Physical AI

RoboPrompt: Intuitive Robot Policy Steering with Sparse Human Input

End-to-end robot policies trained through imitation learning remain constrained by limited data diversity, making reliable zero-shot deployment in real-world settings challenging. Shared-autonomy methods enable human correction through teleoperation, but specialized hardware and operator training hinder deployment at scale. Other approaches incorporate human guidance as additional policy inputs, often requiring architectural changes and dedicated training for steerability, which limits their applicability across policies. We present RoboPrompt, a general-purpose, lightweight robot policy steering system that enables users to guide policy behavior through intuitive, sparse inputs, including drawn traces, target points, and coarse directional instructions. RoboPrompt decouples human-intention translation from the underlying policy: a reusable module converts human guidance into action drafts, which are refined through the diffusion or flow-matching dynamics of the base policy. By controlling action generation in noise space, RoboPrompt balances human intent with the policy prior without modifying the base policy architecture or fine-tuning it for steerability. Experiments demonstrate effective steering across Diffusion Policy, π_0.5, and FastWAM. We further use steered rollouts for online policy improvement through DAgger. After 2-3 rounds of iteration, average success rates increase by 15.5% for π_0.5 across three tasks and by 21.3% across three policies(Diffusion Policy, π_0.5, FastWAM) on the Insert Bread task, while average human intervention counts decrease by 44.0% (2.86 to 1.60) and 81.9% (2.60 to 0.47), respectively.

Yanwen Zou, Chenyang Shi, Guoxuan Xu

HuMBLE: Human Motion-Driven Behavior Learning for Embodied Locomotion

Despite recent advances in humanoid locomotion, controllers optimized for command tracking and robustness tend to produce mechanical gaits, whereas controllers tied to human motion data often fail to generalize to commands outside the data distribution. This work introduces a learning framework that balances these competing objectives to synthesize real-time steerable, robust, and biomimetic locomotion policies from human data. Using an in-house curated locomotion dataset covering diverse speeds and directions, we first learn a natural locomotion prior policy through a teacher-student distillation process. Specifically, we train a full-body reference-conditioned policy with Reinforcement Learning (RL), then distill it into a lightweight prior policy conditioned solely on proprioception and a planar torso-velocity steering command. Next, we fine-tune the prior policy with multi-task RL to expand command coverage and robustness beyond the data distribution, pairing a goal-conditioned task that tracks arbitrary commands with a reference-guided task that tracks the human data as an explicit style regularizer. We validate our framework on three humanoid robots: the Boston Dynamics Atlas R1, Atlas D1, and Unitree G1. Experimental results demonstrate robust performance across real-world scenarios, including direct user-controlled locomotion in indoor and outdoor environments, and integration as the locomotion layer within hierarchical control stacks. Benchmarks against Tabula Rasa RL policies trained without human data and ablation studies confirm that our framework yields a lightweight, deployable policy that reconstructs coordinated whole-body behavior from a steering command, retaining the human gait characteristics while remaining robust and fully steerable.

Mike Zhang, Dongho Kang, Kevin Bergamin

Four Next-Gen Robot Hands Compared: Atlas, Dex5-S, Torso 3, Wuji Hand 2

A side-by-side look at four newly released dexterous hands: the Boston Dynamics hand for the new Atlas, Unitree Dex5-S, Clone Robotics' Torso 3 hand and Wuji Hand 2. Four next-generation designs, four very different philosophies about how a humanoid hand should be actuated, sensed and built.

KingKong Robotics Open-Sources Jumper, a 22-DoF Crab Robot Toolkit

Beijing-based KingKong Robotics open-sourced the software toolkit for Jumper, its 22-DoF six-legged crab robot. Open the repo in an AI coding assistant and one-line prompts train new gaits in MuJoCo-based simulation, design .skin appearances or build .map scenes; trained policies export as .app bundles. The 1.8 kg robot runs on an RK3576 with 22 tactile servos, dToF depth sensing and a 5 MP camera.

AGIBOT A3 Ultra deployed as a salesperson at a car dealership

An official AGIBOT demo: the A3 Ultra humanoid robot works as a salesperson on the floor of a car dealership. The video comes from the official AGIBOT channel, whose title states the model is moving from mass production into real-world deployment.

Exploy: Open-Source Library Bridging Sim-to-Real for RL Deployment

RAI Institute introduces Exploy, an open-source library that packages entire control pipelines into a single ONNX graph, eliminating manual C++ rewrites and numerical discrepancies in sim-to-real RL deployment, validated on Spot, Unitree humanoids, and custom platforms.

