HACMan++ targets long-horizon manipulation for six-dimensional object rearrangement, decomposing actions into three parts: selecting a primitive, anchoring a placement point on the point cloud, and continuous parameters, with unified decision-making via a per-point, per-primitive evaluation map. On this basis, the framework achieves closed-loop chaining of pushing, grasping, moving, and placing.
Based on project experience over the past two years, Hikrobot has upgraded its STRP case storage and picking solution across three dimensions: division of labor, flexibility, and throughput, and introduced the HAPS intelligent picking system. The STRP dual-side solution places tracks on racks on both sides of the aisle, achieving both increased storage capacity and higher task completion rates with 30% fewer T50 units; the industry-first cross-aisle STU connects adjacent aisles with dedicated curved tracks.
REEMAN has published a deployment case at a chemical shipping company in Shanghai, where the Ironhide autonomous forklift links pallet pickup, automatic stretch wrapping, elevator floor transfer, and indoor and outdoor delivery, covering about 600 kg per trip, about 650 meters round trip, about 12 minutes per cycle, and about 50 cycles per day. The forklift exchanges station-ready and wrapping-complete status with the wrapper through an industrial interface, and coordinates requests, entry, and exit with the building elevator control system. During commissioning, the team adjusted drive-wheel downforce to resolve ramp slipping, and worked with the wrapper manufacturer to correct the pallet handover process.
A team led by Cheng Long at the Institute of Automation, Chinese Academy of Sciences, published research at IROS 2026 using tactile-enhanced reinforcement learning to complete sub-millimeter precision insertion, raising the success rate from 52% for a tactile-free approach to 67% under a 0.05 mm clearance.
Yingzhi XBOT announced the consecutive completion of a Series A round of over 200 million yuan and a Series B round of nearly 400 million yuan. Its coffee robot stores have been deployed in over 1,600 locations worldwide, making over 5 million cups of coffee cumulatively. The device occupies less than 2 square meters, houses a coffee machine, ice maker, milk frother, and a six-axis robotic arm, produces a cup of coffee in 46 seconds, and its price has dropped from 700,000 to 800,000 yuan four years ago to over 100,000 yuan. Founder Tang Mu said orders on hand for 2026 are counted at 300 million yuan, with recent monthly deliveries exceeding 100 units and monthly revenue exceeding 20 million yuan.
At its AI DAY on October 29, Critical Point announced that cumulative production of its dexterous hands has surpassed 15,000 units, and launched its self-developed data-collection glove, OmniGlove Pro. The glove uses a fusion of EMF magnetic field positioning and IMU, achieves fingertip positioning accuracy of 0.1 mm, is equipped with 746 tactile sensing points, and has a tactile spatial resolution of 3 mm. The company also introduced the "real-machine Ready" concept, spanning from data collection to real-machine replay.
The authors present MUSBRIC, a modular hardware platform and accompanying MuJoCo simulator for reconfigurable musculoskeletal robots, which uses LEGO-compatible endcaps to turn fiber-reinforced extensile pneumatic artificial muscles into reusable, plug-and-play muscle modules. An elbow-joint prototype validated load-dependent flexion and extension, with pulley-guided routing enabling reversible motion and a mechanical stop stabilizing the straight posture; a robotic hand demonstrated compliant grasping and multi-joint construction, 15 leg configurations completed matched sweeps in hardware and simulation, and a quadruped assembled from four legs achieved walking. The hardware design and simulation code will be open-sourced.
LOGIV is a task-graph framework for long-horizon robot tasks that connects symbolic planning, observation-based verification, and embodied execution. During execution it checks the physical validity of condition-support and resource-ordering edges, and after a failure it uses dependency tracing to determine which parts must be restored.
SimRetract proposes a policy trained only on simulated demonstrations that lets a robotic arm take a retractor from a surgeon's hand and maintain retraction at handoff, achieving zero-shot sim-to-real transfer for surgeon-robot collaboration in open surgery.
The research team released TacClaw-Infinity, a large-scale structured Vision-Tactile-Language-Action (VTLA) dataset for contact-rich manipulation, spanning over 10,000 hours, 1 million episodes, and 1 billion frames across more than 1,000 task types.
