Centralized and Distributed Algorithms for a Cooperative Carrying Task

Many useful tasks require both mobility and manipulation — one or more mobile robots must move to a given location and perform some task with an object there. Our research has focused on mobile robots with nonprehensile (i.e., nongrasping) manipulators. Such manipulators are mechanically simpler than their prehensile counterparts and have the additional advantage of being able to manipulate a wider variety of parts. In this paper, we present algorithms to enable two mobile robots equipped with two degree-of-freedom “palm” manipulators to cooperatively carry an object. As the robots move, the object may slip on the palms, so the robots must make adjustments, lest they drop the object. We first formulate a centralized algorithm that takes the form of a global policy. The correctness of this algorithm is demonstrated through two properties abstracted from the task mechanics. We then show how this centralized algorithm can be transformed into a distributed algorithm so that each robot can operate autonomously, communicating with the other robot only when necessary to ensure successful execution of the task. We have implemented these algorithms and present our experimental results.