Bio-inspired networking: from real-life network study to mobility modelling and mobile networking heuristics

The fast-paced evolution of mobile networking demand has brought problems of ever-increased complexity to the network research domain. Many of these problems often do not have all the elements for a complete solution. We have however to meet the market demand for a denser and simpler, pervasive, web of communicating objects. In as many domains as possible, we have to search for elements of knowledge that will help design better methods of self-configuring, healing, administrating and routing, in autonomous networks. In the context of the development of mobile Ad-hoc connectivity management and routing algorithms, the mobility model is very influent on the results that will be obtained. Hence it is very important to have mobility models that match reality at its best. But do we have very good mobility models at our disposition? For example, nowadays, the main mobility model used in the simulation of mobile ad-hoc algorithms is still the Random Waypoint Model, despite its obvious flaws [1] and the lack of similarity it has with real-life displacements of persons. In parallel, considering that the mother of self-organization is life itself, we chose to harvest this area. Starting from first observations on real-life networks and scale-free network models, we set a certain number of paths to further study, identifying challenges raised and hints at solutions proposed. We developed our research on one of these paths, making an excursion into sociology. Starting from the simple parallel between a behavioural aspect of man-to-group interactions and a dynamic principle of a real-life network model, we devised a mobility model from the ground up, bridging the gap between individual and group mobility models, bringing many new enhancements and fine grained population modelling, while meeting surprisingly well a widespread observation spanning many real-life domains: a scale-free spatial distribution. We are now further studying the relationship between mobility parameters and the characteristics of the inherited topology, in various population evolution scenarios. The future work we plan to accomplish will debut by capturing as precisely as possible, in the frame of a campuswide experimental radio mesh network, spatial and dynamic characteristics of diverse human users, and confront their statistical properties to the results of our model. This will hopefully bring us an indication on valid parameter sets for this model, accompanied with relevance of it, if found to be relevant in this domain. It will also, in a more certain fashion, give us hints on enhancing it. This model will then be useful in two aspects: • Being a behavioural model, it will give us clues on the statistical presence of different parameters of the behaviours it implements. This may serve in feedback to sociologists, and validating certain defined behaviours it would be hard to measure in a dense population, it will hopefully serve as a base for innovative heuristics matching problems such as routing, connectivity establishment, self-healing and security. • It will permit to extend our test scenarios, blending and expanding real-life data, to test, measure, and validate the new solutions that research can bring to consider. One of the prominent aspect that we would like to explore is the clustering phenomenon. In the first place, because it is a prominent factor in many real-life studied networks, which have a particularly high “clustering coefficient1”, compared to random networks [2]. Matching the presence of clustering inherited from different individual behaviours will help finding heuristics at the individual level leading to a network with a persistent defined level of clustering. Many fine solutions to efficient routing, self-configuration and failure resilience do exist that make use of clustering. Remains very interesting questions, such as for example, finding local algorithms that will find the clusters quickly. Probable hints at this could be made, based on the structure of clustering obtained from different individual behaviours. Finally, we would like, by this work, to strengthen the links that are being built between separate preexisting disciplines. This could be made by studying what other behaviours inherited from biologic or sociologic observation could be integrated in the framework, or more generically, to show by