Integrated Task and Motion Planning

Abstract

The problem of planning for a robot that operates in environments containing a large number of objects, taking actions to move itself through the world as well as to change the state of the objects, is known as task and motion planning (TAMP). TAMP problems contain elements of discrete task planning, discrete–continuous mathematical programming, and continuous motion planning and thus cannot be effectively addressed by any of these fields directly. In this article, we define a class of TAMP problems and survey algorithms for solving them, characterizing the solution methods in terms of their strategies for solving the continuous-space subproblems and their techniques for integrating the discrete and continuous components of the search. Expected final online publication date for the Annual Review of Control, Robotics, and Autonomous Systems, Volume 4 is May 2021. Please see http://www.annualreviews.org/page/journal/pubdates for revised estimates.

Learning compositional models of robot skills for task and motion planning

“…A planner must search over sequences of skills as well as parameter values that satisfy these constraints. Finding values that satisfy these constraints is a non-trivial problem; however, existing work in TAMP has shown that a variety of methods can be effective (Garrett et al, 2021). In this work we take a sampling-based approach using the PDDLStream TAMP planner (Garrett et al, 2020a), which we describe in Appendix B.…”

NeRP: Neural Rearrangement Planning for Unknown Objects

“…Perhaps the most relevant area of work to rearrangement planning is the field of task and motion planning (TAMP). TAMP is a broad area of study that looks at integrating discrete high-level planning (which objects to grab, which actions to execute) with continuous low level planning (which positions in which to place objects) [7]. This area generally relies on model-based methods, using known objects and fully-observable domains, although work has been done to weaken these assumptions, particularly in recent years.…”

“…Rearrangement planning is a common subset of task and motion planning [3,15,21,16,7]. It has recently been identified as an interesting challenge area for robotics research [1].…”

“…Rearrangement of unknown objects has been recently identified as a major challenge problem for embodied AI, especially robotics [1]. While many solutions exist that can solve these problems, particularly throughout the sub-field of task-andmotion planning (TAMP) [7], these often come with a wide 1 Please refer to our supplementary video: https://youtu.be/CJb1IzH94eo range of significant limitations that restrict their utility in the real world. In particular, these approaches often rely on having known object models.…”

Combining Task and Motion Planning: Challenges and Guidelines

“…This paper serves as a guideline for navigating the landscape of existing solutions and their caveats when designing combined task and motion planning methods. As opposed to the latest survey on this topic (Garrett et al, 2020), we here analyze a wider scope of concerns within TAMP and their combination; in particular, we discuss several orthogonal aspects, including uncertainty handling and online planning. Also, the discussion is centered on five open research questions and how answering them matters in a selection of industrially-relevant application contexts.…”