Adaptability of a hybrid route choice model to incorporating driver behavior dynamics under information provision
ABSTRACT This paper proposes a seamless framework to incorporate the day-to-day and within-day dynamics of driver route choice decisions under real-time information provision by adapting a hybrid probabilistic-possibilistic model previously developed by the authors. The day-to-day dynamics are captured through the update of driver perception and route choice rules based on the current day's experience. The within-day dynamics are captured through the en-route adjustment of the weights of the driver route choice rules in response to situational factors. Experiments are conducted to analyze the model's ability to capture driver behavior dynamics, and the associated prediction accuracy. The results suggest that the framework can reflect the evolution of driver route choice behavior over time, and adapt to the within-day variability in ambient driving conditions. This is illustrated by its ability to capture phenomena such as inertia, compliance, delusion, freezing, and perception update under information provision, in addition to the effects of familiarity and route complexity. In the within-day context, the results highlight the sensitivity to the situational factors unfolding in real-time. The results also illustrate the better prediction power of the hybrid model compared to that of a traditional multinomial probit model; however, this gap reduces with increasing heterogeneity in driver behavioral class fractions. Elsewhere, the authors show that the proposed framework can be used to predict the ambient driver class fractions, thereby addressing a key deployment limitation of existing dynamic network models for real-time traffic control through route guidance.
Full-textDOI: · Available from: Srinivas Peeta, Jan 05, 2015
SourceAvailable from: Tai-yu Ma[Show abstract] [Hide abstract]
ABSTRACT: This paper proposes a distributed self-learning algorithm based on the regret matching process in games for a dynamic route guidance. We incorporate a user’s past routing experiences and en-route traffic information into their optimal route guidance learning. The numerical study illustrates that the proposed self-guidance method can effectively reduce the travel times and delays of guided users in congested situation.Agent and Multi-Agent Systems Technologies and Applications: Advances in Intelligent Systems and Computing, Edited by Gordan Jezic et al, 08/2014: pages 107-116; Springer.
[Show abstract] [Hide abstract]
ABSTRACT: The paper aims to model the travelers' day-today route choice in case of Advanced Traveler Information System (ATIS). The main focus is related to modeling travelers' route choice learning process. The comparison among different approaches is proposed, in particular, the considered approaches are divided into two main groups: the reinforcement learning (RL) based, including extended RL (ERL), and the belief-learning based, such as Joint Strategy Fictitious Play (JSFP) and Bayesian learning (BL). All analyses have been carried out considering data collected by a web-based Stated Preference experiment, where the respondents are provided with information at different levels of accuracy in order to investigate the effect of information reliability. The result shows that in case of intermediate and high accurate levels, JSFP best predicts the respondents' route choice behavior under information provision, reflecting a best-reply strategy is used by the travelers for their route choice decisions. However, in case of low information accuracy, the result shows a no-learning behavior due to the payoff variability effect. The finding can provide useful insight for supporting effective ATIS design.
[Show abstract] [Hide abstract]
ABSTRACT: The advances in adaptive learning dynamics to pure Nash equilibria in game theory provide promising results for the modeling of selfish agents with limited information in congestion games. In this study, a distributed game-theoretical learning algorithm with real-time information provision for dynamic congestion games is proposed. The learning algorithm is based on the regret matching process by considering a user's previously realised payoffs and real-time information. The numerical studies show that the proposed algorithm can converge to a non-cooperative Nash equilibrium in both static and dynamic congestion networks. Moreover, the proposed algorithm leads to a plausible real-time route choice modeling framework based on a user's perception being updated by incorporating the user's past experience, real-time information and behaviour inertia. © 2014 The Authors. Published by Elsevier B. V. Selection and peer-review under responsibility of the Scientific Committee of EWGT2014.17th Meeting of the EURO Working Group on Transportation, EWGT2014, 2-4 July 2014, Sevilla, Spain; 01/2014