Life-cycle Modeling and the Walking Behavior of the Pedestrian-Group as an Emergent Agent: With Empirical Data on the Cohesion of the Group Formation

TL;DR

The paper investigates pedestrian groups as emergent agents in crowds and develops a dual-state-space framework comprising an agency life-cycle and a formation cohesion variable. It leverages naturalistic surveillance data and manual group annotations to identify state transitions and patterns in group radius, revealing that cohesive groups maintain a bounded radius while transient states induce cohesion disruptions and possible dispersal. The study documents concrete transition patterns (e.g., cohesive→loose→cohesive, cohesive→dispersing→forming→cohesive) and demonstrates their generality across multiple datasets. These findings provide a principled foundation for building more accurate group agents in simulations and for engineering systems designed to interact with pedestrian groups.

Abstract

This article investigates the pedestrian group as an emergent agent. The article explores empirical data to derive emergent agency and formation state spaces and outline recurring patterns of walking behavior. In this analysis, pedestrian trajectories extracted from surveillance videos are used along with manually annotated pedestrian group memberships. We conducted manual expert evaluation of observed groups, produced new manual annotations for relevant events pertaining to group behavior and extracted metrics relevant group formation. This information along with quantitative analysis was used to model the life-cycle and formation of the group agent. Those models give structure to expectations around walking behavior of groups; from pedestrian walking independently to the emergence of a collective intention where group members tended to maintain bounded distance between each other. Disturbances to this bounded distance often happened in association with changes in either their agency or their formation states. We summarized the patterns of behavior along with the sequences of state transitions into abstract patterns, which can aid in the development of more detailed group agents in simulation and in the design of engineering systems to interact with such groups.

Figures (8)

• Figure 1: (Left) An illustration of top-down vs bottom-up modeling of emergent agency and patterns of emergent behavior. (Right) diagram representing individuals as agents and the group as an agent, the intelligence as a unit of each, and relationships between them.
• Figure 2: An image representing a general scene of observations with extracted pedestrian trajectories and group labeling (trajectory color). This is a sample from Student003 dataset lerner2007crowds, with group annotation and illustration from solera2013structured. This figure is borrowed from solera2013structured.
• Figure 3: Sample trajectories extracted to represent a group observation (ego-group in blue) along with other pedestrian in the scene (in orange). Transparency highlights time.
• Figure 4: Group agency and formation state spaces. (A) An illustration for group members within each agency and formation state. (B) High level structure of group agency states and associated possible formation states. (C) Group formation states along with transitions and transition triggers.
• Figure 5: Group radius as a function of time measured for each group observed in ETH-Univ dataset. Radii for groups in different agency and formation states are isolated, as well as for each group size, to highlight patterns of behavior.