Constructing summarizations for V2V traffic data based on sampling methods

Abstract

In vehicle-to-vehicle environment, vehicles interchange traffic data with each other. In this paper, three sampling methods in data stream theory are exploited in order to construct summarizations for these original data, since the original data are potentially unbounded in size and are impossible and unfeasible to be stored directly on the disk. Then traffic states, such as speed on any road, can be estimated based on the summarizations instead of the original data. The two measurements, absolute biases of speed and cover rates of information obtained from the constructed summarizations, are used in validating three methods' feasibility. Preliminary simulation results show three methods are feasible to some extent.