Section: New Results

Optimization in wireless networks

Participants : Jean-Marie Gorce, Nikolaï Lebedev, Hervé Rivano, Fabrice Valois, Anis Ouni, Virgile Garcia, Cengis Hasan.

In the context of the common lab between Inria and Alcatel Lucent Bell Labs and the ANR Ecoscells project, we work on optimizing wireless networks performance. In one side, we work on distributed algorithms for optimal resource allocation and/or mobile-BS asociation. On the other side, we work on mesh wireless networks optimization.

Multi-cell processing, also called Coordinated Multiple Point (CoMP), is a promising distributed technique that uses neighbor cells' antennas [48] . It is expected to be the part of next generation cellular standards such as LTE-A. Small cell networks in dense urban environments are limited by interferences and CoMP can strongly take advantage of this fact to improve cell-edge users' throughput. The present study introduces a distributed criterion for mobiles to select their optimal set of Base Stations (BS) to perform CoMP, and evaluates the impact of this association on the fairness and the total cell throughput. For that, we use a known theoretical expression for the capacity outage probability of CoMP under Rayleigh fading and evaluate the goodputs of antennas associations. The proposed criterion is used in combination with fair resource allocation to perform a joint double-objective optimization of fairness and efficiency. In [48] , [91] , we provide the analysis of the downlink Coordinated Multiple Point (CoMP) used in conjunction with the basic MIMO. The CoMP is the joint multi-cell transmission from several BS to mobiles, coupled here to an open-loop MIMO technique that does not require the perfect channel state knowledge. We show by simulation, that even for $4\times 4$ MIMO transmission, the CoMP can improve the spectral efficiency for some mobiles, depending on capacity outage requirements.

In [33] , we considered downlink transmission in cellular networks where we target to reduce the energy consumption by switching off some base stations by such a way that the distribution of SINR remains unchanged. This is a mean of green networking in cellular networks in downlink consideration. This paper analyzes for line and plane cases, the gain in power consumption obtained after switching off base stations. By computations we observe that the more the operational cost the more the gain in power consumption.

In [47] , we propose an autonomous radio resource allocation and optimization scheme that chooses the transmit power and precoding vector among codebooks for multiple antennas transmitters to improve spectral and power efficiency and provide user fairness. Network self-optimization is an essential feature for supporting the cell densification in future wireless cellular systems. The proposed self-optimization is inspired by Gibbs sampler. We show that it can be implemented in a distributed manner and nevertheless achieves system-wide optimization which improves network throughput, power utilization efficiency, and overall service fairness. In addition, we extend the work and include power pricing to parametrize and enhance energy efficiency further. Simulation results show that the proposed scheme can outperform today's default modes of operation in network throughput, energy efficiency, and user fairness.

In [55] , we focused on broadband wireless networks based on OFDMA resource management, such as LTE systems. We have investigated two optimization problems, one concerning a backhauling mesh infrastructure while the other is the allocation of modulation and coding, subcarriers and power to users in LTE. Considering a realistic SINR model of the physical layer with a fine tuned power control at each node, a linear programing model using column generation has been developed for computing power efficient schedules with high network capacity for wireless mesh backhauling networks. Correlation between capacity and energy consumption have been analyzed as well as the impact of physical layer parameters - SINR threshold and path-loss exponent. With these models, we highlight that there is no significant tradeoff between capacity and energy when the power consumption of idle nodes is important. We also show that both energy consumption and network capacity are very sensitive to the SINR threshold variation. Finally, simulation results show that compared to classic reuse schemes the proposed approach is able to pack more users into the same bandwidth, decreasing the probability of user outage.

In [62] , we focus on broadband wireless mesh networks like 3GPP LTE-Advanced. This technology is a key enabler for next generation cellular networks which are about to increase by an order of magnitude the capacity provided to users. Such an objective needs a significative densification of cells which requires an efficient backhauling infrastructure. In many urban areas as well as under-developed countries, wireless mesh networking is the only available solution. Besides, economical and environmental concerns require that the energy expenditure of such infrastructure is optimized. We propose a multi-objective analysis of the correlation between capacity and energy consumption of LTE-like wireless mesh networks. We provide a linear programing modeling using column generation for an efficient computation of the Pareto front between these objectives. Based on this model, we observe that there is actually no significant capacity against energy trade-off.

In [63] , broadband wireless mesh networks based on OFDMA resource management are studied considering a realistic SINR model of the physical layer with a fine tuned power control at each node. A linear programing model using column generation leads to compute power efficient schedules with high network capacity. Correlation between capacity and energy consumption is analyzed as well as the impact of physical layer parameters - SINR threshold and path-loss exponent. We highlight that there is no significant tradeoff between capacity and energy when the power consumption of idle nodes is impor- tant. We also show that both energy consumption and network capacity are very sensitive to the SINR threshold variation.