Experimental Analysis of Route Choice Behaviour with Real-time Information in Congested, Risky Networks
X Lu, S Gao, University of Massachusetts, US ; E Ben-Elia, University of the West of England, UK
We study the effect of real-time information on route choice behavior under uncertain disruptions using data collected from human subjects who simultaneously and competitively make route choices in controlled PC-based laboratory experiments.
Traffic networks are inherently uncertain with random disruptions creating significant congestion, which usually include incidents, bad weather and work zones. Meanwhile, the development of Advanced Traveler Information Systems (ATIS) is based on the premise that better-informed travelers make route choices that increase their own welfare and collectively reduce system-wide congestion and unreliability. However the premise needs to be verified. The presence of information would affect travelers' choices, which in turn would impact the performance of the traffic system as a whole, resulting in complex interactions between traveler¡¦s choices and uncertain traffic system. Capturing the characteristics of their interactions is valuable for evaluating the effectiveness of ATIS. Recent research investigating route choice behavior incorporating both ATIS and network stochasticity is limited to an uncompetitive situation. On the other hand, most traffic prediction models that consider equilibrium from competitive route choice and incident condition predict ¡§normal¡¨ traffic in an underlying deterministic network and assess the impact of uncertain disruption by loading such ¡§normal¡¨ traffic flows onto the disrupted network and evaluating the resulting traffic conditions. If real-time information is available, diversions from travelers¡¦ ¡§normal¡¨ routes are enabled through certain mechanisms. The implicit assumption of such approach is that when making regular day-to-day travel choices, the possibility of disruptions and the existence of traveler information systems are not considered by the travelers. However, such assumption seems unrealistic. Travelers will potentially recognize the strategic advantage of the information on an uncertain, disrupted network and plan in advance for the information they will have later in the trip. Since investigating route choice behavior in real life is very difficult, if not impossible, due to myriads of complicating factors in such a decision environment, we develop a controlled laboratory experiment, in which the environment can be clearly specified and the effects of control variables obtained through direct comparison.
To the best of our knowledge, this research is the first experimental study to analyze route choice behavior in a competitive situation, considering both uncertain disruptions and real-time information. It sheds light on two central issues, 1) Does real-time traveler information relieve congestion and increase reliability of the network under uncertain disruptions? 2) Whether travelers behave strategically by planning in advance for information or simply react to information only when it is delivered? Understanding of these issues will provide an insightful look into the role of real-time information in randomly disrupted traffic network and promote a paradigm shift in traffic modeling which could better incorporate uncertain disruptions, real-time information and their significant impacts on congestion and travel reliability.
In this paper, a repetitive interactive experiment is conducted in which a relative large group of subjects use a PC-based interface to choose routes link-by-link in a competitive, risky network. The hypothetic network contains three routes to a single destination, a direct one and a route that branches into a link with incident probability of 0.25 and a detour. Two treatments are designed and the difference is that, real-time information on realized incident condition is only provided in treatment 2 through a variable massage sign (VMS) located at the start of the detour. In both treatments, subjects are informed of the incident probability before the trip, and receive feedback of the actual travel times on all alternative routes and realized incident condition after the trip is completed. Travel times are calculated from cost functions that subjects do not know, but they are made aware that a route will be more congested with a larger number of users. Each treatment has a total of 80 rounds.
To analyze the benefit of the real-time information, the comparison between these two treatments is made in probability distributions of route flows, route travel times and the total system travel time. Results from treatment 2 are checked against predictions from two traffic equilibrium models to assess whether considering strategic behavior significantly improves the accuracy of a conventional model assuming reactive route choice under real-time information.
Association for European Transport