Improving Operator Recognition and Prediction of Emergent Swarm Behaviors

Robot swarms are typically defined as large teams of coordinating robots that interact with each other on a local scale. The control laws that dictate these interactions are often designed to produce emergent global behaviors useful for robot teams, such as aggregating at a single location or moving between locations as a group. These behaviors are called emergent because they arise from the local rules governing each robot as they interact with neighbors and the environment. No single robot is aware of the global behavior yet they all take part in it, which allows for a robustness that is difficult to achieve with explicitly-defined global plans. Now that hardware and algorithms for swarms have progressed enough to allow for their use outside the laboratory, new research is focused on how operators can control them. Recent work has introduced new paradigms for imparting an operator's intent on the swarm, yet little work has focused on how to better visualize the swarm to improve operator prediction and control of swarm states. The goal of this dissertation is to investigate how to present the limited data from a swarm to an operator so as to maximize their understanding of the current behavior and swarm state in general. This dissertation develops--through user studies--new methods of displaying the state of a swarm that improve a user's ability to recognize, predict, and control emergent behaviors. The general conclusion is that how summary information about the swarm is displayed has a significant impact on the ability of users to interact with the swarm, and that future work should focus on the properties unique to swarms when developing visualizations for human-swarm interaction tasks.