Interactive Planning for Shepherd Motion.

摘要

Shepherding problem asks how the movement of an agent (e.g., predator, policeman, or sheepdog) can push a group of agents (e.g., prey, crowd, or sheep) from one position to another. The shepherding problem has many important and practical applications in security, environmental protection, agriculture, education and more. While much research has focused on simulating coordinated group behaviors, the shepherding problem has not received as much attention as it should be. This work investigate the possibility of addressing the shepherding problem using motion planning strategies. From an algorithmic point of view, the shepherding problem is extremely challenging due to its extremely large state space. It is clear that the existing motion planning methods cannot provide an efficient way to solve the problem. Instead, we propose an approach that incorporates computer-human interaction techniques with algorithmic robotics: an interactive motion planning method. In our experimental results, we have shown that this combination indeed provides performance improvement over the human-only and the computeronly approach. Although the proposed work shows promising initial results from our first prototype system, there are still a long way to go before we can have a better understanding of the interactive method for motion planning and the shepherding problem.