Section: Scientific Foundations
Characterizing urban networks
A typical urban capillary network will involve a set of different communication technologies like 3G/ LTE, IEEE 802.11, WSN, inter vehicular communications and many others. Each technology relies on a set of mechanisms that were designed to provide a dedicated set of functionalities. Typical mechanisms include resource allocation, scheduling, error detection and correction, routing etc.
Dimensioning the operating parameters of such network mechanisms in order to provide the desired services while ensuring the network efficiency is a classical and yet a difficult issue. There are many directions to address this problem. For instance, one can refer to the network dimensioning and traffic-engineering approaches. Cross layer optimization and Self-organizing networks (SON) paradigm in 3G/LTE are also other perspectives to tackle this issue. However, given the complexity of the problem, most of the efforts concentrate on the mono-technological and/or the mono-service cases.
In the urban scenario, the heterogeneity of the technologies and the particularity of the urban services bring up new network-dimensioning challenges. The optimization has to be extended to the inter-technological perspective and to the multi-services standpoint. The different technologies that compose the capillary network have to inter-operate in a seamless and optimal way so that they can provide user-centric services with the desired quality of experience. Consider, for instance, dimensioning the scheduling mechanism of a mesh network, which has to carry the traffic generated by different WSN in the city. Predicting the time and spatial distribution of the traffic generated by the different WSNs are clearly among the key elements that shall be considered. On the other side, from a downlink standpoint, consider the judicious setting of an WSN aggregation mechanism accordingly with the time varying capacity of the mesh backbone level.
It is quite clear that these questions cannot be addressed without characterizing the features of an urban capillary network. This covers the geographical properties of the networks (distribution, density, nodes degree, mobility etc.) as well as the data traffic characteristics of urban services. Understanding these proprieties and their correlation is still an uncovered area. The main challenge in this case is the production of quantitative traces from real or realistic urban mobility, networks and services. For example, in urban mobility scenarios, how long devices are in radio range of each other gives temporal constraints on the communications protocols that should be understood. In this duration, devices have to self-organize or to hang on the exiting organization and to exchange information.
A second step is to derive analytical or simulation models that will be used for network dimensioning and optimization. Many models already exist in the literature in related scientific fields and they could be considered or adapted to this purpose. This covers different models ranging from radio propagation, vehicular or pedestrian mobility, traffic pattern, etc, the difficulty being on how to mix these models and how to choose the right time magnitude and spatial scale in order to preserve the accuracy of the capillary network features while maintaining the model complexity tractable. The derived models could serve to optimize the different mechanisms involved in the urban capillary network.
The inference between different networks and services is quite complex to understand and to model, therefore a simple approach would be to decouple the models. Choosing the right decoupling technique depends on the targeted temporal and spatial level of the input and output parameters. Again, the latter shall capture for each decoupled model a selected set of significant features of the capillary network. Finally, the purpose of the constructed models is to obtain the optimal dimensioning of the network mechanisms. Several optimization techniques, from exact to heuristics ones, shall be considered to compute the best operating parameters. One of the main challenges here is to maintain the computational complexity tractable by exploiting the specific structure of the problems induced by the city.