Spatial Extrapolation of Anisotropic Road Traffic Data

Abstract

A method of spatial extrapolation of traffic data is proposed. The traffic data is given by GPS signals over downtown Berlin sent by approximately 300 taxis. To reconstruct the traffic situation at a given time spatially, i.e., in the form of traffic maps, kriging with moving neighborhood based on residuals is used. Due to significant anisotropy in directed traffic data, the classical kriging has to be modified in order to include additional information. To verify the extrapolation results, test examples on the basis of a well-known model of stochastic geometry, the Boolean random function are considered.

Spatial Inference of Traffic Transition Using Micro–Macro Traffic Variables

“…Furthermore, the majority of previous works has not been developed for lane-level inference as their focus is at a more coarse-grained level, e.g., at road-level [4], [8], [9], zonelevel [10], or even city-level [11]. A family of geostatistical interpolation techniques called ordinary kriging [22] has been extensively used to spatially and collectively infer zonal traffic conditions over a large area consisting of several road segments.…”

Lane-level traffic estimations using microscopic traffic variables

“…In [4], an extensive investigation is performed on an inference of Annual Average Daily Traffic on weekdays in cities across the state of Texas. Similarly, Ordinary Kriging is employed over rush-hour durations in the city of Berlin to spatially infer traffic congested areas [5]. These proposed Kriging models do not take into account lane-level spatial and temporal variability of the traffic variables across the measurement sites themselves and subsequently, cannot be applied for estimation at lane-level which is a much finer scale.…”

“…The contributions of this paper are as follows. First, the estimation is performed at lane-level which is in a much finer scale than previous approaches [3][4][5][6]. Second, the proposed model is adaptable to different traffic regimes by assessing only changes in microscopic traffic variables, e.g.…”

“…Only a few models have been proposed for the inference of traffic on road segments where local information cannot be measured directly [3][4][5][6], and they are not primarily designed for lane-level estimation under different traffic regimes. In [3], a context awareness model is proposed to infer traffic congestion level on a road segment into {low,medium,high} by using inference rules to incorporate traffic variables from sensors with context attributes such as time of day, weather data and incident report.…”

A Platform for Citizen Sensing in Sentient Cities

“…We are seeking solutions in terms of traffic simulation, as e.g. [15], [6], and [9], including solutions focused on data extrapolation techniques such as [1] and [2].…”