Wireless sensor networks (WSNs) play a major role in industrial environments for data gathering (convergecast). Among the industrial requirements, we can name a few like 1) determinism and bounded convergecast latencies, 2) throughput and 3) robustness against interferences. The classical IEEE 802.15.4 that has been designed for low power lossy networks (LLNs) partially meets these requirements. That is why the IEEE 802.15.4e MAC amendment has been proposed recently. This amendment combines a slotted medium access with a channel hopping (i.e. Time Slotted Channel Hopping TSCH). The MAC layer orchestrates the medium accesses of nodes according to a given schedule. Nevertheless, this amendment does not specify how this schedule is computed. We propose a distributed joint time slot and channel assignment, called $Wave$ for data gathering in LLNs. This schedule targets minimized data convergecast delays by reducing the number of slots assigned to nodes. Moreover, $Wave$ ensures the absence of conflicting transmissions in the schedule provided. In such a schedule, a node is awake only during its slots and the slots of its children in the convergecast routing graph. Thus, energy efficiency is ensured. We describe in details the functioning of $Wave$, highlighting its features (e.g. support of heterogeneous traffic, support of a sink equipped with multiple interfaces) and properties in terms of worst case delays and buffer size. We discuss its features with regard to a centralized scheduling algorithm like $TMCP$ and a distributed one like $DeTAS$. Simulation results show the good performance of $Wave$ compared to $TMCP$. Since in an industrial environment, several routing graphs can coexist, we study how $Wave$ supports this coexistence.

Scheduling in an IEEE802.15.4e TSCH(Time Slotted Channel Hopping 6TiSCH) low-power wireless network can be done in a centralized or distributed way. When using centralized scheduling, a scheduler installs a communication schedule into the network. This can be done in a standards-based way using CoAP. In this study, we compute the number of packets and the latency this takes, on real-world examples. The result is that the cost is very high using today's standards, much higher than when using an ad-hoc solution such as OCARI. We conclude by making recommendations to drastically reduce the number of messages and improve the efficiency of the standardized approach.

This is a joint work with Telecom SudParis: Anis Laouiti.

We are witnessing the deployment of many wireless sensor networks in various application domains such as pollution detection in the environment, intruder detection at home, preventive maintenance in industrial process, monitoring of temporary industrial worksites, damage assessment after a disaster.... Wireless sensor networks are deployed to monitor physical phenomena. The accuracy of the information collected depends on the position of sensor nodes. These positions must meet the application requirements in terms of coverage and connectivity, which therefore requires the use of deployment algorithms. We distinguish two cases: firstly when the nodes are autonomous, and secondly when they are static and the deployment is assisted by mobile robots. In both cases, this deployment must not only meet the application's coverage and connectivity requirements, but must also minimize the number of sensors needed while satisfying various constraints (e.g. obstacles, energy, fault-tolerant connectivity). We distinguished two cases: autonomous and mobile wireless sensor nodes on the one hand, and static wireless sensor nodes on the other hand.

We propose a distributed and optimized deployment of mobile and autonomous sensor nodes to ensure full coverage of the 2D-area considered, as well as network connectivity. With the full coverage of the area, any event occurring in this area is detected by at least one sensor node. In addition, the connectivity ensures that this event is reported to the sink in charge of analyzing the data gathered from the sensors and acting according to these data. This distributed algorithm, called OA-DVFA, can run in an unknown area with obstacles dicovered dynamically. We distinguish two types of obstacles: the transparent ones like ponds in outdoor environment, or tables in an indoor site that only prevent the location of sensor nodes inside them; whereas the opaque obstacles like walls or trees prevent the sensing by causing the existence of hidden zones behind them: such zones may remain uncovered. Opaque obstacles are much more complex to handle than transparent ones and require the deployment of additional sensors to eliminate coverage holes. OA-DVFA is based on virtual forces to obtain a fast spreading of sensor nodes and uses a virtual grid to stop node oscillations and save energy by making sleep redundant nodes. It automatically detects when the maximum area coverage is reached.