PhoneBot: A Low-Cost Open Humanoid Robot Platform Reusing Smartphones

PhoneBot is a roughly 400-dollar open-source 13-DoF bipedal platform that reuses a commodity smartphone as its IMU, camera and onboard computer. A torque-aware DYNAMIXEL XL430 actuator model in MJX, combined with sagittal-mirrored PPO, yields balanced walking, standup and camera-heading policies that run entirely on old or prepaid Android phones.

Ruochen Hou, Quanyou Wang, Daniel Koh

EgoLAP: Learning from Egocentric Human Data through Language-Action Reasoning

Egocentric human data offer a path to scaling robot learning beyond costly robot demonstrations, yet the embodiment gap makes raw human trajectories a poor supervisory target for control. Our key insight is that, although low-level actions are embodiment-specific, their underlying motion intent can capture task-relevant structure that transfers across humans and robots. We introduce EgoLAP, a VLA pre-training framework that jointly learns from human and robot trajectories through a shared language-based action chain-of-thought. EgoLAP expresses motion intent as structured, temporally abstracted language actions and pairs them with motion-level reasoning grounded in scene geometry, physics, and object affordances. Across extensive real-world and simulated experiments, EgoLAP transfers human experience to robot control more effectively than alternative action representations and reaches 80.1% mean real-world task progress, a 2.3x performance gain over alternative action representations. Motion-level reasoning also outperforms a composite reasoning format that combines subtask, object-box, and visual-trace reasoning.

Lihan Zha, Shresth Grover, Tenny Yin

Real-Time Neural Rendering for World Models

A hands-on study of turning world state into pixels, comparing 3D Gaussian splats and neural renderers across fidelity, editability, latency and cost.

BRACE: Adapting Whole-Body References for Force and Terrain Aware Humanoid Motion Tracking

Whole-body tracking has become the interface through which operators drive humanoid robots, yet the references it consumes are recorded on level ground and carrying nothing, so the tracker knows neither the forces the robot must exchange with objects nor the terrain it stands on. Existing controllers address one side of the gap: force-capable policies command an end-effector force but prescribe no whole-body pose, while terrain-adaptive trackers treat loads as disturbances to reject rather than wrenches to command. BRACE folds both into the reference the tracker follows: terrain conformance lifts footholds and root onto the local surface, and a wrench transformation resolves the hand displacement that produces a commanded force together with the center-of-mass and center-of-pressure shifts it induces, bounded by teacher-specific arm-effort limits. Separate exertion and compensation teachers are distilled by DAgger into one flow-matching student that runs on proprioception alone, without a height map or measured wrench, so a teleoperator can supply a flat-ground trajectory while the robot resolves slope and load onboard. On the Unitree G1, in simulation and on hardware, BRACE cuts global tracking error by 40% on average and improves success rate by 18.25 points.

Sudarshan Harithas, Chen Yu, Juan Borbon

BrainCo Revo3 Dexterous Hand: 21 Active DoF, Full-Palm Tactile + Visuotactile Fingertips, Fully Direct-Drive

BrainCo (Hangzhou BCI company) launched its third-generation dexterous hand Revo3: 21 independent motors driving 21 fully active DoF (5 thumb + 4 per finger, replicating the human hand), 0.43 kg, 216 mm height; full-palm tactile array at 0.01 N resolution and fingertip visuotactile sensors detecting 130 μm deformation (V-Touch); fully direct-drive backdrivable joints (no tendons/linkages), 500 Hz comms, 70 N grip, 0.1° repeatability; MuJoCo/Isaac Gym/Omniverse one-click sim-to-real; official demos include cat's-cradle, spinning top and prayer beads. Tech transfer from its Time Best Inventions prosthetic hand heritage.

Stanford Real-Time EXPO-FT makes slow-but-smart VLAs real-time on robots: 42% to 97% with just 10 minutes of online data

New Stanford research (arXiv 2609.18207, Perry Dong, Kuo-Han Hung, Dorsa Sadigh, Chelsea Finn): large VLA models have high inference latency, so the world has moved on by the time their actions execute — and real robots do not pause for inference. Real-Time EXPO-FT splits control into two timescales: a slow asynchronous tier where the pretrained VLA proposes candidate action chunks using its strong behavior prior, and a fast synchronous tier where a lightweight edit policy applies bounded edits to actions at execution time using the latest observation, with a Q-value stage picking the best candidate chunk on the fly. The edit policy is trained with RL (built on EXPO/EXPO-FT; the Q signal touches only the edit, never backpropagating into the VLA backbone), with online robot data capped at 10 minutes and no human intervention. Results: average success 42% to 97% across four dynamic real-world tasks (object passing, ball balancing, table soccer kicking, dynamic object picking); best among delayed and non-delayed methods in 10/10 Kinetix environments. The inversion is notable: approaches like Astra have the slow smart model review the fast policy — here the fast lightweight policy edits the slow smart model. As LeoKharon put it: a big model's value may be its prior, not its control.