Comparison of Agent-Based Transit Assignment Procedure with Conventional Approaches

Abstract

The public transportation system, a key part of a multimodal transportation network, has been widely viewed as an efficient way to reduce road congestion and pollution. Public transportation planners use transit assignment models to forecast travel demand and service performance. As technologies evolve and smart transit systems become more prevalent, it is important that assignment models adapt to new policies, such as traveler information provision. This paper investigates three transit assignment tools that represent three approaches to modeling transit trip distribution over a network of fixed routes. These tools are the EMME/2 Transit Assignment Module (Module 5.35), commonly used by planners; Toronto, Canada, Transit Commission's transit assignment tool, MADITUC; and the newly developed Microsimulation Learning-based Approach to Transit Assignment (MILATRAS). These approaches range from aggregate, strategy-based frameworks to fully disaggregate microscopic platforms. MILATRAS presents a stochastic process approach (i.e., nonequilibrium based) for modeling within-day and day-to-day variations in the transit assignment process in which aggregate travel patterns can be extracted from individual choices. Although MILATRAS presents a different standpoint for analysis in comparison with equilibrium-based models, it still gives the steady state run loads. MILATRAS performs comparatively well with EMME/2 and MADITUC. In addition, MILATRAS presents a policy-sensitive platform for modeling the effects of smart transit system policies and technologies on passengers’ travel behavior (i.e., trip choices) and transit service performance.