Validation of an agent-based microscopic pedestrian simulation model in a crowded pedestrian walking environment

Authored by Mohamed Hussein, Tarek Sayed

Date Published: 2019

DOI: 10.1080/03081060.2018.1541279

Sponsors: No sponsors listed

Platforms: No platforms listed

Model Documentation: Other Narrative

Model Code URLs: Model code not found

Abstract

This study validates a recently developed agent-based pedestrian micro-simulation model in a crowded walking environment. The model is applied to simulate pedestrian movements at a major street in the downtown Vancouver area. The street was closed for traffic to allow people attending a social event to leave the area safely. The calibration of model parameters is conducted using a Genetic Algorithm that minimizes the error between simulated and actual trajectories, acquired by means of computer vision. Validation results confirm the accuracy of the simulated trajectories, as the average error between the actual and simulated trajectories is found to be 0.28 m, and the average error in walking speed is just 0.06 m/s. Furthermore, results show that the model is capable of reproducing the actual behavior of pedestrians during different interactions with high accuracy (more than 94\% for most interactions).

Tags

Agent-based modeling Genetic algorithm Micro-simulation Pedestrians Risk Safety Calibration and validation Computer vision