We also considered 3D volumes and proposed an algorithm, called 3D-DVFA, also based on virtual forces, to ensure full coverage of 3D volumes and ensure network connectivity. This is a joint work with Nadya Boufares from ENSI, Tunisia. Since applications of such 3D deployments may be limited, we focus on 3D surface covering, where the objective is to deploy wireless sensor nodes on a 3D-surface (e.g. a mountain) to ensure full area coverage and network connectivity. To reach this goal we propose 3D-DVFA-SC, a distributed deployment algorithm based on virtual forces strategy to move sensor nodes.

This is a joint work with Telecom SudParis: Anis Laouiti.

Autonomous deployment may be expensive when the number of mobile sensor nodes is very high. In this case, an assisted deployment may be necessary: the nodes' positions being pre-computed and given to mobile robots that place a static sensor at each position. In order to reduce both the energy consumed by the robots, their exposure time to a hostile environment, as well as the time at which the wireless network becomes operational, the optimal tour of robots is this minimizing the delay. This delay must take into account not only the time needed by the robot to travel the tour distance but also the time spent in the rotations performed by the robot each time it changes its direction. This problem is called the Multiple Robot Deploying Sensor nodes problem, in short MRDS. We first show how this problem differs from the well-known traveling salesman problem. We adopt two approaches to optimize the deployment duration. The first one is based on game theory to optimize the length of the tours of two robots (TRDS), and the second is based on a multi-objective optimization, for multiple robots (MRDS). The objectives to be met are: optimizing the duration of the longest tour, balancing the durations of the robot tours and minimizing the number of robots used, while bypassing obstacles.

The TRDS problem is modeled as a non-cooperative game with two players representing the mobile robots, these robots compete for the selection of the sensor nodes to deploy. Each robots tends to maximize its utility function.

We then propose an integer linear program formulation of the MRDS problem. We propose various algorithms relevant to iterative improvement by exchanging tour edges, genetic approach and hybridization. The solutions provided by these algorithms are compared and their closeness to the optimal is evaluated in various configurations.

The objective of this work is to propose an optimal deployment and distributed packet scheduling of multi-sink Wireless Sensors networks (WNSs). We start by computing the optimal deployment of sinks for a given maximum number of hops between nodes and sinks. We also propose an optimal distributed packet scheduling in order to estimate the minimum energy consumption. We consider the energy consumed due to reporting, forwarding and overhearing. In contrast to reporting and forwarding, the energy used in overhearing is difficult to estimate because it is dependent on the packet scheduling. In this case, we determine the lower-bound of overhearing, based on an optimal distributed packet scheduling formulation. We also propose another estimation of the lower-bound in order to simulate non interfering parallel transmissions which is more tractable in large networks. We note that overhearing largely predominates in energy consumption. A large part of the optimizations and computations carried out in this work are obtained using ILP (Integer Linear Programming) formalization.

Sensor networks are often used to collect data from the environment where they are located. These data can then be transmitted regularly to a special node called a sink, which can be fixed or mobile. For critical data (like military or medical data), it is important that sinks and simple sensors can mutually authenticate so as to avoid data to be collected and/or accessed by fake nodes. For some applications, the collection frequency can be very high. As a result, the authentication mechanism used between a node and a sink must be fast and efficient both in terms of calculation time and energy consumption. This is especially important for nodes which computing capabilities and battery lifetime are very low. Moreover, an extra effort has been done to develop alternative solutions to secure, authenticate, and ensure the confidentiality of sensors, and the distribution of keys in the sensor network. Specific researches have also been conducted for large-scale sensors. At present, we work on an exchange protocol between sensors and sinks based on low-cost shifts and xor operations.

This work is a collaboration with Mérouane Debbah (Supelec, France).

The objective of this work is to propose a framework for massive MIMO cooperative communications for Wireless Sensor Networks. Our main objective is to analyze the performances of the deployment of a large number of sensors. This deployment should cope with a high demand for real time monitoring and should also take into account energy consumption. We have assumed a communication protocol with two phases: an initial training period followed by a second transmit period. The first period allows the sensors to estimate the channel state and the objective of the second period is to transmit the data sensed. We start analyzing the impact of the time devoted to each period. We study the throughput obtained with respect to the number of sensors when there is one sink. We also compute the optimal number of sinks with respect to the energy spent for different values of sensors. This work is a first step to establish a complete framework to study energy efficient Wireless Sensor Networks where the sensors collaborate to send information to a sink. Currently, we are exploring the multi-hop case.

This is a joint work with Hanen Idoudi and Leila Saidane from ENSI, Tunisia.

