Singapore-based embodied intelligence startup Ropedia, together with Nanyang Technological University, Duke University, UC San Diego, and Tsinghua University, released the cross-embodiment mobile manipulation world model Ropedia W0, which uses a unified action representation to predict first-person future scenes after humans, robotic arms, and humanoid robots perform actions.
The HumanPlus lab at Xiamen University, together with researchers from MIT and CMU, has launched CLOTHO, a full-body motion capture system that integrates 10 IMUs into loose clothing. The paper was accepted to the SIGGRAPH Asia 2026 journal track and will be published in ACM Transactions on Graphics.
BAAI presented FlagOS 2.2 progress at the 2026 AI Open Computing Conference and FlagOS Technical Conference; the version was released on September 30. Public operators in the nine operator libraries rose from 564 to 1,454 (890 added). In tests covering over a thousand operators and 10 heterogeneous AI chips, cross-chip operator performance improved by about 27% over the previous version. The FlagOS-Compressor weight conversion and quantization tool was released for the first time, enabling a new model to be adapted to as many as 12 chips within one day.
Why it matters: The original text provides specific figures for operators, quantization, and serving throughput, which readers can use to assess the performance and cost potential of existing chips for running new models.
WorldOracle, a world-model evaluation platform jointly developed by researchers from AgiBot, NVIDIA, and several universities, is now live and open for evaluation submissions. The platform uses 335 real manipulation trajectories, with about 40% of interaction segments containing failed or suboptimal operations, testing model prediction against real execution outcomes. Testing of 7 world models found that while some models can reproduce robotic-arm motions relatively accurately, their predictions of object states still show considerable deviation, and performance drops markedly on long-horizon prediction.
X Square Robot has released X2Real, an evolvable simulation evaluation benchmark. Its simulated evaluation and real-robot evaluation achieve a task success rate correlation of 0.96 and a progress score correlation of 0.99.
Why it matters: It provides correlation data between simulated and real-robot evaluation along with a unified integration method, allowing readers to judge whether simulated evaluation can replace part of real-robot testing.
On the afternoon of October 19, 2026, the Healthcare World Model Session of the 2026 Physical AI and World Models Conference was held at the Nanjing International Expo Center, under the theme "Foreseeing Disease, Simulating Life — World Models Reshaping Medical AI," featuring eight keynote speeches and releases of results in six areas.
Daxiao Robotics released the embodied spatial perception model ACE-Brain-0.6, built on Qwen3.5-9B with a frozen vision encoder and supervised fine-tuning of the language model parameters. It surpasses GPT-6 Astra on six of eight Grounding benchmarks, leads with an average of 74.53 points across seven public benchmarks versus Astra's 73.68, and scores 94.13 on RobotGrounding versus 91.10.
On October 16, WorldOracle, an action-driven world-model evaluation platform jointly completed by researchers from AgiBot, NVIDIA, Princeton University, the University of Texas at Austin, and the University of Toronto, was officially released, opening evaluation submissions to developers.
Why it matters: The evaluation compares world-model predictions frame by frame against real execution records, revealing a common problem in which robotic arm motion is replicated but object states are systematically hallucinated, providing concrete reference points for judging model usability.
SUPCON held an industrial AI results conference in Ningbo, launching an industrial AI foundation platform centered on the TPT time-series large model and integrating it with the universal control system UCS to build the AOP autonomous operations factory system, connecting the perception-analysis-decision-execution chain. TPT and AOP have been replicated across 29 sub-sectors, and SUPCON signed industrial chain cooperation agreements with Huawei, Alibaba Cloud and others, while the industrial AI data alliance has more than 300 members. Companies including Wanhua Chemical and Zhenhai Refining & Chemical disclosed the results of TPT in early warning and optimized operations at chlor-alkali and refining units.
At the SIVA2026 Global Spatial Computing Industry Development Forum, PICO, QS Technology, and Pimax all discussed robot data collection, noting that the spatial positioning and motion tracking technologies VR has built up around immersive experiences are now being used for embodied intelligence data collection.
The winning bid for Fuzhou Survey Institute Co., Ltd.'s high-quality dataset construction project for a low-altitude drone perception network has been announced, with China Mobile Communications Group Fujian Co., Ltd. winning at 66.8 million yuan. The project runs until May 30, 2028; within 60 days of winning the bid, customized development of a data annotation system, a dataset management system, and a data value service system must be completed; within 2026, no fewer than 10 high-quality datasets and no fewer than 1 million image-text pairs in the low-altitude economy plus urban governance field must be produced, alongside construction of no fewer than 5 toolchains, and model adaptation and pilot applications must be carried out in no fewer than 5 typical scenarios.
Metro-WM proposes a hierarchical planning framework that, instead of generating unconstrained latent sub-goals, retrieves real states from observed frames in offline expert demonstrations or random-action trajectories as sub-goals, and connects frames from different episodes into paths toward the goal. Experiments show its long-horizon success rate is up to 37.33 percentage points higher than the next-best hierarchical method, up to 10.9 times faster, uses 13-56 times less offline computation, and requires less tuning. The authors also note that leading macro planners frequently output physically unrealizable sub-goals.
