MS, Colorado State University, May 2004
Major Professor: Anthony A. Maciejewski
The study of crowd dynamics, or the large-scale motion of people, has an impact on a wide range of applications including architecture, pedestrian transport, escape planning, stampedes, event organization, and crowd control. The use of autonomous robots is a possible mechanism to improve pedestrian dynamics and modify crowd behavior by eliciting responses from people through instruction and social interaction. The goal is to modify the large-scale patterns of crowd dynamics through the small-scale interactions between people and the robots.
As a first step to achieving this goal, one must understand the range of possible human-robot interactions. The field of social robots gives insight into the various features, such as facial displays, gaze, and gestures, used by robots to interact with people. These features are explored by discussing the related published empirical data and its implications for crowd control robots. To study large-scale crowd behavior, a social force model was implemented and used to simulate pedestrian dynamics. The model consisted of many self-driven particles, i.e., people, which are influenced by internal drives, motivations, and physical forces. Each internal and external influence is realized by forces acting on the individual. Two pedestrian flow scenarios were simulated and studied: escape through a hallway with a widening and lane formation in bidirectional pedestrian flow. To determine the large-scale behavior, an average of how closely actual velocities matched desired velocities was measured to determine overall efficiency. Robots were then added to the simulation in order to improve efficiency through different maneuvers and social interactions.
Preliminary results from the simulations have shown that with the appropriate spatial responses efficiency can be improved for both scenarios. However, many challenges still remain, such as guaranteeing pedestrian safety and robustly dealing with potential malicious interference.