Exploiting probe data to estimate the queue profile in urban networks

Abstract

Queues at signalized intersections are one of the main causes of traffic delays and urban traffic state variability. Hence, a method to estimate queue characteristics provides a better understanding of urban traffic dynamics and a performance measurement of signalized arterials. In order to capture the evolution of queues, we aim at leveraging the collective effect of spatially and temporally dispersed probe data to identify the formation and dissipation of queues in the time-space plane. The queue profile characterizes the evolution of both queue front and back, which consequently can be separated in a two-step estimation process resulting to the queue profile polygon. The evolution of queue front, in the time-space diagram, based on the kinematic traffic shockwave theory is modeled as a line with the known slope of queuedischarging shockwave and estimated with a constrained optimization and a technique known as support vector machine. The evolution of back of queue is more challenging and modeled as a piecewise linear function where slope of segments is between the queue-discharging shockwave and zero. In the proposed method, the input data consists of position and velocity of probe vehicles. The queue profile estimation method does not require any explicit information of signal settings and arrival distribution. The proposed method is tested with various penetration rates and sampling intervals of probe data, which reveals promising results once compared to a uniform arrival queue profile estimation procedure. The proposed method could be beneficial for spillback identification, vehicle trajectory construction, and fuel consumption and emission estimation.

Queue Profile Estimation in Congested Urban Networks with Probe Data

“…() and Hofleitner et al., () also see Vlahogianni and Karlaftis () for incident duration modeling, Ghosh‐Dastidar and Adeli () for delay and queue length estimation at freeway work zones, and Jiang and Adeli () for traffic flow forecasting). A brief description of the developed method without considerations for spillovers, capturing the interdependencies of queuing dynamics of adjacent links and less robust optimization framework to noisy measurements is presented in Ramezani and Geroliminis ().…”

“…Although queue estimation is straightforward given trajectory of probe vehicles with large penetration rates, for realistic cases with low penetration rates and high sampling intervals, an approach that combines data mining, optimization techniques, and physical properties of traffic flow is needed (e.g., see Hao et al (2012) and Hofleitner et al, (2012) also see Vlahogianni and Karlaftis (2013) for incident duration modeling, Ghosh-Dastidar and Adeli (2006) for delay and queue length estimation at freeway work zones, and Jiang and Adeli (2005) for traffic flow forecasting). A brief description of the developed method without considerations for spillovers, capturing the interdependencies of queuing dynamics of adjacent links and less robust optimization framework to noisy measurements is presented in Ramezani and Geroliminis (2013).…”

Probe Vehicle Lane Identification at Signalized Intersection Using Support Vector Machine Method

“…Dalam beberapa studi terdahulu, kendaraan penyidik ini sudah diteliti penggunaannya untuk estimasi waktu perjalanan (Chen and Chien 2001, Chu, Oh et al 2005, Liu and Ma 2009, deteksi insiden (Sethi, Bhandari et al 1995, Dia andThomas 2011), evaluasi keadaan lalu lintas (Dai, Ferman et al 2003, Nanthawichit, Nakatsuji et al 2003, estimasi panjang antrian (Comert and Cetin 2009, Cetin 2012, Comert 2013, Ramezani and Geroliminis 2013, pola pemilihan rute dan bahkan untuk estimasi kekasaran perkerasan jalan.…”

Deriving driver-centric travel information by mining delay patterns from single GPS trajectories

“…For instance, regular single stops at crossways or in front of traffic lights have considerable influence on local measured travel times but not on the local LOS. To overcome such insufficiencies of FCD, previous approaches such as (i) average travel times over several vehicles close in time [2], (ii) average travel time over larger distances [4,7], or (iii) analyzing travel times with situation specific or facility type specific algorithms such as queue detection at intersections [10,13,8] have been proposed. This paper proposes a different approach: for subsequent LOS estimation, GPS trajectories of single vehicles are analyzed and detected delays are classified according to predefined delay patterns.…”