Robotic Swarms are distributed systems whose members interact via local control laws to achieve different behaviors. Robotic swarms are becoming increasingly interesting for large scale applications such as environmental exploration, reconnaissance, surveillance and disaster response. These applications are characterized by uncertainty and potential changes of the environment and mission goals. Additionally, practical missions may require a combination of different swarm behaviors, where these behavioral combinations are not known a priori but could arise dynamically due to change in mission goals. Therefore, human interaction with the swarm (HIS) is needed. Behaviors of robotic swarms can be influenced by a human by (a) altering the behavior of some swarm members, called leaders, (b) altering the control laws that the individual swarm members use or (c) altering the environment in which the swarm operates. We are systematically investigating the effect of influencing the swarm through these three different ways. Our research goal is to understand the following key questions: (1) When does human influence benefit the overall system? (2) Is there a best time for a human operator to inject influence/input? (3) What are the effects of communication limitations (e.g. bandwidth, latency) on system performance in human controlled swarms and can any degradation in performance be mitigated? (4) How does the mismatch in operator understanding of swarm state and swarm member understanding of operator intent affect the performance of the overall system? (5) Are there environmental conditions that result in benefits vs detrimental effects of human control?

Our overall technical approach includes the development of algorithms for autonomous swarm coordination as well as development of models of human swarm interaction and validation of these models via human experimentation. In our experimental evaluation we use multi-robot simulations and a swarm of Turtlebots for environment exploration and search and rescue.