Agent-Based Route Choice with Learning and Exchange of Information

Abstract

Planning models require consideration of travelers with distinct attributes (value of time (VOT), willingness to pay, travel budgets, etc.) and behavioral preferences (e.g., willingness to switch routes with potential savings) in a differentiated market (where routes have varying tolls and levels of service). This paper proposes to explicitly model the formation and spreading of spatial knowledge among travelers, following cognitive map theory. An agent-based route choice (ARC) model was developed to track choices of each individual decision-maker in a road network over time and map individual choices into macroscopic flow pattern. ARC has been applied to both the Sioux Falls and Chicago sketch networks. Comparisons between ARC and existing models (user equilibrium (UE) and stochastic user equilibrium (SUE)) on both networks show ARC is valid and computationally tractable. In brief, this paper specifically focuses on the route choice behavior, while the proposed model can be extended to other modules of transportation planning under an integrated framework.Urban Sci. 2018, 2, 58 2 of 17 accurate description of this complex mechanism, so they are incapable of addressing important issues such as equity. Many researchers [9][10][11][12] have emphasized the importance of equity as a consequence of road pricing and pointed out "equity is an individual, not a group, problem" [13]. To account for this complexity, transport economists and policy makers have long advanced their focus from first-best prices with homogeneous network assumptions to second-best prices under heterogeneity for both network users and service providers [14][15][16][17][18][19][20], which requires explicitly modeling individual travelers' route choice behavior. Although some researchers have studied these problems on small networks, a behaviorally based model, which is not only sufficiently accurate but also applicable on a large network, does not exist [8].It is crucial to recognize that travel decisions are based on travelers' knowledge about the network. Travelers, limited in their capability of acquiring, processing and storing spatial knowledge, can only consider the routes they know. This is the very reason that engineers try to assist travelers to make decisions by providing additional information about the network through technologies like ATIS. As indicated by [2], ATIS cannot be well evaluated without explicitly accounting for the "heterogeneity in behavior" and "the presence of dynamic learning and adjustment processes in user behavior". Other researchers [21,22] also emphasized the role of information and learning in traveler behavior such as route choice decisions. However, limited work has been done to systematically model the mechanism of acquiring, processing, storing of spatial knowledge and its communication among travelers.The route choice model maps travel demand (traditionally defined by an origin-destination (OD) trip table, or in more recent research and practice as a list of travelers with particular attributes going betwee...

Shortest paths, travel costs, and traffic

“…This provides evidence not that individual travelers consider a bundle of costs (though they might) but that each traveler (or class of traveler) weights different costs differently in choosing routes. Thus an agent-based route choice model (Dia, 2002; Manley et al., 2014; Nagel and Flötteröd, 2012; Zhang and Levinson, 2004; Zhang et al., 2008; Zhu and Levinson, 2018), or a multi-class assignment (Dafermos, 1972) with a sufficient number of classes, each having different weights for different cost components will likely make a better prediction than simply considering multiple costs in a single-class assignment. In short user heterogeneity may be a more significant factor than multiple cost categories in explaining resulting traffic flows.…”

Personalized real-time traffic information provision: Agent-based optimization model and solution framework

“…To evaluate the impacts of different integrated multimodal corridor management strategies, Zhou et al (2008) developed a dynamic micro-assignment modeling approach to simulate the route, mode and departure time choices of traveling agents. Zhu et al (2008) proposed an Agent-based Route Choice (ARC) model to track choices of each individual decision-maker in a road network over time and map individual choices into a macroscopic flow pattern. Li et al (2011) constructed a multi-day vehicle-based simulation framework to evaluate the response and benefits of traffic information provision under stochastic demand and capacity conditions.…”