Manipulation with diverse actions

Abstract

We define the Diverse Action Manipulation (DAMA) problem in which we are given a mobile robot, a set of movable objects, and a set of diverse, possibly non-prehensile manipulation actions, and the objective is to find a sequence of actions that moves each of the objects to a goal configuration. We argue that classic sampling-based techniques cannot solve DAMA problems because of the need to move through lower-dimensional subspaces, and we give two sampling-based algorithms for this problem, DARRT and DARRTCONNECT, based on the RRT and RRTCONNECT algorithms respectively. We also show that the DAMA problem can be framed as a multi-modal planning problem [14] and describe a hierarchical algorithm, DARRTH (CONNECT), that takes advantage of this multi-modal nature. This algorithm finds a high-level sequence of transfer manipulations by planning a path only for objects in the domain. It then attempts to achieve each transfer manipulation individually. We present experimental results for all four algorithms for a set of nine problems in two complicated mobile manipulation domains. We show that the bi-directional algorithms are faster than their forward search counterparts and that the hierarchical algorithms perform better than the monolithic searches. We also formally define the conditions under which DARRT is exponentially convergent and prove that these conditions hold for two example manipulation domains, one of which includes nonprehensile manipulation.