2026· IEEE Transactions on Automation Science and Engineering· Vol 23, pp. 19062-19072· 0 citations· 39 references
TL;DR
Virtual-Base Kinematic Optimization (VBKO), a body-assisted mobile manipulation framework that treats the executable body motion of a learned locomotion policy as a virtual base and allocates the desired end-effector velocity between this virtual base and the arm, is introduced.
Abstract
Legged mobile manipulators can substantially extend their manipulation capability by coordinating body motion with arm motion, but such coordination remains difficult when locomotion is executed by a learned policy. The main challenge is that the floating base is not directly actuated and must be realized through policy-executable commands, while manipulation control must still compensate for locomotion-induced body motion and contact disturbances. To resolve this, we introduce Virtual-Base Kinematic Optimization (VBKO), a body-assisted mobile manipulation framework that treats the executable body motion of a learned locomotion policy as a virtual base and allocates the desired end-effector velocity between this virtual base and the arm. The optimization incorporates command-boundary constraints and motion regularization, allowing the robot to use body assistance within reliable policy command ranges. We further propose Policy-Induced Gait Feedforward Compensation for arm torque realization, where the executed policy action is converted through the original locomotion interface into a gait feedforward term and used in an auxiliary inverse-dynamics problem for disturbance compensation, while the locomotion actuators remain controlled by the learned policy. Simulation and hardware experiments, including simulated welding-seam following, demonstrate that the proposed framework improves reachable manipulation range and end-effector tracking accuracy while maintaining stable policy-driven locomotion. Note to Practitioners—This work is motivated by the practical challenge of reusing high-performance legged locomotion policies for industrial mobile manipulation tasks, such as seam tracking, inspection, and object handling around large workpieces. While such policies robustly track body commands, their policy-level interfaces do not directly support body-assisted manipulation or compensation of locomotion-induced disturbances during industrial manipulation tasks. The proposed framework adds virtual-base kinematic optimization and policy-induced gait feedforward compensation on top of an available locomotion policy, facilitating modular debugging and controller replacement while reducing retraining through policy reuse.
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, providing a common end-effector interface for diverse manipulation tasks.
Snake robots can leverage their hyper-redundant morphology for both locomotion and manipulation through distributed body contact but whole-body loco-manipulation is particularly challenging since the entire body serves simultaneously as propulsive mechanism and manipulation tool Prior work on wheeled legged and snake r...
Adarsh Salagame, Henry Noyes, Haroon Hublikar et al.· 0 citations
Changing a humanoid robot's manipulation policy should not require retraining the controller that maintains the robot's balance. This study introduces DUET, a decoupled execution framework that enables locomotion and manipulation controllers to operate simultaneously by combining their commands only at the robot interf...
Kapalik Khanal, Biswash Khatiwada, S. Afrifa et al.· 0 citations
Continuous legged manipulation requires accurate end-effector tracking while the base keeps walking. Combining reinforcement learning (RL) with model predictive control (MPC) suits this task: the learned policy provides robust locomotion, while MPC coordinates the base and arm to compensate for tracking errors. However...
Kuan-Kuan Sima, Yi-Chao Gao, Chen-Xi Gu et al.· 0 citations
Humanoid robots performing lifting and material-handling tasks must maintain balance despite unknown and varying payloads. However, payload uncertainty significantly affects whole-body dynamics and may destabilize conventional controllers. We propose an adaptive whole-body control architecture that integrates three com...
Manh Cuong Nguyen, D. Vu, V. A. Vu et al.· 0 citations
OIL, an Object Interaction Layer that extends the kinematic command interface of whole-body control to include desired object motion and contact, is introduced and zero-shot generalization to novel behavior combinations and sim-to-real transfer on the Digit humanoid robot is demonstrated.
Pranay Dugar, Mohitvishnu S. Gadde, Rajesh Mangannavar et al.· Proceedings of the AAAI Symp...· 0 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.