Physical AI

LEGS: Fine-Tuning Teleop-Free VLAs for Humanoid Loco-manipulation in an Embodied Gaussian Splatting World

Training vision-language-action (VLA) policies for humanoid loco-manipulation is constrained by the high cost and complexity of collecting human teleoperation demonstrations. VLA policies fine-tuned in simulators have, until now, failed to transfer effectively in humanoid loco-manipulation tasks. We present LEGS (Loco-manipulation via Embodied Gaussian Splatting), a hybrid simulator that composites a mesh foreground (robot, objects, props) over a photorealistic 3D Gaussian Splatting (3DGS) background reconstructed from a handheld scene capture. LEGS uses a procedural motion-primitive generator to synthesize labeled demonstrations at scale without human teleoperation, and a deterministic two-stage color calibration to align the rendered 3DGS image to the robot's deployment camera. On a Unitree G1 humanoid robot, across three pick-and-place tasks of increasing whole-body difficulty and three VLA backbones (psi_0, pi_0.5, GR00T N1.6), a policy trained purely on LEGS data matches or exceeds one trained on human teleoperation demos on every experiment. It also outperforms a mesh-only simulation baseline that ablates the effect of the 3DGS background, showing that photorealistic rendering is a key enabler for synthetic data transfer. Humanoid motion is recorded independently of scene appearance in LEGS, allowing the same auto-generated demonstrations to be re-rendered under new backgrounds and object meshes–covering a new scene at more than 15x lower cost than teleoperation–to augment training data for robustness to scene variations. Under combined object-and-scene appearance shift, the policy trained on re-rendered LEGS-AUG data maintains task success while the baseline trained on teleoperation data fails entirely. Our project page is located at https://legsvla.github.io/.

Kim, Hojune, Chen, Timothy, Sun, Jiankai

Learning Terrain-Aware Whole-Body Control for Perceptive Legged Loco-Manipulation

Legged manipulators integrate exceptional terrain adaptability along with mobile manipulation capabilities, which make them highly promising for deployment in human-centric environments. By coordinating the control of both legs and arms, a whole-body controller can significantly expand the operational workspace of legged manipulators. However, many existing whole-body controllers primarily depend on proprioception and do not incorporate the critical exteroception required for effective terrain topology perception. This limitation can hinder their ability to adapt to varying environmental conditions and navigate complex terrains effectively. In this paper, we introduce TA-WBC, a terrain-aware whole-body control framework for legged manipulators, which features a novel RL-based unified policy tailored to whole-body loco-manipulation tasks in various terrains. Specifically, we employ a hybrid exteroception encoder to extract terrain features, providing an essential basis for the robot to proactively adapt posture and footholds. Furthermore, to facilitate stable cross-terrain loco-manipulation, we propose a novel end-effector sampling method based on the foot contact plane, decoupling manipulation target from base fluctuations. Moreover, a dual-policy distillation module is introduced to integrate expansive whole-body motion with terrain adaptability without catastrophic forgetting. The simulation and real-world experiments validate the robustness of our proposed controller, which leads to a larger reachable space, less tracking error, and reduced unexpected stumbles. This unified policy highlights the promising capabilities of legged manipulators in performing loco-manipulation tasks across complex terrains.

Sikai Guo, Yudong Zhong, Guoyang Zhao

SSR: Scaling Surefooted and Symmetric Humanoid Traversal to the Open World

Extending humanoid traversal to the open world is key to practical deployment in human environments, but remains challenging. The robot must use vision to ensure safe and reliable foot placement on heterogeneous terrain under highly dynamic motion, while producing coordinated, natural whole-body behaviors. We propose SSR, an efficient end-to-end framework for egocentric vision-based humanoid traversal that jointly learns these capabilities. SSR introduces imagined foothold guidance, which learns to model forthcoming swing-foot contacts and evaluates their support to guide pre-touchdown swings toward stable regions, reducing edge slips. It further employs equivariant latent-space symmetry augmentation to efficiently induce bilateral coordination under high-dimensional visual observations, and uses terrain-specific multi-discriminator motion priors to encourage human-like behavior across scenes. Extensive experiments show that SSR achieves safe, stable, and high-quality locomotion on diverse real-world terrains, including stairs with varied structures and extreme challenges such as wide gaps and high platforms, while enabling reliable long-horizon traversal in open outdoor environments.

Ruiqi Yu, Yiwen Wang, Yuan Hao

Efficient On-policy Visual-RL via Stochastic Decoupled Policy Gradient

We present the stochastic decoupled policy gradient (SDPG), a lightweight visual reinforcement learning (RL) method that trains diverse visuomotor control policies end-to-end within a few hours on a single NVIDIA RTX 4080 GPU. SDPG estimates policy gradients via random perturbations of trajectory rollouts, requiring orders of magnitude fewer batch-rendered environments and substantially reducing compute and memory overhead. On visual MuJoCo benchmarks, SDPG consistently outperforms baseline methods in training time, memory usage, and rewards. Finally, to support future research, we introduce a suite of realistic visual robotics benchmarks spanning dexterous manipulation, challenging locomotion, and demonstrate effective sim-to-real transfer on physical hardware.

