Main impact factor of the project

2013 2014 2015 2016 2017 2018 Total
Best Papers Award 2 2
Patents 1 1 2
Invited Professor 1 1
PhD Mobility 1 2 2 1 1 6
Defended PhD 1 1 4 1 7
International Journals 4 1 2 3 9 20
International Conferences 4 9 10 5 28
National Conferences 3 3 3 9

To see more details about the indicators, please go here.

You can also download the official report here.

Presentation of the thesis

Title PhD Student Advisors
WLAN density and energy efficiency Dareen Shehadeh

(Defended 2017/12/04)

Nicolas Montavont, Alberto Blanc
Delay Tolerant Users: The Solution to End-to-End Energy Efficiency Samantha Gamboa (parially funded)

(Defended 2015/06/06)

Alexander Pelov, Nicolas Montavont
Spectral Efficiency – Energy Efficiency Trade-off in Interference Limited Wireless Networks Ahmad-Mahbubul Alam

(Defended 2017/03/30)

Jean-Yves Baudais, Philippe Mary, Xavier Lagrange
Decision Making for Green Cognitive Networks Navikkumar Modi

(Defended 2017/05/17)

Christophe Moy, Philippe Mary
Jointed PAPR Optimization and Pre-Distrortion technics Ali Cheaito

(Defended 2017/03/10)

Matthieu Crussière, Yves Louet, Jean-François Helard
Study of a New Multi-Carrier Waveform Marwa Chafii

(Defended 2016/10/07

Jacques Palicot, Rémi Gribonval
Cross-Layer Design Analysis for Energy Efficient Wireless Networks Laudin Molina

(Prevision 2018/05)

Alberto Blanc, Nicolas Montavont

Introduction to the results

Wireless networks have been designed for delivering maximal data rate at peak usage resulting in high power consumption from network devices (e.g., base stations (BS), core network) with little or no dependency on the traffic load, as illustrated in the Figure below. However, the network is not always running at its peak and the studies in the TEPN project were focused on techniques allowing the network energy consumption to be proportional to its load: near zero power consumption when there is little or no traffic, and a proportional increase of power w.r.t. the load. We have first identified the elements the consume most of the power. These include the Power Amplifier (PA), cooling, and the BS. We have proposed solutions whose power consumption is indeed proportional to the load and that also reduce significantly the power consumption of the network. These solutions include signal processing for PA linearity, new energy-efficient waveforms, and strategies to turn on and off BSs

Scientific Contributions

There was and there is still a huge need for developing sustainable information technologies that justified the European project FP7 EARTH that investigated the energy efficiency of mobile communications. Given the different scale of funding and resources of the EARTH project, the approach of TEPN -rather than competing with EARTH-  was to tackle the problem under a cognitive angle in its largest sense, i.e., to add advanced processing in the network in order to save energy as much as possible.

All the work done in this project has the following equation as baseline

where NTRX is the number of RF chains, P0 is the power consumption independent from the load, Pout is the power radiated by antennas and pis the potentially varying slope according to the use of PA. The most significant contributions of this project allowed to optimize these terms in order to reduce the energy consumption of the network.

Systemic Approaches

Part of the energy consumption is due to the static power consumption of a BS and can be minimized by switching on, off or asleep some BS depending on the load of the network. This issue has been covered by the PhDs of Ahmad Mahbubul Alam (a), Navikkumar Modi (b), Samantha Gamboa (c), Dareen Shehadeh (d) and Laudin Molina (e). In particular, we studied how the energy efficiency – spectral efficiency (EE-SE) tradeoff behaves in a large scale network and if we can get insights from an engineering perspective. We proposed an analytical expression of EE-SE tradeoff based on Poisson-point processes and large random matrix theory. We found that there exists an optimal transmission power and an optimal number of BS antennas depending on BS and user density ratios that maximises both EE and SE. Due to the combinatorial nature of selecting which BS will stay ON and which one will be switched OFF, we used a Markovian decision process (MDP) with a restless multi-armed bandit strategy. The reinforcement learning approach proposed has been coupled to a transfer learning strategy, and we reached 5% difference from the optimal configuration in terms of EE (during night time, i.e., when there are many possible strategies). In order to study the dynamicity of the system, we introduced a possible delay for users before they are served. Our numerical evaluation, ns-3 simulation, and experimentation in city downtowns showed up to 78% of power reduction for an offered load of 7 erlangs, and up to 45% and 17% for 10 and 14 erlangs respectively. In the experimentation, we also evaluated that only 5-10% of the BS are needed to maintain the network coverage (in case of low traffic demand). More specifically, we evaluated at 30 percentage points the amount of possible additional power reductions when users are put to sleep in period of transition between BS states.

Signal processing

Another part of the energy consumption is due to non linearities of PA, among other things. Ali Cheaito (PhD thesis (g)) and Marwa Chafii (PhD thesis (f)) covered the Peak to Average Power Ratio (PAPR) issue with the objective to increase the PA efficiency and reduce the power consumption. We proposed closed form expressions to predict the Error Vector Magnitude value of a transmission and provided new analytical tools for the overall optimization of transmitters. Only upper-bounds or approximated expressions were available in the literature, which did not bring sufficient insight into the performance and interactions between the various modules of transmitters. The new analytic forms derived during the project allow for the elaboration of a complete set of analytic laws that are expected to help designing and properly setting the different modules of a transmitter among which PAPR reduction methods, predistortion function and various PA models. The provided results show how it is possible to play with many degrees of freedom, address some performance/efficiency/complexity trade-off and evaluate their impact on power consumption.


Taking a different approach, we also worked on the PAPR derivation itself where it has been proved  that the PAPR depends on this modulation structure. Moreover, the behavior of the PAPR regarding the modulation waveforms was analysed and the PAPR reduction problem was formulated as an optimization problem. Furthermore, a necessary condition for designing waveforms with better PAPR than OFDM has been derived. This necessary condition is particularly satisfied by a wavelet basis. Finally, two new adaptive waveforms were proposed, allowing significant gain in terms of PAPR, while keeping the advantages of multicarrier modulations : the first one being based on wavelet packet and the second one on the Fourier Transform.

Comments are closed.