Continuous agent-based modeling of adult-child pairs based on a pseudo-energy: Relevance for public safety and egress efficiency

TL;DR

This paper extends an agent-based crowd model to include adult-child pairs, capturing their interactions and safety implications during evacuations and flow, with potential applications in public safety design.

Contribution

The authors introduce a pairing potential into a crowd simulation model to accurately represent adult-child pairs and their effects on crowd dynamics.

Findings

Wide corridors reduce flow friction during evacuations.

Tightly held pairs improve safety and flow efficiency.

Loose pairs are more likely to split at high densities.

Abstract

Pushes, falls, stampedes, and crushes are safety hazards that emerge from the collective motion of crowds, but might be avoided by better design and guidance. While pedestrian dynamics are now getting better understood on the whole, complex heterogeneous flows involvinge.g. adult-child pairs, though widely found at e.g. crowded Chinese training schools, still defy the current understanding and capabilities of crowd simulation models. We substantially extend a recent agent-based model in which each agent's choice of motion results from the minimization ofa sum of intuitive contributions, in order to integrate adult-child pairs. This is achieved by adding a suitably defined pairing potential. The resulting model captures the relative positions of pair members in a quantitative fashion, as confirmed by small-scale controlled experiments, and alsosucceeds in describing collision avoidance between pairs. The model is used to simulate mixed adult-child flows at a T-junction and test the sensitivity to the design and pairing strategies. Simulation shows that making the post-confluence corridor wide enough is critical to avoidfriction in the flow, and that tight hand-holding is advisable for safer evacuations (whereas more loosely bound pairs get split at high density) and, more marginally, more efficient egresses in normal conditions.