A genetic algorithm for pedestrian origin-destination flow estimation

Abstract

Urbanenvironments are facing increasing complexity, mainly due to rising populations, vehicle usage, and tourism. This intensifies urban flows, including traffic, daily commuting, and tourist movements, posing challenges for congestion management, environmental sustainability, and overall quality of life. In popular tourist destinations, where urban infrastructure often limits carrying capacity, managing the combined flows of residents and visitors becomes particularly difficult. This can degrade both the visitor experience and the mobility of local communities. To address these challenges, mobility flows can be modeled using trajectory data, typically obtained via GPS. However, such data is costly and often unavailable. In this study, we leverage a more accessible but less granular dataset from the Lisboa Aberta initiative, which provides aggregated mobile device counts across Lisbon by location and time. While this dataset lacks individual identifiers or trajectory information, we aim to reconstruct pedestrian origin-destination (OD) flows from it. We frame this as a dynamic OD estimation problem, which is an inherently under-determined optimization task. To address this, we present a genetic algorithm constrained by a simplified pedestrian mobility model. This approach enables us to infer plausible mobility patterns from aggregated data, offering a cost-effective method for urban flow analysis and contributing to more sustainable urban planning.