WorldOracle is a benchmark for action-conditioned world models (ACWMs) that uses identical initial observations and low-level action sequences to test whether predicted rollouts can reproduce real-world physical consequences and task-state transitions. The evaluation covers four dimensions: physical state correctness, interaction spatial alignment, key interaction area fidelity, and multi-view action coherence, and is paired with long, multi-segment real-world data containing both successful and failed executions. Experiments on seven representative ACWMs reveal different models' strengths and failure modes. The dataset, evaluation code, and leaderboard are publicly available.
The research team proposes DexBioLab, a simulation benchmark and framework for dexterous manipulation in real biological laboratory environments, comprising 14 tasks across 4 robot embodiments, covering functional tool use, precision manipulation, language-conditioned visual grounding, and instrument operation.
The CoRL 2026 paper proposes OpenDexGrasp for open-vocabulary task-oriented dexterous grasp generation, enabling a robot to infer functional intent from free-form language and, combining multi-view visual observations with object geometry, generate executable high-degree-of-freedom grasps.
The paper proposes RoboReel, a unified benchmark for evaluating models that learn policies from human videos (Learning from Observation), featuring real human demonstration videos, simulated robot trajectories, and evaluation environments for ten manipulation tasks.
ADEPT is a large-scale reinforcement learning framework that first pre-trains a dexterous policy on a generic object-reposing task, then post-trains downstream policies with this prior, enabling long-horizon tasks to be solved directly from raw visual-tactile perception across high-DoF embodiments. To prevent naive fine-tuning from degrading zero-shot reposing ability, the authors propose a stable post-training scheme combining behavior-cloning distillation, critic warm-up, and conservative on-policy updates, and introduce a joint-space Geometric Fabric that coordinates the policy with the robot.
This paper studies compositional generalization in sequential robot manipulation tasks from the perspective of instruction-space coverage, decomposing the generalization gap into three sources: marginal instruction shift, instruction-compositional shift, and context-action shift. Results show that exhaustive instruction-tuple enumeration is unnecessary; a structured subset covering action-relevant dependencies can achieve strong out-of-distribution performance. Sparse-training failures often stem from instruction steering rather than missing low-level skills, and fine-tuning with just one demonstration per task raises out-of-distribution success from 0.4% to 54.7%.
The paper proposes OptCar, a method that specializes a generalist forward kinodynamic (FKD) model into a vehicle-specific model while preserving cross-terrain generalization. At its core, a FiLM mechanism summarizes recent state-action history into a single dynamics context token that conditions multi-step predictions in a Transformer FKD architecture; specialization then combines limited real-vehicle data with targeted synthetic trajectories generated by system identification.
WireCraft is a simulation benchmark for industrial deformable linear object (DLO) manipulation, covering three task families: connector insertion, clip routing, and channel seating. It supports two physics models, articulated and deformable, and the released trajectories come from both simulation and a real UR5 robotic arm.
SidewalkBench is a benchmark for visual navigation on urban sidewalks built on NVIDIA Isaac Sim. It provides GPU-accelerated, high-fidelity sidewalk environments, including procedurally generated and real-world scanned scenes, and adds event-driven pedestrian reactive behaviors. The authors evaluated 9 visual navigation models on 330 unit-test scenarios, 800 pedestrian-reactive scenarios, and 105 long-horizon scenarios. They found that pedestrian interaction and long-horizon robustness remain key bottlenecks for existing models, and that scaling up sidewalk training with synthetic data is a viable direction.
OrbiSim proposes a robot simulation paradigm that recasts world models as a fully differentiable physics engine, connecting structured scene assets, neural dynamics, and downstream reinforcement learning. The entire simulation loop, from explicit state transitions to visual observation generation, supports end-to-end differentiability, enabling tasks such as differentiable contact modeling, gradient-based policy optimization under sparse rewards, and physical reasoning. Experiments show that OrbiSim significantly outperforms existing world models in prediction fidelity and control performance, and responds consistently to asset configurations and physical parameters.
PanoAir proposes a panoramic visual-inertial SLAM (VI-SLAM) framework and builds a real-world panoramic dataset covering different illumination, altitudes, trajectory lengths, and motion dynamics. The method uses panoramic feature extraction and panoramic loop closure to strengthen feature constraints and global consistency. On its own dataset and public benchmarks, it completed all 17 sequences without tracking failure and reduced ATE by up to 47.1% compared with the best baseline. Deployment on an embedded platform verified computational efficiency comparable to the PC implementation. The source code and dataset are publicly available.
HydroShear is a non-holonomic fluid-elastic tactile simulator accepted to RSS 2026, addressing tactile sim-to-real policy transfer for contact-intensive tasks. It models stick-slip transitions, path-dependent force and shear accumulation, and full SE(3) object-sensor interaction, and uses signed distance functions (SDFs) to track the displacement of indenter surface points as they interact with the sensor membrane, generating physically grounded and computationally efficient force fields from arbitrary watertight geometries.
Schrödinger's Navigator is a belief-aware framework for zero-shot object navigation (ZSON) that reasons at inference time over trajectory-conditioned imagined 3D futures generated from multiple candidate paths, rather than prematurely converging on a single map.