Haoxiang You, Yilang Liu, Davis Zong

Training a Humanoid Robot for Hard Work

Boston Dynamics shares how Atlas learns whole-body lifting via RL: from reference trajectories and reward design to GPU-scale simulation and real-hardware iteration, demonstrating fundamental building blocks of physical intelligence.

Scaling Behavior Foundation Model for Humanoid Robots

Researchers scale the Behavior Foundation Model to fully unleash humanoid robot manipulation potential.

Ouster REV8: Native Color LiDAR, Ruggedized and Maximized

REV8 unifies structure and color in a single sensor: the first native color lidar, able to read road signs, interpret brake lights and build survey-grade colorized maps. The OS family is reimagined across OS0, OS1 and OSDome, and gains a new 256-channel OS1 Max flagship with double the range and resolution of Rev7, engineered for functional safety and built for affordability at scale. Footage covers Yosemite Valley, a busy Chinatown intersection, a warehouse and a freeway drive.

FreeOcc: Training-Free Embodied Open-Vocabulary Occupancy Prediction

Existing occupancy prediction relies on large-scale voxel annotations and generalizes poorly across environments, collapsing to near-zero per-class scores on unseen scenes. FreeOcc is the first training-free framework for open-vocabulary occupancy prediction: a SLAM backbone supplies poses and sparse geometry, geometrically anchored Gaussian updates build a dense 3DGS map (ray-aligned anisotropic initialization with s_parallel = gamma * s_perp and Gaussian centers fixed to SLAM points), an off-the-shelf vision-language model attaches open-vocabulary semantics to Gaussian primitives, and a probabilistic Gaussian-to-occupancy projection yields a voxel field queryable online, with no learnable parameter serving the occupancy task. On EmbodiedOcc-ScanNet it reaches 31.29/13.86 monocular and 34.40/15.84 RGB-D IoU/mIoU, more than double the self-supervised baselines; geometric IoU improves 76.1% over the next-best DROID-Splat. The paper also introduces ReplicaOcc, where learned baselines collapse to 0.00-0.01 while FreeOcc transfers zero-shot to 55.65/20.90 and runs online at 25 FPS on a live RealSense D435i stream.

Jiang, Zeyu, Zhou, Changqing, Zuo, Xingxing

How to Build a Humanoid Robot at Home: The Asimov 1 Build Guide

Asimov opens up its $20k DIY humanoid kit: 25-DoF hardware, fabrication codes, CAN wiring traps, and the 200 Hz sim / 50 Hz asymmetric actor-critic behind zero-fine-tuning sim-to-real walking.

Neural Harmonic Textures: Real-Time Novel View Synthesis

Neural Harmonic Textures achieves real-time novel view synthesis, outperforming all 3DGS and NeRF derivatives including ZipNeRF in quality across all benchmarks. Code released (Apache 2.0).

Learning Athletic Humanoid Tennis Skills from Imperfect Human Motion Data

Human athletes demonstrate versatile and highly-dynamic tennis skills to successfully conduct competitive rallies with a high-speed tennis ball. However, reproducing such behaviors on humanoid robots is difficult, partially due to the lack of perfect humanoid action data or human kinematic motion data in tennis scenarios as reference. In this work, we propose LATENT, a system that Learns Athletic humanoid TEnnis skills from imperfect human motioN daTa. The imperfect human motion data consist only of motion fragments that capture the primitive skills used when playing tennis rather than precise and complete human-tennis motion sequences from real-world tennis matches, thereby significantly reducing the difficulty of data collection. Our key insight is that, despite being imperfect, such quasi-realistic data still provide priors about human primitive skills in tennis scenarios. With further correction and composition, we learn a humanoid policy that can consistently strike incoming balls under a wide range of conditions and return them to target locations, while preserving natural motion styles. We also propose a series of designs for robust sim-to-real transfer and deploy our policy on the Unitree G1 humanoid robot. Our method achieves surprising results in the real world and can stably sustain multi-shot rallies with human players. Project page: https://zzk273.github.io/LATENT/

Zhikai Zhang, Haofei Lu, Yunrui Lian

SteadyTray: Learning Object Balancing Tasks in Humanoid Tray Transport via Residual Reinforcement Learning

Stabilizing unsecured payloads against the inherent oscillations of dynamic bipedal locomotion remains a critical engineering bottleneck for humanoids in unstructured environments. To solve this, we introduce ReST-RL, a hierarchical reinforcement learning architecture that explicitly decouples locomotion from payload stabilization, evaluated via the SteadyTray benchmark. Rather than relying on monolithic end-to-end learning, our framework integrates a robust base locomotion policy with a dynamic residual module engineered to actively cancel gait-induced perturbations at the end-effector. This architectural separation ensures steady tray transport without degrading the underlying bipedal stability. In simulation, the residual design significantly outperforms end-to-end baselines in gait smoothness and orientation accuracy, achieving a 96.9% success rate in variable velocity tracking and 74.5% robustness against external force disturbances. Successfully deployed on the Unitree G1 humanoid hardware, this modular approach demonstrates highly reliable zero-shot sim-to-real generalization across various objects and external force disturbances.

Anlun Huang, Zhenyu Wu, Soofiyan Atar