The unlicensed spectrum bands become overcrowded causing an increased level of interference for current wireless sensor nodes. Cognitive Radio Sensor Networks (CRSNs) overcome this problem by allowing sensor nodes to access opportunistically the underutilized licensed spectrum bands. The sink assigns the spectrum holes to the secondary users (SUs). Therefore, it must rely on reliable information about the spectrum holes to protect the primary users (PUs). We focused on the MultiChannel Time Slot Assignment problem in CRSN and tackled this problem: first at the level of a cluster (i.e. Intra-cluster multichannel scheduling), second at the level of several clusters coexisting in the same CRSN (i.e. inter-cluster multichannel scheduling).

In 2013, we proposed an Opportunistic centralized TIme slot assignment in COgnitive Radio sensor networks (OTICOR) for the Intra-cluster multichannel scheduling. OTICOR differs from the existing schemes in its ability to allow non-interfering cognitive sensors to access the same channel and time slot pair. OTICOR takes advantages of spatial reuse, multichannel communication and multiple radio interfaces of the sink. We proved through simulations that a smaller schedule length improves the throughput. Applying OTICOR, we showed that, even in the presence of several $PUs$, the average throughput granted to $SUs$ remains important. We also showed how to get the best performances of OTICOR when the channel occupancy by $PUs$ is known.

In 2014, we extended this Intra-cluster multichannel scheduling algorithm by proposing two ways for the sink to determine the available channels and alert the SUs if an unexpected activity of PU occurs. Our objective is to design an algorithm able to detect the unexpected presence of PUs in the multi-hop network while maximizing the throughput. If the estimation of PU presence is accurate, a channel sensing at the beginning of the slotframe is sufficient. This algorithm, called Frame-ICMS (Frame Intra-Cluster Multichannel Scheduling), takes advantage of the slots dedicated to the control period by allowing noninterfering cognitive sensors to access the control/data channel and time slot pair. We showed through simulations that using the control period also for data transmission minimizes the schedule length and maximizes the throughput. However, if the estimation of PU presence is not accurate, channel sensing should be done before each slot. We proposed the Slot-ICMS algorithm.

In 2015, we focused on inter-cluster multichannel scheduling algorithm. We considered the coexistence of different clusters in a same CSRN, each cluster having an intra-cluster multichannel scheduling algorithm. Our goal is to obtain a better scalability without loosing the properties provided by OTICOR:

However, the co-existence of several clusters in the same environment may lead to conflicts in the allocation of time slots and channels between these clusters. To avoid inter-cluster collisions, we donot require that different clusters use different channels, we assign distinct channels only to nodes having one-hop neighbors out of their cluster. Once the problem of inter-cluster collision is avoided, each cluster head schedules the transmissions of its members independently. This whole solution exhibits good performances as shown by the simulation results.

This work is a joint work with Dominique Barth (University of Versailles-Saint-Quentin). To guarantee an efficient and dynamic management of network resources and services we intend to use a powerful mathematical tool: prediction and learning from prediction. Prediction will be concerned with guessing the evolution of network or network components state, based on knowledge about the past elements and/or other available information. Basically, the prediction problem could be formulated as follows: a forecaster observes the values of one or several metrics giving indications about the network state (generally speaking the network represents the environment). At each time $t$, before the environment reveals the new metric values, the forecaster predicts the new values based on previous observations. Contrary to classical methods where the environment evolution is characterized by stochastic process, we suppose that the environment evolution follows an unspecified mechanism, which could be deterministic, stochastic, or even adaptive to a given behavior. The prediction process should adapt to unpredictable network state changes due to its non-stationary nature. To properly address the adaptivity challenge, a special type of forecasters is used: the experts. These experts analyse the previous environment values, apply their own computation and make their own prediction. The experts predictions are given to the forecaster before the next environment values are revealed. The forecaster can then make its own prediction depending on the experts' "advice". The risk of a prediction may be defined as the value of a loss function measuring the discrepancy between the predicted value and the real environment value. The principal notion to optimize the behavior of the forecasters is the regret, seen as a difference between the forecaster's accumulated loss and that of each expert. To optimize the prediction process means to construct a forecasting strategy that guarantees a small loss with respect to defined experts. Adaptability of the forecaster is reflected in the manner in which it is able to follow the better expert according to the context.

