Abstract
Can the collective behavior of human crowds be explained as an emergent property of local pedestrian interactions? More specifically, is it possible that the coordinated movement of crowds emerges from hierarchical chains of leader-follower pairs (Nagy, Akos, Biro, & Vicsek 2010)? To address this issue, we collected data from 5 groups of 4 pedestrians steering toward a common goal. Participants began in a square configuration (0.5, 1.0, 1.5, or 2.5 m sides) and walked in a 12 x 12 m room while their head position and orientation were recorded using an inertial/ultrasonic tracking system (IS-900, 60 Hz). After walking forward for 1 m, they were verbally directed toward one of three goals located 8 m away. Each participant also walked alone at their preferred speed. In a preliminary analysis (Bonneaud, Rio, Chevaillier, & Warren 2010), we showed that participants adjust their individual walking speeds toward a common group speed. Here, we analyze these adjustments based on position in the group. We find that speed adjustments are more pronounced for pedestrians in the back of the group (R² = 0.71) than in the front (R² = 0.46). This implies both a side-to-side coupling [because pedestrians in front adjust their speed] as well as a front-to-back coupling [because adjustments are more pronounced in the back]. These results are supported by our data for pairs of pedestrians, which shows that a simple speed-matching model is sufficient to account for following (Rio, Rhea, & Warren VSS 2010) as well as side-by-side walking with a weaker coupling (Page & Warren VSS 2012). We model these local interactions to simulate the speed coordination among participants in the 4-person groups. In this way, by decomposing crowds into hierarchical pairs of pedestrians, we can begin to link individual locomotion with collective behavior.
Meeting abstract presented at VSS 2012