The authors propose the self-building laboratory (SBL) system, which can design, configure, and operate laboratory setups based on high-level scientific goals. Starting from an empty room or an existing self-driving laboratory, it fills in the laboratory layout, digital twin, robot skills, and experimental workflows. The system was demonstrated on a variety of robots and setups, and can scale from a single workstation to a laboratory with multiple robots working together. Its end-to-end capabilities were validated through a real-world quantum-dot synthesis experiment. The authors say this shifts the role of human researchers from manually configuring and programming device automation to specifying scientific goals.
The paper releases TriSAR-3D, a trimodal person detection dataset collected with a 128-channel LiDAR, an RGB camera, and a thermal camera in indoor and outdoor settings at different times of day, covering diverse poses including lying down; it also proposes a fusion method, PaintFusion.
The paper proposes EmbodiedMidtrain, which bridges the gap between vision-language models (VLMs) and vision-language-action models (VLAs) through mid-training. The authors first characterize the data distribution gap between the two, finding that VLA data occupy compact regions largely separated from the VLM distribution. They then build a mid-training data engine that uses a lightweight learnable proximity estimator to select the candidates most aligned with VLA from a large VLM pool, producing the EmbodiedMidtrain-1M dataset for mid-training VLMs.
This study trains blind assembly policies that observe only joint positions, velocities, and a short history of their own actions, without using visual, force, torque, or tactile sensing, and uses reinforcement learning in simulation to complete peg insertion, gear meshing, and nut tightening. The peg insertion policy transfers zero-shot to the real world and successfully inserts pegs of different shapes and sizes in 73% of trials; controlled experiments show that the blind policy generalizes better to unseen part shapes and sizes but may be slower.
HexaShell is a lightweight framework that lets coding agents control a dual-arm robot via the command line. It achieved a 39.7% success rate and a 48.0 progress score on all 42 RoboDojo tasks, a new best result for RoboDojo.
The research team achieved Baoding ball rotation on the Shadow Dexterous Hand Lite using only proprioception and binary tactile feedback, enabling zero-shot sim-to-real transfer. In a single continuous trial on the physical hand, it completed 225 180° rotations (112.5 full rotations) at about 0.36 rotations per second, with no vision or object-state information throughout. The authors also compared the roles of touch and pure proprioception, identifying the hand regions where tactile information is most critical.
The authors propose LAW, a closed-loop VLM-WAM framework that unifies language reasoning, action generation, and world modeling through semantic subtask progress, achieving an average success rate of 94.21% across 50 RoboTwin tasks.
PASS is a trustworthy precision approach simulation system for fixed-wing aircraft CAT III integrity algorithm development. Within a unified spatiotemporal geometric framework, it integrates configurable airport environments, prescribed trajectories, inertial simulation, calibrated stereo images, and continuous runway-relative ground truth, supporting navigation and integrity research under controllable sensor errors and environmental variations.
The paper proposes the MAP (Masked Action Space) interface, which unifies spatiotemporal references such as keypoints, flows, orientations, and key poses as partial constraints in the same temporal Cartesian action space, and models controllable manipulation as an action inpainting problem.
An arXiv paper proposes the Hardware-as-Regularization method, which uses a low-cost passive joint module (an adjustable mechanical clutch plus a 3D-printed spiral spring) to pull each joint toward a user-defined neutral posture, so that the passive restoring force acts as a physical prior for joint-space L2 regularization.
SenseSwarm runs a lightweight 709K-parameter map predictor on the swarm's shared occupancy map and folds the predicted structure back into a stigmergy-based guidance field, fine-tuning it on the swarm's own multi-front observations; a single-robot predictor fails on such views.
This work proposes a visuomotor imitation learning pipeline for pedicel-targeted strawberry harvesting that unifies perception, motion planning, and contact manipulation: data is collected from human teleoperation demonstrations, an Action Chunking with Transformers (ACT) policy with a ResNet-50 visual backbone is trained, and the policy is deployed autonomously on a Jetson AGX Orin edge platform at 15 Hz.
FCTS is a failure-conditioned terrain curriculum framework that uses localized traversal failures to automatically generate the terrain a policy should train on next. Each failure's location, termination mode, robot state, and surrounding terrain geometry jointly condition the generation process, and Monte Carlo Tree Search explores alternative curriculum trajectories under a fixed training budget.
Mothership proposes a two-tier autonomous refuse collection system in which an electric compaction truck carries two low-speed sidewalk robots, RoboCarts, that fetch and deliver standard two-wheel bins to the truck, return them to their positions, and re-dock for charging. A discrete-event simulation over synthetic campus layouts at 2-16 bins/ha quantified throughput: a two-cart configuration serves 20-29 bins/h, while three to four carts can reach 35-49 bins/h at an 80 m docking radius; the optimal docking radius must be chosen together with fleet size.
PhysSkin++ proposes a physics-informed framework that uses a Neural Skinning Fields Autoencoder to learn continuous neural skinning fields from static geometry, constructing a compact deformation subspace without per-object network training or simulation trajectories.