Anticipatory collective human behavior: Path-seeking of pedestrians through a crowd  Hisashi Murakami , Claudio Feliciani; Yuta Nishiyama; Katsuhiro Nishinari (The University of Tokyo, Kobe, Japan)   C15

Collective animal groups often seem to behave as if having one body and/or one mind as a whole. Without external controls, a globally ordered state emerges from inter-individuals interactions. Human crowds also show various collective patterns similar to other animal groups such as bird flocks or fish schools. Most previous models describe pedestrian dynamics as a repulsive potential as in physical particle systems. Recent advances in image processing technique, however, have revealed that physics-inspired models do not completely explain the empirical observations. Particularly, it has been found that pedestrians anticipate future motions of their neighbors to seek their paths in dynamical pedestrian flow, unlike distance-dependent repulsive interactions in physical systems. This path-finding behavior results in deviation of pedestrians from the direct path to their destination. However, the deviations of individual movements from the desired directions in human crowds are poorly understood, though the path searching activities of pedestrians seem to influence collective-level behavior directly. Do individuals in human crowds exhibit conspicuous searching strategy? If so, how is such a strategy influenced by the ability of anticipation of motions of neighbors? Moreover, does it influence collective-level crowd dynamics? We here show that the deviation derived from the anticipation of motions of neighbors can be statistically described, and how it impacts on the global crowd dynamics, via two experiments with fifty-four participant pedestrians. First, we conducted an experiment of lane formation, a striking example of collective patterning in human crowds, where pedestrians waiting at each side of the experimental corridor begin walking towards the opposite side and then spontaneously separate into several lanes with uniform walking directions. We found that during lane formation, the fluctuation of pedestrians from their desired directions (i.e., the direct path to their destination) follows a scale-free movement strategy called a Levy walk process. The Levy walk, which is a special class of random walks in which many small steps are interspersed with occasional long steps, is theoretically optimal when searching unpredictably distributed resources. This result suggests that lane formation efficiently emerges through an optimal path-seeking strategy by anticipation. To check this suggestion, we conducted the second experiment which was similar to the first experiment; however, we asked some participants to perform walking while texting on the smartphone. Hence, these pedestrians were clearly distracted to anticipate the motions of neighbors. We found that the walking strategy of distracted pedestrians differs from non-distracted ones, influences on motions of other pedestrians as well, and decreases the degree of collective patterning (lane formation). These results imply that anticipatory behavior impacts on the global crowd patterns. Interestingly, also in other animal groups, anticipatory behavior is known to be crucial to perform robust collective behavior: individual noisy movements derived from anticipating or negotiating neighbors, which seem to collapse the global order, can facilitate interactions with various neighbors and contribute to dynamic collective behavior. Our findings suggest that anticipation is an underlying mechanism across a wide range of animal groups and is a key factor to develop the collective mind analogy.