Why it matters: This dataset provides a collection interface and quality control pipeline for ten-thousand-hour-scale tactile manipulation data, and its cross-embodiment transfer validation approach can serve as a reference.
The research team proposes I3L, a hardware and software system that combines whole-body correction with iterative policy refinement to mitigate the out-of-distribution degradation of offline imitation learning policies in long-horizon whole-body mobile manipulation.
UniCross proposes a unified framework that models four dexterous manipulation skills—grasping, relocation, in-hand rotation, and in-hand translation—as the same relational-target formulation, and distills a single policy that does not depend on skill identifiers. The policy performs well on each individual skill, generalizes to unseen objects, and supports long-horizon skill chaining and execution of relational targets unseen during training; the framework also applies to different hand morphologies.
The authors propose DR-LfD, a framework that integrates visuomotor policies into a TAMP-gated decision-making system, decomposes human demonstrations into atomic skills based on contact relationships, and then reproduces and reorganizes them using visuomotor policies or object-centric primitives. This method shifts the demonstration data requirement from growing exponentially with the skill sequence to growing with the number of skill types, and on two DexMimicGen tasks it outperforms a diffusion imitation policy trained on 1,000 demonstrations while using only 100 demonstrations. The paper has been accepted by T-RO.
The authors propose SPARK (Sequential Planning via Anchored Robotic Keypoints), which describes tasks as a symbolic behavior tree, anchoring to the current 3D observation only when each action is invoked; it verifies the plan at runtime and re-anchors objects on local failure, requesting a new plan only if this step fails.
NDPP-Grasp proposes injecting physical plausibility constraints directly into the denoising process of a task-aligned grasp diffusion model, so that guidance can be applied throughout generation even when the constraints are non-differentiable. Existing diffusion methods first train a grasp diffusion model for task alignment and then apply physical plausibility correction after generation, leaving the generation trajectory itself unguided by physical constraints. The work was accepted to NeurIPS 2026.
The authors propose the TacO framework to systematically evaluate tactile sensors based on manipulation policy performance, covering four modalities: visual, acoustic, magnetic, and resistive. They train separate policies for three tasks—pick-and-place with unknown mass, object reorientation, and plug insertion—and analyze how sensor properties such as spatial resolution, shear sensing, tactile representation, and material friction affect task performance. The results show that the value of tactile information depends on the sensor modality, material properties, and specific manipulation task, rather than being universally beneficial; the sensors, code, data, and hardware designs will be made public on the project website.
FlowTouch is a view-invariant visuo-tactile prediction model that uses an object's local 3D mesh to encode information for predicting tactile patterns, enabling tactile feedback even before the object is touched. The method combines scene reconstruction with a Flow Matching-based image generation model. Results show that it narrows the sim-to-real gap and generalizes to new sensor instances, and the generated tactile images can also be used for downstream grasp stability prediction. Code, dataset, and videos are publicly available.
The authors propose π-StepNFT, an online reinforcement learning framework for streaming VLAs that treats transitions in SDE sampling as supervision units and uses variance-normalized transition errors to distinguish positive and negative velocity branches, without computing the final-action likelihood.
EquiContact proposes a hierarchical framework consisting of a high-level vision planner, Diff-EDF, and a low-level compliant visuomotor policy, G-CompACT, for learning contact-rich manipulation policies that generalize across task configuration spaces.
MotorSense is the first large-scale paired egocentric video and surface electromyography (EMG) dataset, containing synchronized video, bimanual EMG, IMU, and audio across 100 hours of unscripted manipulation. The research uses this data for cross-modal pre-training between video and EMG to learn an action organization shared across visual and motor spaces, and transfers it to tasks including action recognition, 3D hand reconstruction, and 3D contact reconstruction. Across four action-related visual tasks and eight video datasets, this sensorimotor-aligned representation outperforms conventional video self-supervised methods.
RoboArchitect lets a coding model observe robot rollouts, diagnose failures, and iteratively revise the agent harness around a frozen policy. Without updating any model weights, it raises average success from 9.7% to 43.3% (4.5x) on unseen scenes and from 12.7% to 40.2% (3.2x) on unseen tasks.
