aiqbench analysis
This research advances bipedal loco-manipulation by introducing a unified reinforcement learning controller that autonomously coordinates locomotion, postural adaptation, and arm reaching based solely on end-effector targets. By integrating Transformer encoding with GRU memory for temporal context estimation, the system effectively balances dynamic trade-offs between tracking accuracy, movement stability, and object interaction without explicit low-level commands. Real-robot validation confirms the approach's robustness across diverse inputs, including VR teleoperation and diffusion policies, marking a significant step toward versatile mobile manipulators.
Learning Holistic Whole-Body Loco-Manipulation with a Bipedal Mobile Manipulator
arXiv:2609.18930v1 Announce Type: new Abstract: Bipedal loco-manipulation enables robots to interact with objects beyond the nominal workspace of their arms by coordinating locomotion and manipulation. Realizing this capability requires a low-level whole-body controller that translates task-level manipulation goals into coordinated arm and leg motions while maintaining balance. We present a unified whole-body controller trained with reinforcement learning that directly maps 6-DoF end-effector targets to coordinated actions for the bipedal base and robotic arm. Given only an end-effector target, the learned controller autonomously coordinates reaching, postural adaptation, and stepping without explicit base-velocity or footstep commands. A reward-gating strategy regulates the trade-offs among end-effector tracking, locomotion, and balance during training, while a temporal context estimator combines windowed Transformer encoding, recurrent GRU memory, and auxiliary dynamics prediction to extract dynamics-relevant information from observation history. Real-robot experiments demonstrate that the same controller supports reaching, postural adaptation, and stepping under commands from VR teleoperation, a learned diffusion policy, and scripted trajectories, providing a common end-effector interface for diverse manipulation tasks.
Abstract:Bipedal loco-manipulation enables robots to interact with objects beyond the nominal workspace of their arms by coordinating locomotion and manipulation. Realizing this capability requires a low-level whole-body controller that translates task-level manipulation goals into coordinated arm and leg motions while maintaining balance. We present a unified whole-body controller trained with reinforcement learning that directly maps 6-DoF end-effector targets to coordinated actions for the bipedal base and robotic arm. Given only an end-effector target, the learned controller autonomously coordinates reaching, postural adaptation, and stepping without explicit base-velocity or footstep commands. A reward-gating strategy regulates the trade-offs among end-effector tracking, locomotion, and balance during training, while a temporal context estimator combines windowed Transformer encoding, recurrent GRU memory, and auxiliary dynamics prediction to extract dynamics-relevant information from observation history. Real-robot experiments demonstrate that the same controller supports reaching, postural adaptation, and stepping under commands from VR teleoperation, a learned diffusion policy, and scripted trajectories, providing a common end-effector interface for diverse manipulation tasks.
Robotics (cs.RO)
arXiv:2609.18930 [cs.RO]
(or arXiv:2609.18930v1 [cs.RO] for this version)
https://doi.org/10.48550/arXiv.2609.18930 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Full text under the publisher's open licence · 253 words
Related coverage
CMP: Robust Whole-Body Tracking for Loco-Manipulation via Competence Manifold Projection
arXiv · Oct 15, 2026
Towards a General Humanoid Loco-Manipulation Model via Egocentric Whole-Body Human Data Pretraining
arXiv · Oct 2, 2026
ViLoMan: Learning Visual-Proprioceptive Whole-Body Loco-Manipulation Skills for Humanoid Robots
arXiv · Sep 18, 2026
Learning Terrain-Aware Whole-Body Control for Perceptive Legged Loco-Manipulation
arXiv · Sep 4, 2026