularly helpful were also the research activities and collaborations pursued together with Prof. S. Lasaulce, Prof. V. Belmega, Dr. P. Mertikopoulos and Dr. I. Stupia, which allowed the fellow to further improve his knowledge on game-theoretic tools and quasi-variational inequality theory for maximal energy efficiency in cellular networks.

POTENTIAL IMPACT OF THE PROJECT

In addition to the scientific results published in international peer-reviewed IEEE journals and conferences, the socio-economic impact achieved by DENSE4GREEN lies in patentable algorithms and techniques, that can be employed by the networks nodes of mobile wireless systems. This includes base stations and user mobile terminals in existing 4G networks, and in future 5G networks, with possible interest from most companies and service providers working in developing next-generation broadband wireless communication systems.

Furthermore, the international network of collaborations initiated between the project team and many research centres and universities across Europe during the project shows potential for future collaborations for European public projects and for some industrial spin-offs in the field of 5G networks.