Section: New Results

Wireless sensor networks

Participants : Nizar Bouabdallah, Fatma Bouabdallah-Othman, Mario Rivero, Bruno Sericola.

Wireless sensor networks (WSNs) can be viewed as a particular case of wireless mesh networks (WMNs), where designers have to cope with the limited power, buffering and processing capacities of the sensor nodes. Despite these limitations, the growing capabilities of these tiny motes, which consist of sensing, data processing, and communicating, enables the deployment of reliable WSNs based on the cooperative effort of a large number of sensor nodes.

In contrast to the traditional networks aim to achieve high QoS levels, sensor network protocols focus primarily on power conservation, because of the limited capacity of the sensor nodes' batteries. However, this should be accomplished while respecting certain constraints on the information reliability and reporting delays. The physical phenomenon should be indeed reliably detected and estimated from the collective information provided by sensor nodes, in order to be able to initiate right actions.

Achieving these two opposite requirements, i.e., the trade-off between energy conservation and information reliability, is the key driver of our work on WSNs [9] .

To reduce the transmission of redundant information while respecting the QoS application requirements, we proposed to profit from the natural temporal and spatial correlation among the observations of the densely deployed sensor nodes.

For instance, continuous monitoring applications require periodic refreshed data information at the sink from the sensor nodes. Profiting from spatial correlation to asleep redundant nodes or using data aggregation is usually not desired with such applications, where the network is designed to provide a continuous tracking of the individual sensor nodes' measurements. The sink node is usually interested in the separate measurements of sensor nodes, which may be used to control or test the behavior of different components of a new product. To date, this entails the need of the sensor nodes to transmit continuously in a periodic fashion to the sink, which may lead to excessive energy consumption.

To reduce the transmission of redundant information without affecting the application requirements, we proposed to profit from the natural temporal correlation among the successive observations of each sensor node. Specifically, we proposed to perform smarter data reporting by avoiding the transmission of non relevant data information [11] . By relevant data we refer to data that contains different information from the previous data transmitted by the same sensor.

It has been verified that our scheme can allow for an improvement in the network lifetime while ensuring the continuous monitoring task. The gain greatly depends on the rate of variation of the phenomenon that the sensors are monitoring [58] .

Considering continuous monitoring applications, we also proposed to further improve the network lifetime by balancing efficiently the energy consumption within the WSN.

The routing protocols should indeed avoid the energy depletion of nodes with naturally higher load, a typical issue in conventional routing schemes such as with MTE (minimum total energy) routing. The MTE routing consists in finding the route that minimizes the total consumed energy between any pair of source and destination nodes. Nevertheless, routing always through the path with the minimum energy consumption, will deplete quickly the energy of the sensor nodes contained therein, causing thus a premature death of the WSN.

A perfect routing protocol would hence drain energy slowly and uniformly among nodes, leading to the death of all the sensors nearly at the same time. Typically, an ideal routing protocol would spread efficiently the traffic inside the network and avoid the fast drain of sensor nodes with natural high energy consumption.

To achieve this, we proposed balancing the energy consumption throughout the network by sending the traffic generated by each sensor node through multiple paths instead of forwarding always through the same path [12] . The problem consists then in determining the set of routes to be used by each sensor node and the associated weights (i.e., the routing configuration) that maximize the network lifetime. As a main contribution of this work, we showed that by efficiently balancing the traffic inside the network, significant energy savings up to 15 % can be achieved compared to the basic routing protocols.

So far, we have described different solutions to reduce the energy consumption in WSNs at different layers: physical (transmission range adjustment), MAC (sleep schedule, correlation-based schemes) and network layers (load balancing). While these protocols may achieve very high performance; in essence, they have not been jointly designed to maximize the overall network performance, specifically to minimize the energy expenditure. The main threat is that the gain achieved at one layer can be ruined at the other layers. The cross-layer design stands out thus as an attractive solution to enable further energy conservation and to cope with the relative inefficiency of traditional layered protocol architectures.

In view of this, we considered a cross layer optimization of the routing and the link layers. To do so, we extended the load balancing routing strategy to work jointly with a MAC level optimization strategy [37] .