Vehicular Sensor Networks for Traffic Flow Reconstruction

Abstract

This Doctoral Thesis is about the characterization of traffic flow using data from a small part of the vehicles that comprise such flow. We propose the use of an ad-hoc wireless network formed by a fraction of the passing vehicles, the probe or sensor vehicles, to periodically recover their positions and speeds. These vehicles, together with wireless bridges located close to the road shoulder, the Road Side Units (RSU), compose the Vehicular Sensor Network. Gathered data are then rearranged in a time-space diagram as a part of microscopic traffic flow representation. Finally, the speed/position information or Space-Time Velocity (STV) field is reconstructed in a Data Fusion Center by means of interpolation techniques. We have used widely accepted theoretical traffic models (car-following, multi-lane and overtakeenabled) to replicate the nonlinear characteristics of the traffic flow in representative situations along several experiments with different traffic-related parameters. In order to obtain realistic packet losses, we have simulated the multihop ad-hoc wireless network with an IEEE 802.11p PHY layer. The interpolation is based on the generation of Triangular Irregular Network, to our knowledge, is the first time such a interpolation is used in traffic context. In addition, we have performed discrete optimization to recover the most relevant time-space regions (cells) and the relation of such cells with traffic flow and the ocurrence of probe vehicles. Finally, we have derived a local density-flow diagram from sensor vehicles which occurs in selected cells. This Doctoral Thesis conclude that: 1) for relevant configurations of both sensor vehicle and RSU densities, the wireless multihop channel performance does not critically affect the STV reconstruction error; 2) the system performance is marginally affected by transmission errors for realistic traffic conditions, 3) the STV field can be recovered with minimal error for a very small Fraction of Sensor Vehicles (FSV) 9% in any traffic condition, it can be further reduced for congested traffic; 4) for that FSV value, the probability that at least one sensor vehicle transits the spatio-temporal regions that contribute the most to reduce the STV reconstruction error, sharply tends to 1. 5) In addition, for certain traffic conditions, the local density-flow diagram is rendered from a mere 3%. Thus, a random and sparse selection of wireless sensor vehicles in realistic traffic conditions, together with the proper interpolation and space-time data analysis techniques, is sufficient to get an accurate reconstruction of any STV field.