HOC2R introduces a hand-object contact trajectory extractor that combines frozen DINOv3 and SAM 2.1 features with learned heads to perform hand-object association and contact localization, extracting robot-usable motion cues from human videos.
PRISM represents reusable skills as parameter-level primitives, storing each skill family as a dedicated low-rank action-expert adapter, and combines and invokes them at inference via a block-wise VLM-conditioned router. On RoboCasa365, PRISM achieves a 56.43% atomic skill success rate and a 20.00% success rate on long-horizon composite tasks, compared with 17.86% and 17.50% for standard two-stage π0.5-LoRA fine-tuning.
The authors propose an object-centric reinforcement learning framework that treats each part as an agent and learns a shared policy for continuous assembly control. A graph neural network represents the parts, their target poses, and their spatial relationships, and predicts translational and angular velocities for each component; per-agent PPO with a centralized graph critic learns motion and temporal coordination within a single feedback policy. The study validates the approach on planar Tangram assembly, using procedurally generated target configurations and randomized initial layouts to support training and evaluation, and demonstrates parallel assembly in simulation and transfer to sequential execution on a real robot.
ForPC is a force-aware whole-body control framework that regulates a humanoid robot's cart-pushing speed through two-stage reinforcement learning. The method first infers force latent variables and state latent variables from onboard observations without external force sensors, then uses a residual policy to coordinate arm, waist, and leg movements; in simulation, cart-pushing speed RMSE is reduced by up to 68.2% compared with a locomotion-only control baseline, and lateral error on multi-curve routes is reduced by up to 74.2%.
CS-Pi proposes a "sample-project-renoise" flow-matching policy sampler that satisfies constraints absent from demonstrations at inference time while keeping actions plausible under the policy. The method projects each sample onto the nearest safe action, then adds noise back according to the size of the projection displacement and reruns the original policy until the sample is safe. On a 7-DoF shelf grasping task it achieves an 81% safe success rate, and 55% in zero-shot deployment on Franka, versus 55% and 40% for projection alone. The paper has been accepted to CoRL 2026.
The Meituan Robotics team has released Meituan-Robotics-0 (MR0), a vision-language-action foundation model and embodied-agent system for fixed-base desktop manipulation. Its training data includes 147 million vision-language samples and 15,060 hours of robot trajectories covering 12 embodiments.
The 18th installment of the Dexterous Hand Practice series discusses how to integrate impedance control and admittance control into the closed loop to avoid crushing objects during grasping. The content focuses on how these two control methods are integrated into the force-control closed loop of a dexterous hand.
The 19th installment of Gu Yue Ju's "Dexterous Hand Practice" series explains how to teach a robot using motion capture gloves, focusing on how teleoperation data is collected. The material provides only the title and summary; it does not give specific devices, procedures, or data metrics.
The 17th installment of the Dexterous Hand Practice series focuses on the differences between piezoresistive and capacitive tactile sensor approaches, and discusses how to use them to determine contact and slip. The content addresses the tactile perception stage in dexterous hand R&D and is part of this tutorial series.
The implementation path of dexterous hand force control from position loop and current loop to impedance control is systematically reviewed. The content is organized around these three control layers, explaining the role of each layer in dexterous hand force control and how they connect.
REEMAN completed on-site deployment of the Fly Boat factory delivery robot (AMR) at a large garment factory in Dongguan in about half a day, and configured multiple pickup and drop-off points. The robot has a rated payload of 300kg and actually carries about 200–250kg per trip (including a custom rack of about 80kg). It lifts the rack with a 10cm autonomous lift, navigates via laser SLAM, travels about 150m per loaded trip, and performs 70–90 tasks per day on average. Workers trigger deliveries via a physical one-button call button. Reference price is about USD 5,780, depending on configuration.
Kassow Robots, Joytek, Bosch Rexroth and Reply have jointly developed an automated kitting system for automotive manufacturing that combines a Joytek AMR carrying two Kassow KR1018 seven-axis collaborative robotic arms with a unified control architecture.
Episode 16 of Dexterous Hand Practice focuses on analyzing contact with grasp matrices and explains how to hold a cup steadily. The content centers on the grasp matrix as an analysis tool, used to assess the contact state and stability when a dexterous hand grasps a cup.
