Section: Scientific Foundations

Highly scalable protocols

The networks formed in an urban environment can sometimes be particularly challenging for the MAC layer protocols and QoS support, especially if the network is not centralized or synchronized: very high node degree, unstable and asymmetric links, etc.

MAC layer protocols are either very difficult to implement in distributed and self-organized environment or present serious scaling issues. Studies focusing on distributed TDMA showed that MAC protocols from this class can be successfully designed to accommodate channel access for a high number of contending nodes. However, scalability is always obtained following a learning phase with relatively high convergence time. This means that in a dynamic network scenario like the one encountered in most urban capillary networks, the MAC protocol spends most of the time in the learning phase, where it achieves a reduced performance. The same problem appears when trying to distribute other usually centralized schemes, such as OFDMA or CDMA. On the other hand, CSMA/CA protocols are distributed by their nature.

However, the current leading solutions in this area are based on the IEEE 802.11 Distributed Coordination Function (DCF), a channel access method designed and optimized for Wireless LANs with a central access point and a maximum of 10-20 contending stations. The DCF is well-known for its scalability issues, especially in multi-hop dynamic networks, and adding energy constraints usually existing in wireless sensor networks does not improve its performance. While multiple MAC layer congestion control solutions have been proposed in the context of mobile ad-hoc networks, the approach is usually based on the idea of reducing the number of neighbors, either through transmission power control or data rate adjustment. However, this is just a workaround and the search for a truly scalable MAC layer protocol for high density wireless networks is still open.

Regarding the network layers, in order to have multi-service platforms deployed in practice, all the requirements of telecommunication operators should be present, in particular in wireless sensor and actuators networks, within the key notion of Service Level Agreement (SLA) for traffic differentiation, quality of service support (delay, reliability, etc.). Moreover, because the world becomes more and more connected to Internet, IP should be supported in wireless sensor networks. The IETF proposes the use of RPL (Routing Protocol for low power and lossy networks), where it is clear that the support of several Destination oriented Directed Acyclic Graphs (DoDAG) is required, and a complete traffic management is needed. Moreover, RPL assumes a static topology but the classical sensor networks give way to urban sensing, where the user's smartphone give the physical measures to the operators. Therefore, the data collection becomes distributed, sometimes local, the network is now dynamic. In such a scenario, inconsistencies stemming from data collected using different calibration process raise a lot of interests. Moreover, data aggregation and data gathering is, in capillary networks, at the heart of the issues related to the limited capacity of the networks. In particular, combining local aggregation and measurement redundancy for improving data reliability is a promising approach.