Trading accuracy and resource usage in highly dynamic vehicular networks

Vehicular networks bring new ways of viewing road traffic management and safety applications. For the first time, it will be possible for vehicles to exchange information and build fine-grained knowledge about the current situation, estimating risks and adapting their driving. Central to these applications is the need to exchange information in a highly dynamic environment, building a view of the current situation before the conditions change. This is turn requires the distributed algorithms used to converge on low error margins quickly. In this paper, we investigate the performance of such a distributed algorithm which aims to build a common assessment of the risk level among vehicles to trade off accident risks with road throughput. In particular, we examine how the convergence rate is affected by network size and node density, and also how the error in the algorithm´s output is affected by the rate at which nodes send out update beacons. We develop a variable-rate beaconing scheme in order to find a trade-off between accuracy of outputs and network resource usage. We then formulate this as a more general optimisation problem applicable to other applications or distributed algorithmic problems in highly dynamic distributed systems such as VANETs.