Section: New Results

Data Collection in Multi-hop Networks

Participants: Jin Cui, Jad Oueis, Hervé Rivano, Razvan Stanica, Fabrice Valois.

Data Aggregation in Wireless Sensor Networks

Wireless Sensor Networks (WSNs) have been regarded as an emerging and promising field in both academia and industry. Currently, such networks are deployed due to their unique properties, such as self-organization and ease of deployment. However, there are still some technical challenges needed to be addressed, such as energy and network capacity constraints. Data aggregation, as a fundamental solution, processes information at sensor level as a useful digest, and only transmits the digest to the sink. The energy and capacity consumptions are reduced due to less data packets transmission.

As a key category of data aggregation, aggregation function, solving how to aggregate information at sensor level, was investigated in the Ph.D. thesis of Jin Cui [1]. In this work, we make four main contributions: firstly, we propose two new networking-oriented metrics to evaluate the performance of aggregation function: aggregation ratio and packet size coefficient. Aggregation ratio is used to measure the energy saving by data aggregation, and packet size coefficient allows to evaluate the network capacity change due to data aggregation. Using these metrics, we confirm that data aggregation saves energy and capacity whatever the routing or MAC protocol is used. Secondly, to reduce the impact of sensitive raw data, we propose a data-independent aggregation method which benefits from similar data evolution and achieves better recovered fidelity. This solution, named Simba, is detailed in [15] as well. Thirdly, a property-independent aggregation function is proposed to adapt the dynamic data variations. Comparing to other functions, our proposal can fit the latest raw data better and achieve real adaptability without assumption about the application and the network topology. Finally, considering a given application, a target accuracy, we classify the forecasting aggregation functions by their performance. The networking-oriented metrics are used to measure the function performance, and a Markov Decision Process is used to compute them. Dataset characterization and classification framework are also presented to guide researcher and engineer to select an appropriate functions under specific requirements.

Energy Harvesting in Wireless Sensor Networks

Energy harvesting capabilities are challenging our understanding of wireless sensor networks by adding recharging capacity to sensor nodes. This has a significant impact on the communication paradigm, as networking mechanisms can benefit from these potentially infinite renewable energy sources. In [23], we study photovoltaic energy harvesting in wireless sensor networks, by building a harvesting analytical model, linking three components: the environment, the battery, and the application. Given information on two of the components, limits on the third one can be determined. To test this model, we adopt several use cases with various indoor and outdoor locations, battery types, and application requirements. Results show that, for predefined application parameters, we are able to determine the acceptable node duty cycle given a specific battery, and vice versa. Moreover, the suitability of the deployment environment (outdoor, well lighted indoor, poorly lighted indoor) for different application characteristics and battery types is discussed .

In a second contribution [22], we study the consequences of implementing photovoltaic energy harvesting on the duty cycle of a wireless sensor node, in both outdoor and indoor scenarios. We show that, for the static duty cycle approach in outdoor scenarios, very high duty cycles, in the order of tens of percents, are achieved. This further eliminates the need for additional energy conservation schemes. In the indoor case, our analysis shows that the dynamic duty cycle approach based solely on the battery residual energy does not necessarily achieve better results than the static approach. We identify the main reasons behind this behavior, and test new design considerations by adding information on the battery level variation to the duty cycle computation. We demonstrate that this approach always outperforms static solutions when perfect knowledge of the harvestable energy is assumed, as well as in realistic deployments, where this information is not available.

Data Collection with Moving Nodes

Patrolling with mobile nodes (robots, drones, cars) is mainly used in situations where the need of repeatedly visiting certain places is critical. In [24], we consider a deployment of a wireless sensor network (WSN) that cannot be fully meshed because of the distance or obstacles. Several robots are then in charge of getting close enough to the nodes in order to connect to them, and perform a patrol to collect all the data in time. We discuss the problem of multi-robot patrolling within the constrained wireless networking settings. We show that this is fundamentally a problem of vertex coverage with bounded simple cycles (CBSC). We offer a formalization of the CBSC problem and prove it is NP-hard and at least as hard as the Traveling Salesman Problem (TSP). Then, we provide and analyze heuristics relying on clusterings and geometric techniques. The performances of our solutions are assessed in regards to robot limitations (storage and energy), networking parameters, but also to random and particular graph models.

Also related to data collection, in [3], we advocate the use of conventional vehicles equipped with storage devices as data carriers whilst being driven for daily routine journeys. The road network can be turned into a large-capacity transmission system to offload bulk transfers of delay-tolerant data from the Internet. The challenges we address include how to assign data to flows of vehicles and while coping with the complexity of the road network. We propose an embedding algorithm that computes an offloading overlay where each logical link spans over multiple stretches of road from the underlying road infrastructure. We then formulate the data transfer assignment problem as a novel linear programming model we solve to determine the optimal logical paths matching the performance requirements of a data transfer. We evaluate our road traffic allocation scheme using actual road traffic counts in France. The numerical results show that 20% of vehicles in circulation in France equipped with only one Terabyte of storage can offload Petabyte transfers in a week.

Network Resilience

The notion of Shared Risk Link Groups (SRLG) captures survivability issues when a set of links of a network may fail simultaneously. The theory of survivable network design relies on basic combinatorial objects that are rather easy to compute in the classical graph models: shortest paths, minimum cuts, or pairs of disjoint paths. In the SRLG context, the optimization criterion for these objects is no longer the number of edges they use, but the number of SRLGs involved. Unfortunately, computing these combinatorial objects is NP-hard and hard to approximate with this objective in general. Nevertheless some objects can be computed in polynomial time when the SRLGs satisfy certain structural properties of locality which correspond to practical ones, namely the star property (all links affected by a given SRLG are incident to a unique node) and the span 1 property (the links affected by a given SRLG form a connected component of the network). The star property is defined in a multi-colored model where a link can be affected by several SRLGs while the span property is defined only in a mono-colored model where a link can be affected by at most one SRLG. In [4], we extend these notions to characterize new cases in which these optimization problems can be solved in polynomial time. We also investigate the computational impact of the transformation from the multi-colored model to the mono-colored one. Experimental results are presented to validate the proposed algorithms and principles.