Back to AI News
arxiv

PGX Framework Splits Long-Horizon Humanoid Loco-Manipulation Into State Machine, VLA Skills and World Locking

Researchers present Physically Grounded Execution (PGX), a framework that offloads temporal sequencing and precision control from a monolithic vision-language-action model to two training-free components. PGX divides long-horizon humanoid loco-manipulation into a predicate-based state machine for task progression and recovery, VLA skills confined to active manipulation phases, and world-coordinate locking for spatial consistency across base keeping, object-relative positioning and contact targets. Experiments cover three simulated tasks and one real-world task on a Unitree G1 humanoid, where the authors report gains in stage progression and task success rates. A project page accompanies the arXiv preprint.

AI Brief

It offers a concrete architectural recipe for making VLA models work on multi-stage humanoid loco-manipulation, a known weak point for end-to-end robot policies.

arXiv:2611.26596v1 Announce Type: cross Abstract: Robotic vision-language-action (VLA) models can execute diverse individual manipulation skills, yet they still struggle with long-horizon loco-manipulation tasks, where temporal structure, skills, and precision are tightly coupled within a monolithic model. We propose Physically Grounded Execution (PGX), a framework that offloads temporal and precision control to two classical training-free components, allowing the VLA to focus on stage-specific skill dynamics. PGX decomposes long-horizon loco-manipulation into three layers: (i) a predicate-based state machine governs task progression, using physical predicates to gate phase transitions, trigger recovery, and coordinate navigation and waiting. Shared predicate semantics connect offline demonstration segmentation, online execution, and stage-wise evaluation; (ii) VLA skills operate within active manipulation phases, with flat or stage-trained models concentrating on task-relevant actions; and (iii) world-coordinate locking (WCL) maintains spatial consistency through base station keeping, object-relative positioning, and world-anchored contact targets. We demonstrate the effectiveness of PGX with extensive experiments conducted on three simulated tasks and one real-world task using a Unitree G1 humanoid robot for long-horizon loco-manipulation, showing improvements in stage progression and task success rates. Project page is available at: https://kl-research.github.io/pgx
Be the first to rate this article.

Comments & Questions (0)

Captcha
Please be respectful — let's keep the conversation friendly.

No comments yet

Be the first to comment!