5arXiv cs.AI (Artificial Intelligence)·
PAC-MAN: Perception-Aware CBF-RL Framework for Whole-Body Humanoid Dodgeball with 95% Real-World Success
Researchers present PAC-MAN, a framework combining control barrier functions (CBF) with reinforcement learning for whole-body safety in humanoid robots, demonstrated via a dodgeball evasion task. The system uses only onboard depth camera with semantic segmentation, no privileged state information, and is deployed zero-shot on a Unitree G1 humanoid achieving 95% dodge success on real throws. The work advances perception-aware safety for humanoid locomotion by showing that fixed onboard cameras are adequate for reactive whole-body evasion when paired with appropriate barrier structure.
Related guides (1)
Related events (8)
Adaptive safety margins for visual robot navigation via context-conditioned safety critic
A new arXiv paper proposes a context-conditioned safety critic that learns adaptive clearance preferences for ranking diffusion-based trajectory proposals in indoor robot navigation. The critic decomposes into safety, efficiency, and distance-constraint matching terms, trained with privileged ESDF geometry in simulation and distilled into a perception-only selector via teacher-student learning. The method achieves top success rate and SPL on PointGoal navigation benchmarks (HM3D, MP3D) and transfers zero-shot to a Unitree G1 humanoid robot without task-specific tuning.
PAC-ACT: Reinforcement learning post-training framework for Action Chunking Transformer robot policies
PAC-ACT is a reinforcement learning post-training framework that fine-tunes pretrained Action Chunking Transformer (ACT) policies for precision industrial contact manipulation tasks. The method reformulates policy optimization at the chunk level, introduces an actor-critic architecture adapted for ACT, and uses a hybrid behavior-prior constraint to prevent distribution shift during online RL fine-tuning. Experiments on industrial contact benchmarks show significant improvements in task success, contact stability, and force safety — including a 46x reduction in force readings above 60 N on a contour-following task — while preserving low latency and GPU memory usage.
Mana framework achieves zero-shot sim-to-real transfer for dexterous articulated tool manipulation
Researchers introduce Mana (Manipulation Animator), a sim-to-real framework that reframes dexterous robotic manipulation as an animation problem using a coarse-to-fine pipeline of procedurally-generated grasp keyframes, motion planning, and reinforcement learning. The system requires minimal human input (under one minute per tool) and achieves zero-shot sim-to-real transfer across four articulated tools with varying joint types and scales. The work addresses a longstanding gap in dexterous robotics where articulated tool use—requiring coordination of internal degrees of freedom and contact-rich interactions—has been underexplored relative to rigid object manipulation.
OmniAct: Hierarchical framework for persistent embodied agents with unified cyber-physical action spaces
Researchers present OmniAct, a hierarchical asynchronous architecture for embodied agents that unifies cyber (APIs, IoT) and physical (manipulation, navigation) action spaces under a single multimodal semantic planner. The system incorporates adaptive hierarchical memory with event-boundary-driven compression to maintain sub-linear context growth, and an asynchronous visual preemption engine for closed-loop failure recovery during physical execution. Evaluated across 40 real-world long-horizon tasks on two robotic platforms coordinating four IoT devices, OmniAct achieves consistent end-to-end success improvements and elevates mid-scale open-weight models to proprietary-level performance. The work directly addresses the fragmentation between planning, memory, and verification in existing embodied agent systems.
DEED framework bridges lab-to-store gap for humanoid retail robots using data-efficient VLA post-training
Researchers present DEED (Data-Efficient Post-Training and Experience-Driven Learning), a systems-level framework for deploying Vision-Language-Action humanoid robots in real retail environments. The approach is evaluated on a supermarket chip-restocking task using a Unitree G1-Edu humanoid and NVIDIA's GR00T N1.6 foundation model, combining data-efficient post-training, experience-driven refinement adapted from RECAP, and a latent-space analysis tool. Key finding: bridging lab-to-real performance gaps is primarily a systems integration challenge rather than an architectural one, achievable with careful data design on a single GPU.
HANDOFF: Unified humanoid whole-body controller distilled from complementary specialist teachers
HANDOFF is a single whole-body controller for humanoid robots that uses a compact, explicit command-space interface bridging task planning and motor control. It is trained via multi-teacher KL distillation into a mixture-of-experts student from three specialists: whole-body motion tracking, locomotion, and fall-recovery. Evaluated on the Unitree G1, it matches state-of-the-art velocity tracking and demonstrates natural-language-driven task execution via a VLM-based agentic planner without task-specific fine-tuning. The work is relevant to the AI/robotics intersection as it shows a practical path to deploying language-driven agentic planners on physical humanoid hardware.
πR² enables reactive real-time replanning for robot manipulation flow policies
Researchers introduce πR², a method that makes large action-chunking flow policies reactive and real-time by splitting conditioning into fast proprioception and slow vision-language channels, plus a latency-adaptive denoising schedule. Applied to NVIDIA's GR00T-N1.7 on a real xArm6+XHand platform, it achieves ~25Hz closed-loop replanning (roughly 4× faster than the base policy), acting on fresh observations every 40ms. The approach improves task success rates by up to 23% in simulation and 30% in real-world manipulation over the strongest baseline, requiring only minimal architectural modification and fine-tuning from a pretrained checkpoint.
REGRIND: Minimalist retargeting-guided RL pipeline for dexterous manipulation from single human demo
Researchers introduce REGRIND, a pipeline that learns dexterous robot manipulation policies from a single human demonstration by retargeting hand-object motion to robot kinematic references and training a residual RL policy to track object-centric keypoints. The system achieves zero-shot sim-to-real transfer on multi-fingered hands for contact-rich tasks including scissors operation and screwdriver turning. The paper systematically analyzes key factors governing sim-to-real transfer in dexterous manipulation, extending a recipe previously validated for humanoid whole-body control into the harder contact-rich manipulation domain.