Steady State of Pedestrian Flow in Bottleneck Experiments

TL;DR

This paper introduces an improved algorithm to automatically detect steady states in pedestrian flow experiments, enabling more consistent comparisons across different initial conditions and experimental setups.

Contribution

An enhanced Cumulative Sum Control Chart algorithm is proposed for automatic steady state detection in pedestrian flow data, calibrated with an autoregressive model.

Findings

The flow in all states converges to steady state flow when pedestrian-to-bottleneck ratio exceeds 115.

The improved algorithm provides more reproducible steady state detection compared to manual methods.

Steady state detection facilitates better comparison of pedestrian flow experiments.

Abstract

Experiments with pedestrians could depend strongly on initial conditions. Comparisons of the results of such experiments require to distinguish carefully between transient state and steady state. In this work, a feasible algorithm - Cumulative Sum Control Chart - is proposed and improved to automatically detect steady states from density and speed time series of bottleneck experiments. The threshold of the detection parameter in the algorithm is calibrated using an autoregressive model. Comparing the detected steady states with previous manually selected ones, the modified algorithm gives more reproducible results. For the applications, three groups of bottleneck experiments are analysed and the steady states are detected. The study about pedestrian flow shows that the difference between the flows in all states and in steady state mainly depends on the ratio of pedestrian number to bottleneck width. When the ratio is higher than a critical value (approximately 115 persons/m), the flow in all states is almost identical with the flow in steady state. Thus we have more possibilities to compare the flows from different experiments, especially when the detection of steady states is difficult.