Whole-Body Dynamic Throwing with Legged Manipulators

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Munn, Humphrey, Tidd, Brendan, Böhm, Peter, Gallagher, Marcus, Howard, David
Format: Preprint
Published: 2024
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866912302056341504
author Munn, Humphrey
Tidd, Brendan
Böhm, Peter
Gallagher, Marcus
Howard, David
author_facet Munn, Humphrey
Tidd, Brendan
Böhm, Peter
Gallagher, Marcus
Howard, David
contents Throwing with a legged robot involves precise coordination of object manipulation and locomotion - crucial for advanced real-world interactions. Most research focuses on either manipulation or locomotion, with minimal exploration of tasks requiring both. This work investigates leveraging all available motors (full-body) over arm-only throwing in legged manipulators. We frame the task as a deep reinforcement learning (RL) objective, optimising throwing accuracy towards any user-commanded target destination and the robot's stability. Evaluations on a humanoid and an armed quadruped in simulation show that full-body throwing improves range, accuracy, and stability by exploiting body momentum, counter-balancing, and full-body dynamics. We introduce an optimised adaptive curriculum to balance throwing accuracy and stability, along with a tailored RL environment setup for efficient learning in sparse-reward conditions. Unlike prior work, our approach generalises to targets in 3D space. We transfer our learned controllers from simulation to a real humanoid platform.
format Preprint
id arxiv_https___arxiv_org_abs_2410_05681
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Whole-Body Dynamic Throwing with Legged Manipulators
Munn, Humphrey
Tidd, Brendan
Böhm, Peter
Gallagher, Marcus
Howard, David
Robotics
Throwing with a legged robot involves precise coordination of object manipulation and locomotion - crucial for advanced real-world interactions. Most research focuses on either manipulation or locomotion, with minimal exploration of tasks requiring both. This work investigates leveraging all available motors (full-body) over arm-only throwing in legged manipulators. We frame the task as a deep reinforcement learning (RL) objective, optimising throwing accuracy towards any user-commanded target destination and the robot's stability. Evaluations on a humanoid and an armed quadruped in simulation show that full-body throwing improves range, accuracy, and stability by exploiting body momentum, counter-balancing, and full-body dynamics. We introduce an optimised adaptive curriculum to balance throwing accuracy and stability, along with a tailored RL environment setup for efficient learning in sparse-reward conditions. Unlike prior work, our approach generalises to targets in 3D space. We transfer our learned controllers from simulation to a real humanoid platform.
title Whole-Body Dynamic Throwing with Legged Manipulators
topic Robotics
url https://arxiv.org/abs/2410.05681