In the context of densifying the mobile network in urbanized areas, it is a common practice to increase the capacity by updating existing cell sectors. This paper studies upgrades consisting in adding new cellular frequencies on existing sectors. Using historical data aggregated at sector level spanning over 18 months, we proposed methods to i) following a sector upgrade, evaluate the trending growth rates of resource availability and of the proportion of resource usage per user of already existing cells, ii) use machine learning to predict this trend given the sector existing frequencies, planned frequencies to be added, previous performance indicators and urban fabrics, iii) prioritize cell deployments by computing a predictive score based on the previous predictions. To evaluate the efficiency of machine learning methods, gradient boosting, random forest and SVM models were compared with a baseline, using several metrics. The baseline model was designed to predict the traffic activity evolution as the mean evolution of the same added and existing configurations seen in the training set. This baseline was beaten by all the models trained with our methods. The ranking made by the best ML models had an average precision higher than the baseline by 0.19, a reciprocal rank higher by 0.11 and normalized discounted cumulative gain higher by 0.12.

Many studies have shown how to process mobile network data with machine learning algorithms to infer land uses or predict mobile traffic behavior. However, there are few, if any, machine learning applications to help deploy new cells. Current forecasting models are designed to predict the traffic of existing cells based on the historical data they produced. In a previous work, we have proposed a method to predict the class activity of a future cell, given its modeled area, demographic and geographic features. In this paper, we extend it by estimating static hour-by-hour median weekly activity aggregated at base station level, based on the same static inputs. We tested several machine learning models including transformers, compared their results and showed that our method beats simple baselines.

Strong spatial relationships exist between call volume and human activities. In this paper, we show machine learning models can classify base station coverage areas into classes of mobile call profile based on areas geographic and demographic properties. We propose the novel approach of taking clustered Call Detail Records (CDR) based time series as ground truth, and passing heterogeneous features as inputs. These features were land use and points of interest retrieved from OpenStreetMap database, and demographic data provided Facebook Research. Together, they allowed the creation of a representation of urban fabric. CDR clusters were then characterized by these features through SHAP model interpretation. Three models were tested in the region of Dakar with CDR coming from D4D-Senegal Challenge. Results exceeded the accuracy of systematic most common class classification and the accuracy of models trained only on population dataset. Additionally, models generalization capacity were evaluated in Thies, with results equalling those of the baseline.

Industrial Partners (IPs) with Mobile Network Operators (MNOs) are extending the mobile network infrastructure with Small Cells (SCs) in order to meet the growing mobile traffic demand. Due to the increasing number of telecommunication market competitors and the scarcity of radio resources, static sharing schemes are no more efficient. New dynamic schemes should be considered to meet both user expectations and economic success. In a crowd networking context, we propose in this work a dynamic radio resource scheme based on combinatorial double auctions. The participants in these auctions are the MNOs considered as buyers and the IPs, providers of SCs, considered as sellers. The commodity is a bundle of radio resources geographically distributed in several sites. This market place is regulated by an auctioneer responsible of pricing process and market clearing. The auctions are launched periodically to increase the elasticity. Theoretical analysis shows that this model satisfies the most important properties of auctions namely, strong balanced budget, individual rationality, incentive compatibility and economic efficiency. The proposed auction model is also simulated to evaluate the social welfare, the utilization percentage of the supplied resources and the service rate of the buyers demand. Results show that this model ensures a positive utility values for both sides with a service rate that exceeds 50\% and an utilization rate that reaches 50\% for some IPs.

The deployment of small cells is one of the technical solutions to meet the challenge of data traffic rise. It reduces the cost of radio access networks. The efficiency of this solution depends on the success of cell planning to mitigate the interference. This work aims to optimize the incremental cell planning scheme that considers a preliminary macro-cell network infrastructure and expands it with new small cells to enhance the coverage and the capacity. In order to reduce the cost of installing new small cells, we consider a crowd networking business model where the Mobile Network Operator retrieves sites to deploy small cells above. We formulate the small cells placement problem as an Integer Linear Programming. It aims to maximize the coverage, ensure the required capacity and mitigate the cross-tier interference. Since the problem is NP-hard with large number of sites, we propose an heuristic to select the optimal sites locations around each macro cell. We tested the proposed heuristic with different simulation scenarios. The proposed Sequential Deployment scenario is better in coverage uniformity among dense and non dense zones whilst the Dense Zones First ensures more optimized selected sites number with better capacity in the dense zones in terms of users. The last scenario proves that higher candidate sites optimize the number of small cells in the non dense areas without compromising the coverage.

Heterogeneous networks (HetNets) which include macrocells with short range small cells proved a better coverage and higher user data rates compared to classical networks. Using the same spectrum as the macrocells, small cells would allow increased spatial reuse of bandwidth. In industrialized countries, the deployment of new small cells by another actor (tier) to cover the outage improves the service with a lower cost. In this paper, we investigate cell association issue in heterogeneous networks composed of small and macrocells operating in the same spectrum. In contrast to the related work, we consider that the small cells are a cellular network belonging to another tier. Hence, there is competitiveness between the two tiers in order to selfishly maximize the gain while respecting the User Equipment (UE) Quality of Service (QoS) requirements. We propose a model based on game theory in order to get the best distribution of user equipments among small and macro base stations.

