Some Observed Features of Freeway Traffic Oscillations

Abstract

Some recent published works on freeway traffic have focused on congested traffic conditions in which oscillations form and propagate upstream through vehicle queues. Some features of oscillating waves were examined through empirical evidence. The data measured on multiple days along an approximately 3.5-km three-lane freeway in Istanbul, Turkey, present pronounced features of oscillations, such as their formation, propagation, and growth. It was identified that oscillations form immediately upstream of a bottleneck, they propagate at certain speeds against the direction of traffic flow, and their amplitude grows while propagating. In reports published earlier, it was stated that oscillating waves propagate backward at nearly constant speed with no dependence on flow. It has been observed in this study that propagation speeds cluster into two groups with averages of 21.29 and 15.52 km/h at the downstream and upstream study segments, respectively. It was shown that any difference in propagation speeds may be attributable to the features of the segment studied and, in turn, to the slope of the congested part of the associated fundamental diagram. A theory on reduction of oscillation amplitude at merge sites proposed in the literature was tested, and predictions of the theory were validated with some acceptable error.

Spatiotemporal filtering method for detecting kinematic waves in a connected environment

“…Subtracting the moving averages from the oblique cumulative vehicle count curves amplifies the temporal changes, and the propagations of KWs can be tracked visually [2,3]. However, such procedures require the manual determination of the parameters in each road segment, which makes the systematic implementation of the method challenging.…”

“…The data within the window, W 1 spanning V D1 were compared with the data observed at an upstream location within the same window and a varying time lag, Δl as in Fig 4B: Δl reflects the time for the KW to travel to the upstream location. The time lag ranged from 20 to 120 seconds, considering that the speed of KW can range from 6.3 km/h (3.9 mph) to 37.8 km/h (23.4 mph) [2,3].…”

“…Fig 6D shows the performance of the model in terms of false-positives (FPs), false negatives (FNs), true positives (TPs), and true negatives (TNs) based on the true KW shown in Fig 6E. The red boxes in Fig 6E mark the KW that emanated further downstream of D, displaying a better-defined width of KW and a stabilized speed. The blue boxes in Fig 6E mark the KW that emanated near D. Most of the FPs and FNs occurred in detecting KWs that emanated near D. This could be the result of KWs that were not stabilized until they traveled further upstream, and it also could indicate the increase in the amplitude of KW as it was propagated [3,16,43]. Fig 7 shows the effects of the level-of-congestion (C) and asynchronous changes in speed and flow (H) on detecting KW in the MELR excluding random effects.…”

“…Most of the falsely-identified KWs were emanated near D, while the bottleneck remained active. (See Fig 6E.) This may be the result of KWs not yet becoming stable since this occurs as they travel further upstream [3,16,37]. Therefore, when the proposed method is applied in practice, the several RSUs collecting traffic data should be installed at specific intervals to detect the downstream's false negatives in the upstream.…”

“…Backward-moving kinematic waves (KWs) signal changes in traffic conditions (i.e., traffic shockwaves [1] and stop-and-go traffic conditions [2,3]) such that detecting traffic conditions leading to the emanation of KWs and monitoring their propagations have become important topics of research. The information gleaned from KWs is vitally important in the real-time traffic management [4], in the evaluation of traffic safety [5,6], and connected vehicle control strategies [7][8][9][10][11].…”