Tsinghua University and other institutions have proposed the ART (Agentic Robot with Tool-use) framework, which can fine-tune any end-to-end VLA model into an agent capable of plug-and-play use of external tools. The method uses 30,000 tool-call trajectories and two-stage LoRA fine-tuning, achieving average success rates of about 75% in LIBERO simulation and about 62% on a real Astribot S1 robot, an improvement of about 20% over baselines such as OpenVLA and π0.
Robot integration is enabling mid-sized manufacturers to achieve high-precision production with compact, low-cost solutions, without overhauling the entire production line. Torneria Novoletto uses the SCHUNK R-C2 automation concept, combining gripping components with robot loading and unloading.
SUPMAST brought its "Satongxue" AGV unmanned transport robot and in-plant smart logistics solution to the 22nd China International Grain and Oil Expo in Xiamen, held November 6 to 8, with live demonstrations of fork picking, stacking, and alignment correction at Booth C123 in Hall C1. The AGV uses laser SLAM + reflective pillars + IMU fusion navigation and requires no magnetic strip guide rails; the accompanying Yunshu S7 RDS scheduling system supports a maximum concurrency of ≥20 units and task allocation latency of ≤500ms. In the Yihai Grain and Oil project, it handles a production takt of 3 minutes per pallet, reducing the error rate by more than 95% and the cargo damage rate by 94%.
At CeMAT ASIA 2026, Hai Robotics is exhibiting the FlashPick system's double-deep solution and the K1200 heavy-duty latent lifting robot. Global signed cooperation for the FlashPick system has exceeded 20,000 units, with the largest single project surpassing 5,000 units. The double-deep solution can store 45,000 totes within 1,000 m² of space, with throughput of 4,000 totes per hour; the K1200 can collaborate with ACRs and supports unified scheduling of up to 6,000 robots on the same site.
Nuoyin Intelligence says KnowinAgent achieved a team self-tested composite score of 87.60 and an overall success rate of 82.83% across all 54 task variants in RoboDojo, once again ranking TOP1 on the leaderboard.
Yifei Technology, together with Korean distributor Jookang Robotech, showcased a complete palletizing solution comprising the Elephant collaborative palletizing all-in-one machine and Camel series AGVs at Robot World 2026, with a live demonstration of the full cycle of end-of-line carton palletizing and pallet and material transfer.
Boston Dynamics this week launched its second-generation autonomous mobile unloading robot, Stretch Gen 2, which can unload up to 1,000 boxes per hour, has a battery that can last two shifts, and can charge to 90% in about 75 minutes.
Boston Dynamics this week released the autonomous mobile unloading robot Stretch Gen 2, which can unload up to 1,000 boxes per hour, has a battery that can last two shifts, and fast-charges to 90% in about 75 minutes. The new machine reduces internal cabling by 81%, uses automotive-grade connectors and continuously rotating joints, and its gripper has independent valves and sensors for each suction cup, enabling it to grasp multiple boxes in a single motion. The company says software is updated monthly, and the most recent update increased the unloading rate of the in-service fleet by 24%; the complete machine has been deployed at multiple customer sites.
Changan Automobile announced that its first painting visual AI adhesive robot has officially gone live at the Liangjiang plant, with risk detection rate and reporting accuracy both exceeding 99.5%. The robot uses a large-model detection system built on a deep convolutional neural network (CNN) to lock in optimal adhesive application points and standardized operating trajectories, with single-inference response at the millisecond level. After clearing the PVC area bottleneck, line JPH increased by about 8%.
During Double 11, the Cainiao Zengcheng Tmall Supermarket warehouse in Guangzhou deployed 135 ZeeBot climbing robots to complete 28 hours of peak order fulfillment in a storage area with 9-meter-high, 20-tier racks and 42,000 bins, achieving a 99.99% on-time rate for wave outbound orders. The robots accumulated over 140 km of vertical travel, with a peak of 11 km in a single hour and a vertical climbing speed of 1.2 m/s; they attach to racks with dual grippers and require no dedicated tracks. Overall warehouse storage efficiency is about 3 times the original, labor efficiency has risen to about 2.5 times, and output per workstation equals 2 to 3 picking employees.