This paper proposes a separable catenary model for suspended-fabric Robotic Manipulation Surfaces (RMS) that can quickly compute the surface shape for given actuator heights, as well as the actuator heights needed to achieve a desired local slope. The closed-form inverse solution lets a simple PID controller position objects on the fabric surface, and the model relates fabric length, module span, and actuator stroke to the reachable workspace, yielding a lower bound on minimum fabric length and design charts.
The study proposes language-conditioned behavioral cloning for autonomous endoscope guidance, defining 6 clinically relevant guidance behaviors and collecting 235 subtask demonstrations. In preliminary phantom experiments, a single visuomotor policy achieved 100% success on focus, overview, and stop, 80% on search-tip and follow-tip, and 50% on horizon leveling. The authors note that the failure modes indicate more complex guidance tasks require temporal context.
The study proposes an online method for learning closed-loop dynamics under a frozen locomotion policy, using a contaminated Gaussian model to distinguish predictable from unpredictable responses and combining a Gaussian process with random Fourier features to build an online learner that reduces bias from heterogeneous observations. Combined with model predictive control (MPC), the method predicts floating-base motion more accurately; in simulation and hardware experiments, end-effector tracking error was 0.01 m, and a finite-sample error bound for the online learning is provided.
SketchMate has a vision-language model build a sketch model for insertion tasks, consisting of a few geometric primitives and two mating coordinate frames, which a planner and compliant controller then execute. In a strict kinematic test of 69 simulated insertion episodes, the method inserted 88% of parts at the assembled pose, while ReKep and MOKA using the same VLM achieved no successful insertions.
ResMerge proposes a two-stage framework that learns mergeable residual policies atop a frozen pre-trained policy: residual reinforcement learning obtains task-specific corrections during adaptation without additional expert demonstrations, while parameter-space merging consolidates compatible residuals to reduce stored parameters.
The authors propose the Direct Manifold Action Transformer (DMAT), which reframes bimanual action generation as directly modeling clean, observation-conditioned joint-space action chunks from stochastic interpolants, enabling single-step inference without iterative denoising or trajectory straightening.
RTPolicy is a multi-view visuomotor policy that uses view-aware feature fusion and visual compression to aggregate multi-camera information into a compact token representation, and uses a flow-matching action generator to reduce action-generation overhead. On the 50 tasks of the RoboTwin 2.0 single-task clean evaluation, it achieved an average success rate of 52.26%, higher than the strongest leaderboard baseline π0 at 46.4%.
BioAgent is an embodied agent for long-horizon laboratory protocols that acquires new skills by matching human-reviewed requirements with validated skills, selecting tool-based implementations for missing behaviors, and requesting human demonstrations as needed. On the second task, it reused 17 skills unchanged, modified 2 existing implementations, and added 5 new skill types; on two dual-arm laboratory tasks with 20 and 31 ordered skill invocations respectively, it achieved complete-task success rates of 66% and 58%, with each task run 50 times and autonomous recovery enabled.
HandMimicX proposes an object-centric framework for learning dexterous manipulation skills from egocentric human RGB-D videos. The method defines the reinforcement learning objective in terms of object motion, decoupling skills from specific hand trajectories, and imposes constraints on hand-object interaction states during grasping and manipulation to obtain more anatomically plausible hand configurations. Simulation and real-robot experiments show that the method learns efficiently from human videos and transfers skills to novel spatial configurations.
The authors propose LEXP, a framework for learning from physical experience through context-space adaptation that combines task-level planning with an experience module, reusing execution experience to generate contextual feedback while model weights and the action interface remain fixed. Across 12 real-world manipulation tasks, LEXP achieved 23/36 successes (63.9%), compared with 19/36 (52.8%) using memory alone and 2/36 (5.6%) for vanilla ChatGPT-6-Astra.
TinyAstral proposes a two-stage method: first freeze an open-weight teacher model and optimize its harness through recursive self-improvement (RSI), then distill a 4B student model to replace the teacher within the same harness. On RoboTwin, harness evolution raised the teacher's success rate from 33.4% to 63.6%, and the final 4B student reached 62.8%; in measured deployment episodes, the student used about one-fifth of the teacher's accelerator time.