The wireless industry is preparing the fifth generation wireless systems with several requirements regarding capacity and coverage area. A promising solution to meet this challenge is a densification of the network via a large scale deployment of small cells from different sizes. The advantage here is that small cells are easy to deploy and cost-efficient compared to macro base stations. However, the challenge is to ensure a reliable backhaul links. Mesh routers backhauling is one of the candidate solutions here, since it is easy to deploy, doesn't use wired connections between locations and cost-efficient. Nevertheless, the issue is to deploy routers in the suitable locations that provide the convenient and cheaper power sources. In this paper, we address the problem of mesh routers placement to ensure small cells backhauling. We propose a novel approach that aims to minimize the number of mesh routers and optimize their placement among locations offering the possibility to use solar panels. We first model the problem as an Integer Linear Program (ILP) and since it is hard to find the optimal solution with a large number of mesh routers and small cells, we propose also an heuristic approach called GPMR (Green Placement for Mesh Routers). We study the performance of both, the ILP model and the proposed heuristic GPMR, through simulations scenarios. Results show that GPMR provided performances close to those provided by ILP when the number of small cells is less than 10.We also show that the results provided by our heuristic GPMR follow an evolution close to the one observed in the optimal solutions when the number of small cells is high (more than 20).Thus, our heuristic approach can be used to obtain a near-optimal performance for ultra-dense networks.

Logically centralized network control design enables data-plane and control-plane separation in communication networks. Until today, traditional fault restoration mechanisms based on such designs were restricted to ‘one’ centralized network manager to detect faults and manually reroute traffic around effected areas. Nonetheless, this design is challenged by the overall scalability, restoration latency and convergence delay of the physically centralized controller. As a result, in this paper we argue that one can do a far more efficient network design. Here we propose, model, and design the problem of placing multiple centralized controllers, among multiple network topologies, so as to maximize the resilience and scalability of such networks- within the bounds of convergence delay. We begin with formal model definitions to optimize the controller placement locations. Based on the model, new ILP formulations have been defined. Subsequently, a simple but efficient greedy based heuristics has been proposed to solve the ILP. Performances are quantitatively evaluated and our results show significant reliability improvements in a given topology.

The emergence of Internet access and advanced wireless technologies has its limitations across the globe, i.e. today there exist several rural regions, especially in developing countries, that do not afford Internet connectivity. In this paper, we present a design of a cost-effective wireless network architecture that aims at providing Internet in fix-usage within those countries. We claim that with few design changes to the 3GPP architecture, it is possible to extend Internet connectivity within suburban and rural areas by deploying numerous hotspots based on sharing tasks and revenues with local actors.

Competitive Mobile Network Operators (MNOs) are typically long-known for their shrewdness to conceal their underlying network topology information. Having said this, in this article, we propose a quasi-distributed topology information sharing framework for network operators based on logically centralized controllers. Through our approach, we present a topology information sharing scheme in which two or more MNOs can cooperatively and more importantly, discreetly reveal their topology information for the sake of utilizing the unused available resources of each other, at times of network failure situations. Our approach has been formulated and developed based on a novel key metric to ‘tune’ the amount of information sharing. Based on extensive simulations, we then investigate the impacts of network topology information sharing on the network capacity. The overall feasibility is illustrated through significant numerical results.

In this article we illustrate that, management of network resources can yield notably different performances leading to different restoration behaviors -- for different network topologies. In consideration towards this, we first briefly summarize our earlier work on a novel fault restoration mechanism. The proposed mechanism addresses the problem of finding the minimum bandwidth cooperative route in the backhaul of mobile network operators that is shared among multiple operators. We then present our recent results on the extended performance evaluations of the same, for four different synthetic network topologies carried out using extensive simulations. Through these evaluations, we conclude on the best topology that mobile network operators should construct in order for them to maximally benefit by the proposed cooperative routing scheme to yield optimum performance for their capital expenditure.

In a recent work [1], we proposed LTE4V2X, a novel framework for a centralized vehicular network organization based on 4G LTE network. We demonstrated the efficiency of our framework for an FCD (Floating Car Data) application. Such applications are based on data collected from vehicles (localization, speed, direction, etc.) in order to feed a traffic management server. In the continuity of this work, this new paper presents two extensions of LTE4V2X. The first one is the multi-hop extension, which uses multi-hop communications to deal with areas where there is no LTE coverage (e.g. tunnels). The second extension deals with the adaptation of LTE4V2X framework for a dissemination application that aims to disseminate a specific message in a given geographical area. We analyze the performances of LTE4V2X using NS-3 simulation environment and a realistic highway mobility model. The results show that the multi-hop extension leads to an improvement of LTE4V2X performances, for applications based on both data collection and data dissemination.