The Hong Kong University of Science and Technology, Beijing Academy of Artificial Intelligence, Peking University and other institutions jointly proposed DexRoam, which uses first-person human whole-body demonstrations to train mobile dual-arm dexterous manipulation policies. The work has been accepted to CoRL 2026. The system narrows the human-robot motion distribution gap through three stages: embodiment alignment, action semantic alignment, and temporal alignment, reducing MMD and SWD metrics by an average of 41.7% and 31.1%, respectively.
AgiBot launched a 24-hour convenience store work challenge in which 6 Genie G2 robots took on roles including reception, pickup and delivery, drink preparation, inspection and restocking, and cleaning, completing 1,730 tasks in total; in the end, 4 passed probation and 2 did not. The robots run on the embodied Agent framework Genie Master, which uses Physical RSI to turn success and failure experience from execution into long-term memory and iteratively refine action strategies; capabilities such as judgment for restocking and cleaning come from this prior self-review.
On November 3, Zhijian Dynamics and Hangcha Group completed the deployment of the first robot sorting line for a supermarket front warehouse, connecting the full workflow from AMR merge feeding, robot autonomous planning and dexterous grasping and turning, and scan verification to order distribution. The line uses the iX robot body that Zhijian Dynamics customized for Hangcha, and relies on visual perception and VLA task planning to handle boxed, bagged, bottled, and other product forms without the need to model each SKU separately. Measured single-machine sorting efficiency reaches up to 450 items per hour, with accuracy stable above 98%, and under full load it can support 24-hour continuous fault-free operation.
Why it matters: Readers can learn about the actual deployment conditions of the supermarket front-warehouse sorting line and the single-machine efficiency and accuracy data.
Six AGIBOT G2 wheeled humanoid robots completed a 24-hour trial at an operating convenience store in Xuhui District, Shanghai, performing a total of 1,730 tasks, with four of them passing the final assessment.
Why it matters: AGIBOT disclosed the task distribution and assessment results of six G2 robots over 24 hours in a real convenience store, offering a view of how embodied intelligence performs in continuous operation.
The REEMAN X-ant autonomous forklift at a large auto parts factory in Shanghai achieves about 125–140 pallet transport cycles per day, carrying 800–1,000 kg per trip over a full cycle of about 100 meters. The narrowest point of the plant's transport route is only 1.55 meters, and the pickup and drop-off area offers about 2.8 meters of operating space. With a one-button call, the forklift automatically picks up goods, transports them under load, drops them off, and returns to its charging station to stand by. The X-ant JBOT20F is 695 mm wide, has a turning radius of 1,448 mm, and a rated load of 2,000 kg.
German retail group Otto will deploy AI dual-arm robots in phases at a logistics center in Poland to inspect, identify, and sort e-commerce returns. The first phase plans to deploy 10 systems, and whether 15 more will be added later depends on initial operating results. Returned goods are often crumpled, repackaged, or presented in various orientations, so the robots need to combine visual information to decide how to grasp and adjust their actions based on feedback; simulation can model different orientations and packaging to assist training. Currently, the project still lacks enough public data to judge actual efficiency and return on investment.
Beijing Humanoid released the Hi-WBC general whole-body control model on the "Huisikaiwu" platform on November 2. The model combines human motion references with real physical constraints, enabling robots to adjust how they execute tasks as payload, external forces, and terrain change. It uses a continual skill expansion framework based on MoE and policy distillation, Contact-aware Learning that incorporates contact force prediction, and Perceptive Control that introduces terrain awareness.
Why it matters: The original text discloses Hi-WBC's technical approach to handling payload, external force, and terrain constraints beyond motion tracking, offering insight into how whole-body control is moving from demonstrations toward real work.
The research team proposes A2-MEAS, an embodied multi-agent framework that uses code as a scaling interface. It combines correction codes with corrective language, spatial evidence, execution history, and action proposals, while MetaBrain provides training supervision and runtime guidance.
