Hierarchical Distributed Task Assignment for UAV Teams
David W. Casbeer, Matthew E. Argyle · Infotech@Aerospace 2011 · 2011
In this paper, we discuss four impediments to distributed UAV architectures: poor UAV to UAV communication, inadequate processing capability (applies to both computation and sensors), operators’ desire for a continuous stream of video and images, and pilot safety in airspace with both unmanned and manned aircraft. These impediments motivate centralized architectures for multi-UAV planning and coordination. The key problem stems from the need for a human to continuously monitor the UAVs’ data streams. We discuss a hierarchical distributed task assignment algorithm, called the team consensus based bundle algorithm, that keeps these four impediments in check. Furthermore, a multi-socket manager (MSM) is presented, which was designed with these four impediments in mind. The MSM is an interface that allows testing of hierarchical and distributed planning architectures that keep these fours factors in check, e.g., the team consensus based bundle algorithm. Currently deployed unmanned air vehicles (UAV) are controlled by a group of operators that actively manage different components of the UAV. One operator might control the motion of the vehicle, either directly through “stick-and-rudder” inputs or through a waypoint navigation mode. Simultaneously, a different group of operators is remotely controlling the onboard sensors and evaluating any received data. Recent research has been focused on inverting this ratio of operators per UAV to have a single operator manage a team of UAVs. 1 Planning algorithms have been developed to aid the operator with the management of a team of UAVs, and current research is improving on these algorithms. 2‐4 To date, planners are generally implemented