Dealing with Difficult Instances of Object Rearrangement

Abstract

Rearranging multiple objects is a critical skill for robots so that they can effectively deal with clutter in human spaces. This is a challenging problem as it involves combinatorially large, continuous C-spaces involving multiple movable bodies and complex kinematic constraints. This work initially revisits an existing search-based approach, which solves monotone challenges, i.e., when objects need to be grasped only once so as to be rearranged. The first contribution is the extension of this technique to a method that addresses many non-monotone challenges. The second contribution is the use of either the monotone or of the new non-monotone method as a local planner in the context of a higher-level task planner that searches the space of object placements and which provides stronger guarantees. The paper aims to emphasize the benefit of using more powerful motion primitives in the context of task planning for object rearrangement than an individual pick-andplace. Experiments in simulation using a model of a Baxter robot arm show the capability of solving difficult instances of rearrangement problems and evaluate the methods in terms of success ratio, running time, scalability and path quality.

Object Rearrangement with Nested Nonprehensile Manipulation Actions

“…The proposed high-level NRS algorithm is given in Algorithm 1 and is an adaptation of an existing rearrangement planning algorithm [2], which itself is an adaptation of a NAMO approach for clearing obstructed paths [19]. The original algorithm of [2] is limited to grasping actions, while the current algorithm considers pushing actions. Moreover, the NRS algorithm also considers nested actions that are applied on sets of objects, as opposed to the mono-object manipulation actions used in [2].…”

“…The original algorithm of [2] is limited to grasping actions, while the current algorithm considers pushing actions. Moreover, the NRS algorithm also considers nested actions that are applied on sets of objects, as opposed to the mono-object manipulation actions used in [2]. On the other hand, the algorithm of [2] deals also with non-monotone instances, while we focus in this work on monotone instances for simplicity's sake.…”

“…Moreover, the NRS algorithm also considers nested actions that are applied on sets of objects, as opposed to the mono-object manipulation actions used in [2]. On the other hand, the algorithm of [2] deals also with non-monotone instances, while we focus in this work on monotone instances for simplicity's sake. The subroutines related to pushing and nested operations are presented in the subsequent sections.…”

Sample-Based Methods for Factored Task and Motion Planning

“…Planning problems in which the goal is not just to move the robot without collision but also to operate on the objects in the world have been addressed from the earliest days of motion planning to this day, for example [28,27,40,2,1,33,35,36,38,17,3,23,15,10]. In recent years, there have been a number of approaches to integrating discrete task planning and continuous motion planning [5,32,21,12,13,31,7,37], aimed at increasing the capabilities of autonomous robots.…”

Incremental Task and Motion Planning: A Constraint-Based Approach

“…Navigation among movable obstacles (NAMO) [21,54,73,74,77], Minimum Constraint Removal/Displacement (MCR/MCD) [34,35] find motion plans in the presence of movable obstacles. Rearrangement Planning bases the goal not on the robot's position but rather on positions of the objects to be moved [48]. In contrast to NAMO, MCR, MCD, and Rearrangement Planning, general TMP considers multiple, arbitrary task actions.…”