Why it matters: The framework uses code as a scaling interface, injects corrective language and execution history into the policy, and presents comparison data across multiple benchmarks and real-robot experiments.
Ambi Robotics is installing new versions of its AmbiSort sortation system at 8 sites ahead of peak season and expects customer throughput to grow 9% to 10% year over year. The company sends about half its team to sites for inspections while the other half stands by to handle issues; it also pre-positions spare parts at sites in advance and pushes AI system improvements only after controlled-environment A/B testing. Its customer service hotline is answered by the same people responsible for on-site training and repairs, in order to build customer trust.
At CeMAT, Geek+ exhibited the general-purpose humanoid robot Gino 1, the RoboShuttle Hyper case-to-person system, and the SkyCube pallet-to-person system, and demonstrated a comparison of prediction and execution by its self-developed Gravity embodied model in clearing bins of mixed items and picking unfamiliar items.
LiberAI, founded by Liu Songming, has released Liber-0, a new-generation world action model that uses three expert streams for language, vision, and action along with Multimodal Attention to jointly predict future frames and actions.
Why it matters: The technical report uses controlled experiments to break down the contributions of human data, action supervision, and embodiment diversity to dexterous manipulation, which can inform judgments about data-mix strategies.
Sun Yat-sen University, Peng Cheng Laboratory, and other institutions propose RePO-VLA, which segments recovery trajectories, trains a progress-aware semantic value function, and performs value-conditioned fine-tuning so that dual-arm VLA policies can continue advancing tasks after disturbances. The study also builds FRBench, a failure-recovery benchmark; the RoboTwin simulation contains 46 dual-arm tasks and 23,453 trajectories. On real robots in water-pouring and towel-folding tasks, as recovery demonstrations per task increase from 50 to 200, the average disturbed success rates are 30%, 65%, and 75%, respectively.
Tristan, co-founder of YC F26-incubated company Asim, said the team is building a standardized benchmark for physical intelligence in humanoid robots, with a focus on evaluating dexterous manipulation capabilities. The benchmark uses repeatable scenarios to examine spatial reasoning, grasp adaptation, coordination, and how robots respond to physical interaction, with the goal of pinpointing where capabilities fail and providing reproducible failure cases. The team also noted the need to address comparability issues caused by differences in hardware, sensors, and control interfaces.
Geek+'s wheeled humanoid Gino 1 was released on February 10, featuring 14 degrees of freedom in its dual arms, three-fingered dexterous hands, a Jetson Thor compute platform, and a VLA fast-slow collaborative system; it can perform four types of work: picking, box handling, packing, and inspection. The article compares it with cases such as Agility Robotics' Digit and Robot Era's L7, pointing out that the key to humanoids entering warehouses lies in interfaces, data sources, and the maintenance response radius, rather than the form factor itself. Geek+ says Gino 1 already has mass-production capability and can connect to its own fleet formed by its scheduling system, picking stations, and AMRs for storage, picking, handling, and inspection.
Elite Robot and Shanghai Xijia Semiconductor jointly implemented an 8-inch silicon carbide substrate laser lift-off + thinning fully automated production line. The laser unit processes a single wafer in 20–25 minutes, with yield above 99.5%. A composite robot plus automation system handles autonomous wafer transfer and high-precision alignment; combined with laser modification and ultrasonic lift-off processes, it raises overall production efficiency by more than 30%. The solution uses chained automation to eliminate material transfer and positioning losses between process steps, and applies intelligent scheduling to unify process parameters and ensure line-wide consistency.
Guyueju released a tutorial explaining the complete workflow from URDF modeling to MoveIt2 grasping operations for a robotic arm plus dexterous hand. The material provides only the title and does not give specific steps, parameters, or code details.
REEMAN completed the on-site deployment of the Ironhide autonomous forklift in about one day at a lighting manufacturing factory in Zhejiang, for pallet transport between a covered outdoor storage area and an indoor warehouse. In actual operation, the forklift has a maximum payload of about 600 kg, a one-way route of about 100 meters, and performs about 140–180 transport tasks per day; the reference equipment price is about USD 9,900. During deployment, engineers also adjusted the transport route based on the actual situation of temporary inventory occupying the aisle.