In this paper, we provide a new availability analysis of backhaul wireless communication networks which are supported by microwave links which evolved out of sharing between two different mobile network operators (MNOs) based on the Multi-State System (MSS) approach using Discrete-state Continuous-time Markov chain model. Our results show that such a jointly- constructed network was available to meet the required demand of the total system more than 99.8 % of the time.

Vehicular networks face a number of new challenges, particularly due to the extremely dynamic network topology and the large variable number of mobile nodes. To overcome these problems, an effective solution is to organize the network in a way which will facilitate the management tasks and permit to deploy a wide panoply of applications such as urban sensing applications. This paper presents LTE4V2X, a novel framework for a centralized vehicular network organization using LTE. It takes advantage of a centralized architecture around the eNodeB in order to optimize the clusters management and provide better performances. We studied its performances against a decentralized organization protocol for a well known urban sensing application, FCD application. We analyze the performances of LTE4V2X using NS-3 simulation environment and a realistic urban mobility model. We show that it permits performance improvement by lowering the overhead induced by control messages, reducing the FCD packet losses, as well as enhancing the goodput.

In mobile ad hoc networks (MANETs), multihop transmissions rely on nodes collaboration to correctly route and forward packets to their final destination. However, collaboration of intermediate nodes is not guaranteed and non-collaborative nodes may drop packets they are asked to forward. Such behaviour, known as packet dropping, results of nodespsila selfishness or maliciousness and has dramatic effects on networks performances. Most of proposed solutions to counter packet dropping are either based on trust management or multipath routing, but no solution combines both trust management and multipath routing. Adaptive Secured Multipath for Ad hoc networks (ASMA) is a security framework which adapts security to the application requirements; evaluates trustworthy relationship based on localized trust model and efficiently combines multipath routing into trust management. In this paper we study ASMA performance against the most classical and simple packet dropping attack: Black Hole attack. Aiming to evaluate ASMA in dense networks, we use a realistic scenario representing a subway environment which brings interesting features and may be a prolific context for MANET application development. Through simulations we compare the performances of DSR and ASMA associated to DSR in attacked pure MANETs. Simulation results illustrate that ASMA-DSR outperforms DSR under different attack configurations and is totally adapted for dense and large networks.

The recent proliferation of wireless devices extends the scope of mobile ad hoc networks (MANETs) applications beyond the military domain to include civil and commercial application scenarios. Since MANETs are composed of mobile terminals with limited resources, guaranteeing their security remains an unsolved and motivating challenge. Due to advantageous features like infrastructureless and spontaneous deployment, MANETs offer an opportunity to set up temporary, dynamic and local networks for low cost. However, some of their characteristics like unreliability of wireless links, dynamic topology, and absence of underlying infrastructure raise serious problems which become critical when security is also considered. Multipath routing scheme enhances the robustness of routing protocols and offers a mean to mitigate networks mobility impact and node misbehaviors. In this paper we compare the performances of five multipath routing protocols: three node-disjoint multipath routing protocols and two routing protocols based on a Untrusted Node Disjoint (UND) path scheme. Comparisons between these protocols highlight the improvements raised by the UND scheme under different attack configurations.

The next generation of wireless networks is expected to support a wide range of real-time multimedia applications with stringent quality of service (QoS) requirements. In this paper, we study the performance of 802.11b wireless multi-hop networks carrying Voice over IP (VoIP) calls. We compare the performance of the standard proactive ad hoc routing protocol, OLSR, which makes use of the hop count as routing metric, with two modified versions: SP-OLSR (Strongest Path OLSR) and S2P-OLSR (Shortest Strong Path OLSR). Both of them consider the Signal to Interference and Noise Ratio (SINR) to build a reliable topology graph. The cross-layer aspect is employed to compute and push the SINR values from the physical layer to the routing protocol. We implemented the proposed protocols in the Qualnet simulator and performed extensive simulations using a proper traffic model tailored for VoIP traffic in such an environment. Obtained results indicate that the quality of VoIP applications improves significantly using SP-OLSR and S2P-OLSR compared to the standard OLSR, and that S2P-OLSR offers the same performances as SP-OLSR with a lower routing overhead cost. Index Terms— Wireless multi-hop networks, performance eval- uation, VoIP, routing metrics.

We consider the problem of mitigating interference and improving network capacity in wireless multi-hop networks. An ongoing aim of our research is to design a routing metric which is cognizant of interference. To address this issue, and based on the measurement of the received signal strengths, we propose a 2-Hop interference Estimation AlgoRithm (2-HEAR). With the use of the received signal level, a node can calculate the signal to interference plus noise ratio (SINR) of the links to its neighbors. The calculated SINR is used to infer the packet error rate (PER) between a node and each of its first tier interfering nodes set. Then, the residual capacity at a given node is estimated using the calculated PERs. Based on the capacity estimation analysis, a new routing metric, EBC (Estimated Balanced Capacity), is proposed. EBC uses a cost function at the aim of load-balancing between the different flows within the network. Extensive simulations shows that EBC improves tremendously the network capacity and also enhances the VoIP calls quality.