Modelling for Wireless Sensor Network Protocol Design

Protocol design for a Wireless Sensor Network (WSN) must jointly take many aspects into account, owing to the embedded nature of the devices, and the needs for energy efficiency, scalability, self-organisation and robustness to heterogeneous and unplanned environments: all these are crosslayer issues. As usual for wireless systems, protocol design normally includes a heuristic design phase, followed by validation and optimisation that can be achieved through simulation, mathematical and/or experimental approaches. In the two former cases, models are needed to specify the environment addressed by the researcher: as usual, models should be simple and realistic at the same time, where these two characteristics must be defined according to the target of the evaluation procedure. Models should also be agreed and shared among the scientific world in order to allow the comparability of results achieved by different researchers. In a wider sense, modelling for protocol design should also include the definition of the performance figures that have to be selected to measure protocol optimality, or to benchmark them. Some models have got in the last year a natural consensus, like for instance the geometrical distribution of nodes, often considered to be (statistically) uniformly distributed over a square area. However, other models for WSN protocol design still need to find consensus. This paper addresses this aspect and discusses the relevance of some of the issues not sufficiently accurate when dealing with modelling for WSN protocol design. To provide a solid base to this discussion, some results achieved through different means, both via simulation, mathematical or experimental approaches, are reported in the paper; as a common denominator to the different approaches and algorithms discussed, the performance of WSNs is measured in terms of network lifetime (related to energy efficiency), for which a new definition is given in this paper.