Corridor geometry effects on the bidirectional pedestrian walking modeled as a multi-agent system

We characterize how variations in corridor geometry affect bidirectional pedestrian dynamics modeled as a situated multi-agent system. Three environment configurations—corridors with side exits, crooked corridors, and patch-containing corridors—are compared against a standard straight corridor. These geometric modifications give rise to distinct collective patterns, revealing new features of self-organization and the system’s adaptive response to environmental constraints. We quantify the dynamics using four order parameters: average velocity, walking flow, velocity autocorrelation function, and informational entropy. Side exits enhance average velocity and autocorrelation while reducing entropy; crooked geometries increase flow, yield higher autocorrelation, and lower entropy; patch-containing corridors decrease both velocity and flow yet preserve autocorrelation. Furthermore, velocity–density, flow–density, and entropy–density relations exhibit clear phase transitions at critical densities. These findings provide a unified characterization of the effects of modified corridors, both on the walking dynamics and on the intrinsic decision making activity of agents in bidirectional pedestrian motion. • We design a MAS system to integrate both dynamics and behavioral activity of pedestrian walking. • Walking patterns are described by integrating four order parameters in different corridor geometries. • Crowd activity tends towards an entropy-autocorrelation balance to form self-organized patterns. • Fundamental diagram for velocity, flow and entropy present phase transitions in the same critical density region.