Data Quantity for Reliable Traffic Information in Vehicular Ad Hoc Networks

Vehicular ad hoc networks (VANETs) are promising platforms from which decentralized traffic information systems can be built. In VANETs, susceptible situations, such as communication disruptions and low market penetration, could cause a dearth of information to be provided. This study examined the minimum quantity of data required for reliable travel time measurement over a variety of traffic conditions. Individual travel time data were obtained from computer simulation. The premise of this study, which is that data quantities can be reduced in certain situations, has important implications for the statistical methodologies used to assess sample quality, since appropriate conditions for the Central Limit Theorem might not be met. We compare such results with a distribution-free method called Acceptance Probability. Importantly, unstable traffic conditions, which are those of greatest interest for the purposes of studying congestion, also contribute to reduction in sample sizes, regardless of the technical quality of the wireless system at any given time. We show that in many realistic cases, standard sample size techniques under-estimate the required sample sizes and therefore a more refined approach is called for when using these data for estimation purposes.