RoboDuet: Whole-body Legged Loco-Manipulation with Cross-Embodiment Deployment


Guoping Pan* 1,4         Qingwei Ben* 1,3         Zhecheng Yuan1,2         Guangqi Jiang5        
Shoujie Li Ji1         Yandong Ji5         Jiangmiao Pang3         Houde Liu1,4         Huazhe Xu1,2,3        
1 Tsinghua University      2 Shanghai Qi Zhi Institute      3 Shanghai AI Lab      4 Jianghuai Advance Technology Center      5 UC San Diego
* Equal conttribution
arXiv Paper Code

Abstract

Fully leveraging the mobile manipulation capabilities of a quadruped robot equipped with a robotic arm is non-trivial, as it requires controlling all degrees of freedom (DoFs) of the quadruped robot to achieve effective whole-body coordination. In this letter, we propose a novel framework RoboDuet, which employs two collaborative policies to realize locomotion and manipulation simultaneously, achieving whole-body control through mutual interactions. Beyond enabling large-range 6D pose tracking for manipulation, we find that the two-policy framework supports cross-embodiment deployment, allowing for the use of different quadruped robots or various robotic arms. Our experiments demonstrate that RoboDuet achieves a 42.5% improvement in average success rate over the baseline in mobile manipulation tasks employing whole-body control. These policies also enable zero-shot deployment across different quadruped robots in the real world.

Method

Cooperative policy for whole-body control. RoboDuet consists of a loco policy for locomotion and an arm policy for manipulation. The two policies are harmonized as a whole-body controller. Specifically, the loco policy adjusts its actions accordingly by following instructions from the arm policy. The goal of the loco policy \( \pi_{loco} \) is to follow a target command \(\mathbf{c_t} \). The goal of the arm policy \( \pi_{arm} \) is to accurately track the 6-DoF pose. The actions of the arm policy consist of two parts: the first six actions \(a^{arm^J}_t \in \mathbb{R}^6\) represent the target joint position offsets corresponding to six arm joint actuators. The rest part of the arm policy \( a_t^{arm^G} \) is used to replace orientation commands, providing additional degrees of freedom for end-effector tracking to cooperate with the loco policy.

Pipeline Image
An overview of RoboDuet

Two stage training. In order to achieve both robust locomotion ability and flexible manipulation ability, we adopted a two-stage training strategy. Stage 1 focuses on obtaining the robust locomotion capability, which design is inspired by the powerful blind locomotion algorithm. Stage 2 aims to coordinate locomotion and manipulation to achieve whole-body large-range mobile manipulation, when the arm policy will be activated simultaneously with all the robotic arm joints.

Real Experiments

Whole body control

Given a series of challenging target poses, RoboDuet enables the legged robot to respond rapidly to the arm’s guide, adjusting body posture to coordinate and maximize proximity to the target. This demonstrates robust whole-body control capabilities.

Mobile Manipulation

Mobile Manipulation

To demonstrate the robot’s mobile manipulation capabilities, we tasked the robot with transporting objects across different heights, specifically targeting four representative levels: ground (0 cm), chair (20 cm), cabinet (60 cm), and standing desk (100 cm). Results indicate that RoboDuet effectively utilizes whole-body control to accomplish these mobile manipulation tasks

Ground
Chair
Cabinet
Standing Desk

Cross-Embodiment

To evaluate the zero-shot transferability across different quadruped robots, we directly deployed policies trained on the Go1+ARX5 configuration onto the Go2+ARX5, which has a 41.6% increase in base's weight. Using VR device to send identical target poses to both systems, results demonstrate that both configurations effectively exhibit agile 6D pose tracking and robust whole-body control capabilities.

More interesting cases

Whole-body control allows the robot to wield the hammer with greater striking force.

Agile 6D pose manipulation enables the robot to regrasp dropped objects.

Discrete Commands Following

Furthermore, by providing a series of discrete target poses during the quadruped robot’s movement, the system effectively adapts its posture to each target while maintaining stable locomotion throughout the whole process.

BibTeX

@misc{pan2024roboduetwholebodyleggedlocomanipulation,
                title={RoboDuet: Whole-body Legged Loco-Manipulation with Cross-Embodiment Deployment}, 
                author={Guoping Pan and Qingwei Ben and Zhecheng Yuan and Guangqi Jiang and Yandong Ji and Shoujie Li and Jiangmiao Pang and Houde Liu and Huazhe Xu},
                year={2024},
                eprint={2403.17367},
                archivePrefix={arXiv},
                primaryClass={cs.RO},
                url={https://arxiv.org/abs/2403.17367}, 
          }