Section: New Results

Self-Organization

Participants : Antoine Gallais, Nathalie Mitton, Valeria Loscri, Farouk Mezghani, Anjalalaina Jean Cristanel Razafimandimby.

Stable parent selection

The Industrial Internet of Things consists in the use of low power lossy networks to enable next industrial applications. To work properly, the network has to provide strict guarantees concerning the delay and the reliability. IEEE 802.15.4-TSCH proposes time synchronized and slow channel hopping medium access control to cope with these requirements. It relies on a strict schedule of the transmissions, spread over orthogonal radio channels, to set up a resilient wireless infrastructure. A routing protocol (e.g. RPL) has then to construct energy-efficient routes on top of this link-layer topology (as investigated in the 6TiSCH IETF working group). Most of existing solutions rely on tree-based topologies, where each node has to select one or multiple parents to forward its traffic to the destination. Unfortunately, the links to the routing parents exhibit time-varying characteristics, due to e.g. obstacles, and external interference, thus leading to oscillations and increased required control of the routing topology. Moreover, the network has to provision enough resources (i.e., time, channel) to cope with those variations, while still being reactive to node/link failures. We investigated the stability of 6TiSCH networks, and especially the impact on routing parent selection. We identified moments of instability due to oscillations in the radio conditions caused by external interference and obstacles, in two indoor testbeds with different channel conditions. We identified the causes of instabilities, and proposed solutions for each of the layers in the 6TiSCH stack. First, at the MAC layer, we demonstrated that a rearrangement of shared cells in the slotframe reduces the probability of collisions for control packets, paving the way to a faster negotiation during topology reconfigurations. Next, we eased the schedule consistency management between two nodes (renegociated from scratch in the current standard, upon detection of a schedule inconsistency). Finally, at the routing layer, we exploited the existing correlation between the broadcast packet reception rate and the unicast link quality to create a two-step parent selection that favors stable parents. We finally obtained a network that converged faster and that reacted accurately during moments of instabilities. Results are available in [46], [42].

Bayesian communications

The amount of data that are generated in IoT devices is huge and the most of time data are highly correlated, by making useless the forwarding of all the raw data generated. Bearing that in mind, we have designated and implemented an effective mechanism to reduce the amount of data sent in the network in [45]. Results are encouraging since there is a size effect of less interfering in the communication system with an important impact on battery consumption for wireless devices that are energy constrained.

Multi-technology self-organization

Opportunistic communications present a promising solution for disaster network recovery in emergency situations such as hurricanes, earthquakes, and floods, where infrastructure might be destroyed. Some recent works in the literature have proposed opportunistic-based disaster recovery solutions, but they have omitted the consideration of mobile devices that come with different network technologies and various initial energy levels. [19], [30] present COPE, an energy-aware Cooperative OPportunistic alErt diffusion scheme for trapped survivors to use during disaster scenarios to report their position and ease their rescue operation. It aims to maintain mobile devices functional for as long as possible for maximum network coverage until reaching proximate rescuers. COPE deals with mobile devices that come with an assortment of networks and aims to perform systematic network interface selection. Furthermore, it considers mobile devices with various energy levels and allows low-energy nodes to hold their charge for longer time with the support of high-energy nodes. A proof-of-concept implementation has been performed to study the doability and efficiency of COPE, and to highlight the lessons learned. Following-up with these results, we performed several experimentations and could benchmark smartphone performances with regards to their multi-communications interfaces. Testing experiments have been carried out to measure the performance of smartphones in terms of energy consumption, clock synchronization and transmission range. We believe that such experimental results can support technological choices for rescue operations but also for many other applications relying on smartphone performances. Results are available in [30].

Heterogeneous Self-organizing (smart) Things

In the panorama of the Internet of Things, one main important issue is the management of heterogeneous objects, that need to communicate in order to exchange information and to interact in order to be able to synergically accomplish complex tasks and for providing services to final users. In this context, the thesis [10] has tried to face the main challenges related to complex heterogeneous systems, where objects are able to self-organize to each other and are equipped with some kind of intelligence in order to dynamically react to the environment changes. Several tools have been exploited ranging from artificial neural networks to genetic algorithms and different solutions have been proposed to make these systems dynamic and responding to the $*$self properties.