Incorporation of Memory Length Parameter in a Day To-Day Modelling Framework and Its Comparison With Stochastic User Equilibrium Model

Route choice modeling is a central component of transit assignment methods, playing a pivotal role in assessing how public transit travelers make decisions in the absence of external information. In reality, even identical travel experiences are perceived differently by different individuals due to variations in learning rate, risk-taking behavior, memory length, and other personal traits. This paper presents a simulation-based microscopic day-to-day (DTD) modeling framework that incorporates the memory length parameter alongside other stochastic elements in the route decision-making process. A congestion-sensitive cost function based on the Bureau of Public Roads (BPR) formulation is integrated into a logit-based route choice model. The day-to-day framework is developed for a dummy transit network and benchmarked against a conventional Stochastic User Equilibrium (SUE) model solved using the Method of Successive Averages (MSA). Sensitivity analyses are carried out with varying demand levels (100, 200, 240 passengers) and dispersion parameters (θ = 0.01, 0.05, 0.1, 0.5). Additionally, memory span tests are conducted in both laboratory and real-world field environments, and a regression model is developed to relate quantifiable personal traits (gender, age, income, education, frequency of transit use) to an individual's memory span. Key findings show that memory span is consistently lower in real-world settings than in controlled environments, that flow distributions in the DTD model are stochastic yet their averages converge to SUE flows, and that higher memory spans yield more stable route flows over time.