Day-to-Day Evolution of Network Flows Under Route-Choice Dynamics in Commuter Decisions

Abstract

The day-to-day dynamics of commuter decisions on urban traffic networks induced by route-choice dynamics is investigated. More specifically, this investigation analyzes the stability and reliability of system performance due to route-choice dynamics under real-time information, relaxing key restrictions associated with user equilibrium models. A simulation-based framework is developed to analyze day-to-day dynamics by integrating an empirically calibrated model of route-choice decisions with a dynamic network assignment model. Computational experiments are used to investigate the effect of certain experimental factors—recurrent network congestion level, market penetration, nature of information, and frequency of information updates—on network performance stability and reliability. The findings provide evidence of considerable day-to-day variations and stochasticity in network flows and performance, even when departure-time decisions and the origin-destination matrix are assumed to be fixed. The results indicate that (a) network performance may deviate significantly from equilibrium because of route-choice decisions under information; (b) within-day route choices have a significant effect on day-to-day network flow evolution; (c) user equilibrium path flows may not be reached or may not be stable when uncoordinated route-choice decisions are made under real-time information; (d) route-choice decisions do not exhibit convergence to an equilibrium, although the switching rate may be near steady-state conditions; (e) advanced traveler information systems (ATIS) information strategies can affect both within-day and day-to-day dynamics; and (f) some information strategies can lead to improved reliability and stability but at the expense of longer trip times. The results indicate that the trade-off between trip time performance and stability must be considered in network analysis and design. These results have important implications for the design of traffic control strategies, more-effective ATIS implementation guidelines, and incident management strategies.

Effects of user equilibrium assumptions on network traffic pattern

“…As could be expected, such studies (Nakayama et al, 1999;Horowitz 1984;Iida et al, 1992;Cantarella and Cascetta, 1995;Jha et al, 1998;Oh et al, 2003;Rossetti and Liu, 2005) found that the drivers' route choice could be affected by their earlier experiences. Recently, Srinivasan (2003) considered the disutility from changing paths to describe route choice behavior, and Nakayama et al (1999Nakayama et al ( , 2000Nakayama et al ( , 2001 developed a rule-based model to study drivers' route choice behavior. However, their process was rather complicated and their route-based if-then rules cannot consider the overlapped attributes between routes.…”

A day-to-day dynamical model for the evolution of path flows under disequilibrium of traffic networks with fixed demand

“…Yang and Zhang (2009) classify the analytical approaches into five categories: the simplex gravity flow dynamics (Smith, 1983), the proportional-switch adjustment process (Smith, 1984;Smith and Winsten, 1995;Huang and Lam, 2002;Peeta and Yang, 2003), the tatonnement process (Friesz et al, 1994), evolutionary dynamics (Sandholm, 2001;Yang, 2005) and the projected dynamical system Nagurney, 1995, 1996;Zhang, 1996, 1997). Several studies adopt simulation-based or field survey approaches to model the DTD dynamics (Mannering et al, 1990;Hu and Mahmassani, 1995;Mahmassani and Stephan, 1988;Caplice and Mahmassani, 1992;Mahmassani and Jou, 2000;Srinivasan and Mahmassani, 2000;Srinivasan and Guo, 2004). While these approaches consider several factors including the socioeconomic and weather-related variables to model the path-shift behavior, the analytical models typically rely on the path features for this purpose.…”

Doubly Dynamic Traffic Assignment

“…[1] initialise the route choice probabilities based on free flow costs (initialisation of (3)); [2] allocate the drivers in various departure periods to routes based on random multinomial experiments (implementation of (5)); [3] compute the departure period dependent experienced route costs based on a dynamic network loading map, (6)- (10); [4] at the end of day n-1, the population mean experienced route costs are updated using the learning model (3) and the costs fed back to the first step above; and [5] compute the summaries viz., means, variances and covariances of route flows at the end of the realisation.…”

“…In fact, the purview of dynamic traffic assignment is much wider and includes day-to-day variations in the demand in addition to the usual within-day variations. Day-to-day evolution of traffic flows was considered by several authors in the past (1)(2)(3)(4), all of whom focused on the evolution of the traffic flows across the days either as a stochastic or a deterministic process, but primarily based on static within-day cost-flow functions. On the contrary, nowadays, more generalised traffic assignment models are being developed which are aimed at addressing both the day-to-day and within-day variations in route flows and such models are called doubly dynamic traffic assignment models, which are the main subject of the present paper.…”