Hierarchical task allocation in robotic exploration
John K. Hawley · RIT Scholar Works (Rochester Institute of Technology) · 2009
Autonomous robotic exploration has long been a topic of interest in robotics research. Robotic exploration promises the ability to explore otherwise unreachable or hostile environments. Autonomous exploration is particularly useful in distant or hostile environments in which real-time communication with a human controller may not be practical, such as deep sea or planetary exploration. In order to more effectively explore a large unknown area, multiple robots may be employed to work cooperatively. While cooperation among multiple robots allows for increased exploration potential, it also entails significantly more complex planning. This complex planning involves allocation of exploration tasks to the robots participating in the exploration. Task allocation for multi-agent systems has applications in a wide variety of fields, but specifically in robotics, it makes a level of autonomy possible that is difficult to achieve otherwise. Task allocation has been approached in a variety of ways, depending largely on the nature of the tasks considered. Some problems present very specific tasks, allowing task allocation algorithms for them to be very domain-specific. This thesis presents an analysis of various task allocation approaches that have been taken specifically for autonomous robotic exploration, and will present a new hierarchical market based approach. This new approach provides agents with a mechanism to form coalitions and to divide a coalition into smaller coalitions. The formation of new coalitions from larger coalitions to pursue multiple avenues of exploration forms an implicit hierarchy of goals as they are discovered.