Whole-Body Dynamic Throwing with Legged Manipulators

TL;DR

This paper presents a deep reinforcement learning approach enabling legged robots to perform accurate, stable, and range-extended whole-body throws, leveraging full-body dynamics and adaptive curriculum for improved real-world applicability.

Contribution

It introduces a novel RL framework for full-body throwing in legged robots, emphasizing generalization to 3D targets and successful transfer from simulation to real robots.

Findings

Full-body throwing enhances range, accuracy, and stability.

Adaptive curriculum improves learning efficiency and balance.

Method generalizes to 3D target locations.

Abstract

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.