Our purpose is to apply on-line learning strategies to:

In 2015, Nesrine Ben Hassine developped an extraction tool to provide real traces from YouTube. these real traces are used as a learning sample by the different prediction algorithms used.

Nesrine Ben Hassine and Dana Marinca extended their simulation tool developed in Python to integrate:

With these tools, we can now tune parameters of prediction strategies and evaluate them.

To predict the popularity of video contents, expressed as the number of solicitations, we compared three prediction strategies: Single Exponential Smooting (SES), Double Exponential Smoothing (DES) and Basic. The best tuning of each strategy is determined, depending on the considered phase of the solicitation curve. For DES, values of the smoothing factor close to 1 probide the best results. We study the behavior of each strategy within a phase and around a phase change, where a phase is defined as an interval of time during which a measured metric remains relatively stable.

Basic expert makes large errors at the phase change, but it quickly corrects its prediction and it is the expert having the closest prediction to the real value within a phase. DES expert provides also good quality predictions within a phase. Since DES and Basic experts outperform the SES expert, we recommend the use of on the one hand, the best DES expert per phase within a phase and on the other hand, the Basic expert to automatically detect phase changes, because of its better reactivity. This self-learning and prediction method can be applied to optimize resources allocation in service oriented architectures and self-adaptive networks, more precisely for cache management in CDNs.

In Content Delivery Networks (CDNs) where experts predict the number of solicitations of video contents, simulations based on real YouTube traces show that the accuracy of prediction is improved by splitting the video content profile in contiguous phases. A phase is an interval of time during which a measured metric remains relatively stable. The best expert per phase outperforms the best expert on the whole video content profile. Different prediction methods are compared and also different phase change-points detection methods are evaluated:

The goal is to identify the method (or method parameters) minimizing the cumulated discrepancy compared to real solicitations of video contents. The use of this machine learning method allows the Content Delivery Network to self-adapt to users solicitations by caching the most popular contents near the end users. More generally, such method can be applied to decide which contents should be replicated to improve the performance of audio and video applications and maximize the satisfaction degree of users.

We propose two analytically tractable stochastic-geometric models of interference in adhoc networks using pure (non-slotted) Aloha as the medium access. In contrast the slotted model, the interference in pure Aloha may vary during the transmission of a tagged packet. We develop closed form expressions for the Laplace transform of the empirical average of the interference experienced during the transmission of a typical packet. Both models assume a power-law path-loss function with arbitrarily distributed fading and feature configurations of transmitters randomly located in the Euclidean plane according to a Poisson point process. Depending on the model, these configurations vary over time or are static. We apply our analysis of the interference to study the Signal-to- Interference-and-Noise Ratio (SINR) outage probability for a typical transmission in pure Aloha. The results are used to compare the performance of non-slotted Aloha to the slotted one, which has almost exclusively been previously studied in context of wired ad-hoc networks.

We study a stationary Poisson pattern of nodes on a line embedded in an independent planar Poisson field of interfering nodes. Assuming slotted Aloha and the signal-to-interference-and-noise ratio capture condition, with the usual power-law path loss model and Rayleigh fading, we explicitly evaluate several local and end-to-end performance characteristics related to the nearest-neighbor packet relaying on this line, and study their dependence on the model parameters (the density of relaying and interfering nodes, Aloha tuning and the external noise power). Our model can be applied in two cases: the first use is for vehicular ad-hoc networks, where vehicles are randomly located on a straight road. The second use is to study a “typical” route traced in a (general) planar ad-hoc network by some routing mechanism. The approach we have chosen allows us to quantify the non-efficiency of long-distance routing in “pure ad-hoc” networks and evaluate a possible remedy for it in the form of additional “fixed” relaying nodes, called road-side units in a vehicular network. It also allows us to consider a more general field of interfering nodes and study the impact of the clustering of its nodes on the routing performance. As a special case of a field with more clustering than the Poison field, we consider a Poisson-line field of interfering nodes, in which all the nodes are randomly located on random straight lines. In this case our analysis rigorously (in the sense of Palm theory) corresponds to the typical route of this network. The comparison to our basic model reveals a paradox: clustering of interfering nodes decreases the outage probability of a single (typical) transmission on the route, but increases the mean end-to-end delay