2026-07-17 具身智能行业日报

Scaling Behavior Foundation Model for Humanoid Robots

arXiv Robotics · 机器人基础模型

In this work, we revisit the scaling recipe for BFMs and demonstrate that substantial performance gains can be achieved through the coordination of three core components: 1) the learning paradigm of motion tracking that reformulates diverse humanoid control problems as the reproduction of integrated whole-body behaviors in the global frame; 2) the strategic synergy between on-policy rollout quantity and reference motion diversity; and 3) the expressive and scalable model architecture termed Humanoid Transformer that facilitates the natural emergence of structured behavioral representations.

Hugging Face model lerobot/pi052_robocasa

Hugging Face LeRobot Models · 开源模型与数据

Tags: lerobot, safetensors, robotics, pi052, dataset:pepijn223/robocasa_pretrain_human300_v4_annotated5, license:apache-2.0, region:us

持续线索：第 2 次进入日报，首次出现 2026-07-16

CosFly-VLA: A Spatially Aware Vision-Language-Action Model for UAV Tracking

Stanford Robotics · VLA

Relative to OpenVLA, CosFly-VLA-0.8B reduces open-loop Average Displacement Error (ADE) by 34.1% on seen-test and 35.3% on unseen-test.

DriftWorld: Fast World Modeling through Drifting

arXiv Robotics · 行业动态

Rather than denoising iteratively at inference, DriftWorld learns an action-conditioned drift during training, allowing it to generate future frames from the current observation and a candidate action sequence in a single forward pass at 30+ fps, which is 17x faster on average than diffusion based baselines.

AeroAct: Action-Centered World-Action Models for Language-Conditioned Quadrotor Flight

NVIDIA · 行业动态

To obtain aligned visual, state, language, and dynamically feasible action data, we build a DiffAero-based pipeline with complementary Isaac Lab and 3D Gaussian splatting renderers.