An Alternative Framework to Capture the Day-To-Day Traffic Flow Evolution After an Unexpected Network Disruption

Abstract

In this study, we proposed an alternative discrete day-to-day dynamic traffic assignment model to capture the evolution of traffic flow after an unexpected network disruption. The model incorporates improved mechanisms for updating travelers' perceptions and learning processes, making it more effective in assessing day-to-day traffic dynamics under disruptive conditions—an aspect often overlooked in previous studies. The framework differs from traditional models in three key ways. First, it introduces a dynamic cost perception updating weight to adjust drivers' anticipation of traffic conditions during disruptions. This weight prioritizes predicted costs immediately after a disruption and gradually shifts to reflect up-to-date experiences as conditions stabilize. Second, the model employs the Exponential Moving Average (EMA) method for calculating perceived travel costs. Unlike conventional deterministic approaches, the EMA method accounts for travel inertia and emphasizes adaptive learning, allowing for smoother adjustments to disruptions. Third, it adopts a link-based network loading approach, avoiding the limitations of traditional path-based assignments. Numerical examples demonstrate the model’s flexibility and applicability. Applied to a simple test network, it effectively captured both short-term traffic fluctuations and long-term adaptations, highlighting its ability to handle disruptions and realistically model day-to-day traffic flow evolution.