The authors propose the Contact Imitation Interface (CIMI) framework, which uses contact geometry and fingertip RGB views to build a robot-free human demonstration interface and transfers skills to a robot hand through contact-alignment-first retargeting.
The research team used a five-fingered robotic hand to grasp and insert RJ45 connectors, evaluating 281 imitation-learning policies spanning seven policy families in approximately 1,700 real-robot trials. Controlled ablations showed that observation and action representations had a greater impact than conventional optimizer and sampling choices. For point-cloud policies, region-of-interest cropping, representation in the robot-base frame, and normalization within the point encoder each improved success by 70 to 100 percentage points. For the vision-language-action policy, switching to absolute action prediction improved insertion success by up to 70 percentage points.
Why it matters: Approximately 1,700 real-robot trials compared 281 imitation-learning policies, indicating that observation and action representations determine real-world success more than optimizer and sampling choices.
CaPtain proposes an open post-training pipeline that trains small open vision-language models (VLMs) into code-as-policy (CaP) robot agents, without relying on VLAs, frontier models, or privileged state at test time. The pipeline first performs supervised fine-tuning on synthetic trajectories generated by GPT-6 Astra, then applies multi-turn reinforcement learning, introducing a physics critic that reads simulator privileged state to assign credit to each request.
A joint team from Fudan University's Institute of Trustworthy Embodied Intelligence and Shanghai Xinzhi Embodied Intelligence Technology Co., Ltd. won first place in the quality inspection category at the 2026 International Embodied Intelligence Skills Competition. The student team, working with the National and Local Center, used the Linglong H wheeled-arm robot and the Linglong L bipedal robot to win two first places, one third place, and one fourth place across three categories: loading and unloading transport, firefighting, and precision assembly. In the firefighting category, the Linglong L bipedal robot completed fire detection, fire suppression, and evacuation without human intervention.
Xiaomi has unveiled a bionic hand with 21 degrees of freedom and 1,500 tactile points across the full palm. The palm weighs 324 grams and is driven by 26 motors, with tendon-driven actuation as the main approach plus 5 direct-drive degrees of freedom. The team selected 1,827 of the 2,579 standard work procedures across 16 work sections at the pilot center of Xiaomi Auto's intelligent manufacturing department, collected multi-angle video and motion-capture data to work backward to the joint configuration, and says that in simulation the average stable-grasp success rate in the 17-to-21 degree-of-freedom range was 15% to 38% higher than in the 6-to-15 range, with gains slowing markedly after 21.
KAIST AI and Holiday Robotics propose a robot-centric pointmap that transforms RGB-D observations into the robot coordinate frame while preserving the H×W grid, and adds it element-wise to RGB tokens to inject it into a pretrained VLA.
After 16 months of R&D, Hongqi Chain has terminated development of its robot unmanned retail system. The company's announcement said the system has not yet met the standard for large-scale commercial deployment, and it is shifting to a 24-hour cloud-staffed operation business. An industry insider close to Hongqi said that in summer, condensation on glass-bottle cold drinks made the robot unable to grip them steadily, and it kept dropping and shattering them; it also could not reliably pick up goods in complex packaging, and operation was even harder when aisles were narrow and customers were nearby. In a robot retail warehouse visited by QbitAI, the robot only handled mobile picking, with each pick taking about 20-plus seconds, and human staff were still needed for restocking and charging. Several practitioners believe robot retail is still in the PoC stage.
Shanghai has proposed deploying 8,000 humanoid robots into factories by 2028, while Jiangsu, Shenzhen and other regions are also advancing real-scenario training and production-line deployment; companies including SAIC and CATL have disclosed application progress. Most deployed cases are concentrated in decomposable processes such as cell grabbing and loading, test loading and unloading, and material handling and sorting. For example, AgiBot's "Nengzai No. 1" works at about one unit every 2 seconds on Buick's battery mass-production line, and Galbot S1 has entered CATL's production line to handle material handling and sorting. The article notes that entering factories still requires passing hurdles in stability, battery life and cost accounting, and cross-factory replication also requires reducing technology adaptation and system integration costs.