Learning Symbolic Operators for Task and Motion Planning (2021-02-28T00:00:00.000000Z)

TL;DR

This work proposes a bottom-up relational learning method for operator learning and shows how the learned operators can be used for planning in a TAMP system, finding this approach to substantially outperform several baselines, including three graph neural network-based model-free approaches from the recent literature.

Abstract

Robotic planning problems in hybrid state and action spaces can be solved by integrated task and motion planners (TAMP) that handle the complex interaction between motion-level decisions and task-level plan feasibility. TAMP approaches rely on domain-specific symbolic operators to guide the task-level search, making planning efficient. In this work, we formalize and study the problem of operator learning for TAMP. Central to this study is the view that operators define a lossy abstraction of the transition model of a domain. We then propose a bottom-up relational learning method for operator learning and show how the learned operators can be used for planning in a TAMP system. Experimentally, we provide results in three domains, including long-horizon robotic planning tasks. We find our approach to substantially outperform several baselines, including three graph neural network-based model-free approaches from the recent literature. Video: https://youtu.be/iVfpX9BpBRo. Code: https://git.io/JCT0g

Authors

L. Kaelbling

12 papers

J. Tenenbaum

25 papers

Tomas Lozano-Perez

9 papers

TL;DR

Abstract

Authors

References32 items

SPOTTER: Extending Symbolic Planning Operators through Targeted Reinforcement Learning

Hierarchical Planning for Long-Horizon Manipulation with Geometric and Symbolic Scene Graphs

Integrated Task and Motion Planning

Inductive logic programming at 30: a new introduction

Deep Visual Reasoning: Learning to Predict Action Sequences for Task and Motion Planning from an Initial Scene Image

Creative Problem Solving by Robots Using Action Primitive Discovery

A review of learning planning action models

Learning sparse relational transition models

Learning Quickly to Plan Quickly Using Modular Meta-Learning

Learning STRIPS Action Models with Classical Planning

Relational inductive biases, deep learning, and graph networks

Active Model Learning and Diverse Action Sampling for Task and Motion Planning

PDDLStream: Integrating Symbolic Planners and Blackbox Samplers via Optimistic Adaptive Planning

Learning to guide task and motion planning using score-space representation

Incremental Task and Motion Planning: A Constraint-Based Approach

Guided search for task and motion plans using learned heuristics

Logic-Geometric Programming: An Optimization-Based Approach to Combined Task and Motion Planning

Combined task and motion planning through an extensible planner-independent interface layer

Inductive Logic Programming

Acquiring planning domain models using LOCM

Active Learning of Relational Action Models

Hierarchical task and motion planning in the now

FF-Replan: A Baseline for Probabilistic Planning

Angelic Semantics for High-Level Actions

Learning Symbolic Models of Stochastic Domains

PDDL2.1: An Extension to PDDL for Expressing Temporal Planning Domains

Planning as heuristic search

Inductive Learning of Reactive Action Models

Learning value functions with relational state representations for guiding task-and-motion planning

PyBullet, a python module for physics simulation for games, robotics and machine learning

PPDDL1.0: The language for the probabilistic part of IPC-4

aSyMov: A Planner That Deals with Intricate Symbolic and Geometric Problems

Field of Study

Journal Information

Name

Page

Venue Information

Name

Type

URL

Alternate Names